WO2007058950A2

WO2007058950A2 - Biological interface system with neural signal classification systems and methods

Info

Publication number: WO2007058950A2
Application number: PCT/US2006/043815
Authority: WO
Inventors: Daniel J. Sebald; Almut Branner; Kirk F. Korver; Andras Pungor; J. Christopher Flaherty
Original assignee: Cyberkinetics Neurotechnology Systems, Inc.
Priority date: 2005-11-10
Filing date: 2006-11-09
Publication date: 2007-05-24
Also published as: WO2007058950A3

Abstract

Systems and methods for neural signal classification for the processing of multicellular signals of a patient are disclosed. The system includes a preprocessing device operatively coupled to an input channel and configured to receive multicellular signals collected from a sensor, at least a portion of the sensor configured to be disposed within the brain of a patient. The preprocessing device is also configured to filter the multicellular signals to extract a neural signal portion of the multicellular signals, the neural signal portion including a neural spike portion and a local field potential portion. The system also includes a neural spike processing device operatively coupled to the preprocessing device. The system is configured to project information associated with a neural spike onto a feature space, the feature space indicative of a correlation of the neural spike with a benchmark signal. The system is also configured to adaptively determining a spike sorting statistical model for the feature space samples. The system is further configured to identify one or more types of voluntary stimuli based on analysis of the feature space, wherein the projected information is grouped in clusters, each cluster defining a particular type of voluntary stimuli.

Description

BIOLOGICAL INTERFACE SYSTEM WITH NEURAL SIGNAL CLASSIFICATION SYSTEMS AND METHODS

[001] This application claims the benefit of priority under 35 U. S. C. § 119(e) of U.S. Provisional Application No. 60/735,158, filed November 10, 2005.

Field of the Invention

[002] Embodiments of the present invention relate to biological interface systems that include one or more devices that receive processed multicellular signals of a patient. For example, a processing unit produces a processed signal based on cellular signals received from a sensor comprising multiple electrodes. More particularly, an exemplary processing unit includes a neural signal classifier for detection, identification, and classification of neural signals received from a patient.

Description of Related Art

[003] Biological interface systems, for example neural interface systems, are currently under development for numerous patient applications including restoration of lost function due to traumatic injury or neurological disease and diagnosis and/or detection of neurological events such as epileptic seizures. These interface systems may include one or more sensors, such as electrode arrays configured to receive electrical signals from living cells. The sensors are implanted in the central nervous system, such as the motor cortex of the brain, and the peripheral nervous system. The cellular signals received may be processed to produce diagnostic, therapeutic and/or control signals, such as a control signal used to operate one or more controlled devices such as, for example, a wheelchair, a prosthetic limb, a robot, a computer, or any other type of controlled device.

[004] Because substantial amounts of electrical activity may occur within a neurological system such as the brain, a large number and variety of signals may be collected by each electrode and/or sensor. These signals may include electric signals associated with neurological activity, such as individual neuron firing (a neural spike), multicellular signals (an integration of multiple neural spikes such as Local Field Potential signals), other electrical physiologic signals (e.g. DC bias), and noise (e.g. environmental noise, physiologic noise and system noise such as noise associated with the electrode or other system electronics).

[005] Early attempts to utilize signals directly from neurons to control an external prosthesis encountered a number of technical difficulties. The ability to identify and obtain neural spikes is difficult, as well as often incomplete and unreliable. Another problem that has been encountered is caused by the changes that occur to the neural signals over time, resulting in a degradation of system performance. Neural interface systems that utilize other neural information, such as electrocorticogram (ECoG) signals, local field potentials (LFPs) and electroencephalogram (EEG) signals, have similar issues to those associated with individual neuron signals. Since all of these signals result from the activation of large groups of neurons, the specificity and resolution of the control signal that can be obtained may be limited. However, if these lower resolution signals could be properly identified and the system adapt to their changes over time, control signals could be generated to control rudimentary devices or work in conjunction with the more selective control signals processed directly from individual neurons.

[006] Early systems were extremely susceptible to additive noise, such as line noise, environmental noise, thermal drift, and other types of noise. Although filters were used to suppress some of the noise, these filters were not able to adequately adapt to variable noise signals. As a result, during prolonged operation of a system, particularly during a learning period, the neural signals became increasingly difficult to detect, due in large part to migration of noise signals back into the neural signal data.

[007] While currently available methods may provide a mechanism to identify some neurological signals, they may be inefficient, ineffective, time consuming, and costly. For example, in order to sufficiently "train" conventional processing systems, patients may be required to undergo several sessions in which data collected from sensors was manually classified by a technician based on the waveform size and shape. Not only is such a practice time consuming, it may also be vulnerable to operator error resulting from incorrectly identifying a noise spike as valid neurological activity and/or inadequately classifying a valid neuron signal for future detection. [008] In addition, conventional systems that rely solely on threshold detection of neural spikes may be incomplete, include false identifications or be otherwise unreliable, as any signal that includes high power spurious noise may be classified as a neural spike, regardless of whether the signal contains neural information. Similarly, signals that include neural content, but do not meet a minimum threshold value may be otherwise discarded. These methods not only produce unreliable results, they significantly decrease system performance.

[009] Furthermore, conventional automated systems that rely exclusively on complex algorithms and statistical models require large amounts of processing time and resource, which may be undesirable in commercial applications where patients may require simple, yet time efficient processing systems. Additionally, costs associated with providing the processing and memory requirements to accommodate conventional automated neural spike sorting systems may also unnecessarily increase costs associated with a commercial neural processing system.

[010] From the above discussion, it is apparent to one of skill in the art that there is a need for a biological interface system with an efficient neural signal classification system that effectively identifies neural spike activity. There is also a need for a neural signal processor that includes parameters that can be adjusted by one or more operators of the system. There is also a need for a neural classification and control system that adapts over time to account for changes in noise and/or changes in the behavior of the neural activity of a patient. There is also a need for a system that can detect multiple neural signals collected from different electrodes and effectively identify one or more neurons corresponding to each detected signal. There is also a need for a system that provides manual, automatic, and/or a combination of manual and automatic configuration of a neural training profile as well as allowing for manual rejection of detected neural spikes in a feature space. There is also a need for a neural signal processing system that includes an adaptive filter to account for additive noise injected onto the neural signal.

[011] The disclosed biological interface systems with neural signal classifier are directed towards overcoming one or more of the problems set forth above. Summary

[012] In accordance with one aspect, the present disclosure is directed to a method for neural signal classification for processing of multicellular signals. The method may include receiving a plurality of multicellular signals from a sensor associated with a biological system of a patient. The method may also include filtering the plurality of multicellular signals to produce a neural signal, the neural signal including a neural spike portion. The method may further include extracting the neural spike portion of the neural signal. The method may further include correlating the neural spike with a benchmark signal, the benchmark signal indicative of an ideal neural spike associated with the patient. The method may also include projecting samples indicative of the correlation on a feature space. The method may also include adaptively determining a spike sorting statistical model for the feature space samples. The method may further include classifying the neural spike based on one or more clusters of data samples observed in the feature space.

[013] According to another aspect, the present disclosure is directed toward a method for classifying neural signals for a biological interface system, comprising receiving a plurality of multicellular signals from a sensor associated with a biological system of a patient. The method may also include filtering each of the plurality of multicellular signals to produce one or more neural signals. The method may further include projecting the one or more neural signals on a feature space according to a characteristic associated with the one or more neural signals. The method may also include identifying one or more neural spikes associated with the one or more neural signals.

[014] According to yet another aspect, the present disclosure is directed toward a neural signal classification system for a biological interface system. The system may include an input channel for receiving at least one signal from a sensor associated with a biological system of a patient. The system may also include a filter operatively coupled input channel and configured to substantially suppress noise associated with the signal. The system may further include a signal separator operatively coupled to the filter for separating the at least one signal according to at least one predetermined frequency threshold. The system may further include a neural signal processor operatively coupled to the signal separator for identifying at least a portion of the at least one signal as a neural spike.

[015] In accordance with yet another aspect, the present disclosure is directed toward a neural signal classification system for identification and classification of neural spike activity. The system includes a preprocessing device operatively coupled to an input channel and configured to receive multicellular signals collected from a sensor, at least a portion of the sensor configured to be disposed within the brain of a patient. The preprocessing device may also be configured to filter the multicellular signals to extract a neural signal portion of the multicellular signals, the neural signal portion including a neural spike portion and a local field potential portion. The neural signal classification system may also include a neural spike processing device operatively coupled to the preprocessing device and configured to determine whether the neural spike portion includes a neural spike, the neural spike indicative of a voluntary stimulus associated with the patient. The system may also be configured to project information associated with a neural spike onto a feature space, the feature space indicative of a correlation of the neural spike with a benchmark signal. The system may be further configured to identify one or more types of voluntary stimuli based on analysis of the feature space, wherein the projected information is grouped in clusters, each cluster defining a particular type of voluntary stimuli.

[016] According to yet another aspect, the present disclosure is directed toward a method for classifying neural signals for a biological interface system. The method may include receiving a plurality of multicellular signals from a sensor associated with a biological system of a patient. The method may also include extracting a neural spike portion of the multicellular signal, wherein the neural spike portion corresponds substantially with a high- frequency portion of the multicellular signal. The method may further include determining whether the neural spike portion includes a neural spike indicative of a voluntary stimulus initiated by the patient. The method may also include correlating the neural spike with a benchmark signal, the benchmark signal indicative of an ideal neural spike associated with the patient, and identifying the neural spike based on the correlation. [017] In accordance with yet another aspect, the present disclosure is directed toward a biological interface system for collecting multicellular signals emanating from one or more living cells of a patient and for transmitting processed signals to a controlled device. The biological interface system may include a sensor for detecting the multicellular signals, the sensor consisting of a plurality of electrodes to allow for detection of the multicellular signals. The system may also include a processing unit configured to receive the multicellular signals from the sensor. The processing unit may also be configured to extract a neural spike portion of the multicellular signal, wherein the neural spike portion corresponds substantially with a high-frequency portion of the multicellular signal. The processing unit may be further configured to determine whether the neural spike portion includes a neural spike indicative of a voluntary stimulus initiated by the patient. The processing unit may also be configured to correlate the neural spike with a benchmark signal, the benchmark signal indicative of an ideal neural spike associated with the patient. The processing unit may be further configured to identify the neural spike based on the correlation. The processing unit may also be configured to transmit a control signal indicative of the neural spike to a controlled device. According to an exemplary embodiment, the biological interface system may also include the controlled device for receiving the processed signals.

[018] In accordance with yet another aspect, the present disclosure is directed toward a system for measuring impedance of an electrode associated with a biological interface system. The system may include a channel amplifier communicatively coupled to an electrode of the biological interface system, wherein the electrode is configured to collect multicellular signals associated with a biological system. The system may also include a reference amplifier communicatively coupled to a reference electrode, the reference electrode configured to collect noise data associated with the biological system. The system may further include a signal generator configured to periodically provide a test signal to the biological system for measuring the impedance of the electrode. The system may also include a processing device coupled to the channel amplifier and the reference amplifier. The processing device may be configured to receive multicellular signals from the channel amplifier when the signal generator is not coupled to the biological system and determine an impedance of each electrode when the signal generator is coupled to the biological system.

[019] Additional objects and advantages of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. The objects and advantages of the invention will be realized and attained by means of the elements and combinations particularly pointed out in the appended claims.

[020] It is to be understood that both the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the invention, as claimed.

Brief Description of the Drawings

[021] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate various embodiments of the present invention, and, together with the description, serve to explain the principles of the invention.

[022] Fig. 1 illustrates a schematic representation of the biological interface system consistent with an embodiment of the present invention.

[023] Fig. 2 illustrates an exemplary embodiment of a portion of the biological system, including sensor electrodes implanted in the brain of a patient and a portion of a processing unit implanted on the skull of the patient, consistent with the present invention.

[024] Fig. 3 illustrates another exemplary embodiment of a biological interface system consistent with the present invention wherein an operator configures the system at the patient site.

[025] Fig. 4 illustrates another exemplary embodiment of a biological interface system consistent with the present invention wherein a patient controls multiple devices and an operator configures the system at a site remote from the patient.

[026] Fig. 5 provides a flowchart illustrating system configuration routines for the biological interface system consistent with the disclosed embodiments. [027] Fig. 6 illustrates an exemplary biological interface system with a configuration routine that includes automated neural spike sorting and a system diagnosis.

[028] Fig. 7 illustrates a series of neural signals and two graphs illustrating a comparison of data associated with a preferred automated spike sorting routine to data gathered from manual spike sorting routines.

[029] Fig. 8 illustrates exemplary sets of manual spike sorting outputs versus automated spike sorting outputs as seen in a human patient model.

[030] Fig. 9 provides a graphs depicting cross correlation of manual and automated spike sorting outputs of Fig. 8.

[031] Fig. 10 illustrates system components for processing neural signals to extract spikes and generate parameters suitable for classification.

[032] Fig. 11 provides a flow chart depicting an exemplary configuration routine associated with the biological interface system consistent with the disclosed embodiments.

[033] Fig. 12 provides a schematic of an exemplary impedance diagnostic device in accordance with the disclosed embodiments.

[034] Fig. 13 provides a schematic of an exemplary circuit for measuring the impedance of a single electrode according to certain disclosed embodiments

[035] Fig. 14 provides a flowchart depicting an exemplary mode of operation of the biological interface system, serving as an impedance measuring system in accordance with the disclosed embodiments.

[036] Fig. 15 provides an output associated with an exemplary impedance measuring software program.

[037] Fig. 16 illustrates a neural signal classification system according to an exemplary embodiment.

[038] Fig. 17 illustrates an exemplary disclosed neural signal classification system including a data processing device consistent with the disclosed embodiments. [039] Fig. 18 illustrates a neural spike processing device associated with the neural signal classification system according to an exemplary disclosed embodiment,

[040] Fig. 19 illustrates exemplary high-frequency and low-frequency component detection modules associated with the neural spike processing device consistent with the disclosed embodiments.

[041] Fig. 20 illustrates a local field potential processing device associated with the neural signal classification system according to an exemplary disclosed embodiments.

[042] Fig. 21 provides a flowchart depicting an exemplary method of operating a neural signal classification system consistent with the disclosed embodiments.

[043] Fig. 22 provides a flowchart depicting a method for preprocessing a multicellular signal associated with the exemplary method of operating a neural signal classification system.

[044] Figs. 23A-23B provide flowchart illustrations of an exemplary method of operating a neural spike processing device consistent with the disclosed embodiments.

[045] Fig. 24 provides a flowchart depicting a method of operating a local field potential processing device according to an exemplary disclosed embodiment.

[046] Fig. 25 provides a graph depicting an exemplary Gaussian mixture model consistent with the disclosed embodiments.

[047] Figs. 26a-26b provide graphs illustrating exemplary neural spike clusters in accordance with the disclosed embodiments.

[048] Fig. 27 illustrates a graph depicting a principle component projection and associated histogram consistent with one exemplary embodiment.

[049] Fig. 28 illustrates a graph depicting a Fisher linear discriminant project and associated histogram consistent with one exemplary embodiment. [050] Fig. 29 illustrates a graph depicting a principle component limits range projection and associated histogram consistent with one exemplary embodiment.

[051] Fig. 30 illustrates a histogram where the "peak trace" line tracks the maximum of all previous bin counts in accordance with certain disclosed embodiments.

[052] Fig. 31 illustrates an exemplary histogram for a low-spike-rate cluster isolated from a high-spike-rate cluster in accordance with certain disclosed embodiments.

Detailed Description

[053] To facilitate an understanding of the invention, a number of terms are defined immediately herebelow.

Definitions

[054] As used herein, the term "biological interface system" refers to a neural interface system or any system that interfaces with living cells that produce electrical activity or cells that produce other types of detectable signals.

[055] As used herein, the term "cellular signals" refers to subcellular signals, intracellular signals, extracellular signals, single cell signals and signals emanating from one or more cells. The term "subcellular signals," as used herein, refers to: a signal derived from a part of a cell; a signal derived from one particular physical location along or within a cell; a signal from a cell extension, such as a dendrite, dendrite branch, dendrite tree, axon, axon tree, axon branch, pseudopod or growth cone; or signals from organelles, such as golgi apparatus or endoplasmic reticulum. The term "intracellular signals," as used herein, refers to a signal that is generated within a cell or by the entire cell that is confined to the inside of the cell up to and including the membrane. The term "extracellular signals," as used herein, refers to signals generated by one or more cells that occur outside of the cell(s). The term "cellular signals," as used herein, include but are not limited to signals or combinations of signals that emanate from any living cell. Specific examples of "cellular signals" include but are not limited to: neural signals; cardiac signals including cardiac action potentials; electromyogram (EMG) signals; glial cell signals; stomach cell signals; kidney cell signals; liver cell signals; pancreas cell signals; osteocyte cell signals; sensory organ cell signals such as signals emanating from the eye or inner ear; and tooth cell signals. The term "neural signals," as used herein, refers to neuron action potentials or neural spikes; local field potential (LFP) signals; electroencephalogram (EEG) signals; electrocorticogram signals (ECoG); and signals that are between single neural spikes and EEG signals.

[056] As used herein, the term "multicellular signals" refers to signals emanating from two or more cells, or multiple signals emanating from a single cell.

[057] As used herein, the term "patient" refers to any animal, such as a mammal and preferably a human. Specific examples of "patients" include but are not limited to: individuals requiring medical assistance; healthy individuals; individuals with limited function; and in particular, individuals with lost motor or other function due to traumatic injury or neurological disease.

[058] As used herein, the term "configuration" refers to any alteration, improvement, repair, calibration or other system-modifying event whether manual in nature or partially or fully automated.

[059] As used herein, the term "configuration parameter" refers to a variable, or a value of a variable, of a component, device, apparatus and/or system. A configuration parameter has a value that can be: set or modified; used to perform a function; used in a mathematical or other algorithm; used as a threshold to perform a comparison; and any combinations of these. A configuration parameter's value determines the characteristics or behavior of something. System configuration parameters are variables of the system of the present invention, such as those used by the processing unit to produce processed signals. Other numerous subsets of configuration parameters are applicable, these subsets including but not limited to: calibration parameters such as a calibration frequency parameter; controlled device parameters such as a time constant parameter; processing unit parameters such as a cell selection criteria parameter; patient parameters such as a patient physiologic parameter such as heart rate; multicellular signal sensor parameters; other sensor parameters; system environment parameters; mathematical algorithm parameters; a safety parameter; and other parameters. Certain parameters may be controlled by the patient's clinician, such as a password-controlled parameter securely controlled by an integral permission routine of the system. Certain parameters may represent a "threshold" such as a success threshold used in a comparison to determine if the outcome of an event was successful. In numerous steps of a system configuration or other function, a minimum performance or other measure may be maintained by comparing a detected signal, or the output of an analysis of one or more signals, to a success threshold.

[060] As used herein, the term "discrete component" refers to a component of a system such as those defined by a housing or other enclosed or partially enclosed structure, or those defined as being detached or detachable from another discrete component. Each discrete component can transmit information to a separate component through the use of a physical cable, including one or more of electrically conductive wires or optical fibers, or transmission of information can be accomplished wirelessly. Wireless communication can be accomplished with a transceiver that may transmit and receive data such as through the use of "Bluetooth" technology or according to any other type of wireless communication means, method, protocol or standard, including, for example, code division multiple access (CDMA), wireless application protocol (WAP), infrared or other optical telemetry, radio frequency or other electromagnetic telemetry, ultrasonic telemetry or other telemetric technologies.

[061] As used herein, the term "routine" refers to an established function, operation or procedure of a system, such as an embedded software module that is performed or is available to be performed by the system. Routines may be activated manually such as by an operator of a system, or occur automatically such as a routine whose initiation is triggered by another function, an elapsed time or time of day, or other trigger. The devices, apparatus, systems and methods of the present invention may include or otherwise have integrated into one or their components, numerous types and forms of routines. An "adaptive processing routine" is activated to determine and/or cause a routine or other function to be modified or otherwise adapt to maintain or improve performance. A competitive routine is activated to provide a competitive function for the patient of the present invention to compete with, such as a function which allows an operator of the system to compete with the patient in a system training task; or an automated system function which controls a visual object which competes with a patient controlled object. A "configuration routine" is activated to configure one or more system configuration parameters of the system, such as a parameter that needs an initial value assigned or a parameter that needs an existing parameter modified. A "language selection routine" is activated to change a language displayed in text form on a display and/or in audible form from a speaker. A "patient training routine" is activated to train the patient in the use of the system and/or train the system in the specifics of the patient, such as the specifics of the patient's multicellular signals that can be generated by the patient and detected by the sensor. A "permission routine" is activated when a system configuration or other parameter is to be initially set or modified in a secured manner. The permission routine may use one or more of: a password; a restricted user logon function; a user ID; an electronic key; a electromechanical key; a mechanical key; a specific Internet IP address; and other means of confirming the identify of one or more operators prior to allowing a secure operation to occur. A "remote technician routine" is activated to allow an operator to access the system of the present invention, or an associated device, from a location remote from the patient, or a system component to be modified. A "system configuration routine" is activated to config. the system, or one or more components or associated devices of the system. In a system configuration routine, one or more system configuration parameters may be modified or initially set to a value. A "system reset routine" is activated to reset the entire system or a system function. Resetting the system is sometimes required with computers and computer based devices such as during a power failure or a system malfunction.

General Description of the Embodiments

[062] This disclosure relates generally to systems and methods for neural signal classification and, more particularly, to systems and methods for neural signal classification for use with a biological interface system. While exemplary embodiments illustrate certain components and/or features associated with biological interface systems, particularly those illustrated in Figures 1 -4, these components and features provide exemplary biological interface systems 100, 100', 100" for performing neural signal classification. It is therefore contemplated that additional and/or different components than those illustrated in Figures 1-4 may be adapted to include systems, methods, and/or processes associated with neural signal classification. For example, a neural signal classification system 501 illustrated in Fig. 16 may include computer executable instructions stored on a computer readable medium that, when executed by a processor, may perform methods consistent with the disclosed embodiments. Neural signal classification system 501 may be integrated within one or more components associated with the exemplary biological interface systems illustrated in Figures 1-4, for example, or, alternatively, may be a standalone computer system, in communication with biological interface systems. According to one exemplary embodiment, neural signal classification system 501 may include a diagnostic system or subsystem that receives the processed signal from the processing unit and performs a diagnosis or provides diagnostic information to a user. According to another exemplary embodiment, neural signal classification system 501 may include a therapeutic system such as an epilepsy detection system that receives the processed signal form the processing unit and performs a therapeutic event such as a stimulation to prevent a seizure.

[063] Furthermore, neural signal classification system 501 may include processes that allow one or more systems to analyze collected signals received from a sensor placed near nerve cells, extract neural spikes from these sensor signals, and process these neural spikes to produce a diagnostic, therapeutic or control signal. It is also contemplated that neural signal classification system 501 may include adaptive processes to periodically update one or more components and/or functions associated with one or more components of a biological interface system. These adaptive processes may provide an integrated learning module designed to reduce the training time associated with customizing a biological interface system with a particular neural fingerprint associated with a particular patient.

[064] Systems, methods, apparatus and devices consistent with embodiments of the invention detect cellular signals generated within a patient's body and implement various signal processing techniques to generate processed signals for transmission to one or,more devices. The system includes a sensor, consisting of a plurality of electrodes that detect multicellular signals from one or more living cells, such as from the central or peripheral nervous system of a patient. The system further includes a processing unit that receives and processes the multicellular signals and transmits a processed signal to a diagnostic, therapeutic or controllable device. The processing unit utilizes various electronic, mathematic, neural net and other signal processing techniques in producing the processed signal.

[065] An integrated system configuration routine is embedded in one or more components of the system. The system stores sets of multicellular signals, and uses the stored signals to generate one or more system configuration parameters, including parameter values such as initial values and modified values. These configuration parameters are used to produce an input-output relationship, such as a transfer function that is applied to subsequent multicellular signals to produce the processed signals. The configuration routine may be a requirement of the system prior to allowing full use of the processed signals by the diagnostic, therapeutic or controllable device. The configuration routine may adapt over time, such as to improve system performance and/or reduce the patient requirements of subsequent routines that are performed. The configuration routines may provide a system configuration plan, and the configuration plan may be adjusted based on the measurement of one or more parameters that are collected prior to requiring the patient to control a controlled device or a surrogate of a controlled device. The configuration routine may require no operator other than the patient, or may work with an operator at a remote location, such as a clinical site or a service group of the manufacturer of the biological interface system.

[066] The configuration routine may include a visual representation of a human figure. The representation may be picture based, such as pictures from a video or digital camera of an actor providing the human movements, or may be a digital image or animation of one or more drawing or computer generated human fig. graphics. The human fig. may be adjustable, such as by the patient, these adjustments including whether the movements are accomplished by left or right side body limbs, and which gender should be represented. Modifications such as these can be accomplished with the use of a patient input device, such as a tongue or neck switch or other input device. Additional feedback can be provided to the patient, simultaneously or at a different time, such as audio feedback provided through one or more speakers. This audio feedback may include combinations of tones or spoken language. The additional feedback Ts provided to improve the quality of the system configuration parameters generated, to generate additional system configuration parameters, and/or provide an additional function. Other forms of feedback can be additionally or alternatively provided to the patient, such as feedback selected from the group of: visual; tactile; auditory; olfactory; gustatory; electrical stimulation such as cortical stimulation; and combinations of the preceding. Additional visual feedback may include a second visual representation of a human figure, provided simultaneously with the first human fig. or at different times.

Detailed Description of the Embodiments

[067] Reference will now be made in detail to the present embodiments of the invention, examples of which are illustrated in the accompanying drawings. Wherever possible, the same reference numbers will be used throughout the drawings to refer to the same or like parts.

[068] Fig. 1 illustrates a biological interface system 100 according to an exemplary disclosed embodiment. Biological interface system 100 may include a sensor 200, a processing unit first portion 130a, a processing unit second portion 130b, and/or one or more controlled devices 300a-d. One or more components associated with biological interface system 100 may be implanted within the body of a patient or, alternatively, may be located substantially external to the body of the patient. For example, as illustrated in Fig. 1 , sensor 200 and processing unit first portion 130a may be implanted inside the body, under the skin of a patient, while processing unit second portion 130b, controlled devices 300a-d, and a neural classification system 501 (shown in Fig. 16) may be located external to the body of the patient. It is contemplated that one or more components illustrated as being located within the body may be located external to the body. Similarly, one or more components illustrated as being located external to the body may be located within the body. Furthermore, although sensor 200, processing unit first portion 130a, processing unit second portion 130b, and a neural classification system may be separate, it is contemplated that one or more of these components may be combined in a single, integrated unit. Biological interface system 100 may include additional, fewer, and/or different components than those listed above. [069] Sensor 200 may include a plurality of electrodes, not shown, for detecting multicellular signals. Sensor 200 may take various geometric forms and include numerous materials of construction. All exposed surfaces, such as surfaces that come in contact with tissue or bodily fluids, comprise biocompatible materials well known to those of skill in the art. In a preferred embodiment, sensor 200 includes a ten by ten matrix of electrodes; the electrodes are included at the tip of individual projections, these projections spaced at approximately 400 μm with a length of 1.0 to 1.5 mm; and the electrodes typically have an impedance between 100 kOhm and 1 MOhm. Sensor 200 may be placed at various locations internal and/or external to a patient, and may comprise multiple discrete components that are placed at one or more locations proximate to one or more living cells.

[070] Another element of system 100 is a processing unit that receives the multicellular signals from sensor 200, and utilizes one or more signal processing techniques to produce processed signals. Depicted in Fig. 1 is processing unit first portion 130a and processing unit second portion 130b which are each a component of the processing unit of an embodiment of the present invention. Additional components may also be part of the processing unit, all of the components collectively performing the receiving of the multicellular signals and the production of the processed signals. Processing unit discrete components can be implanted within the patient, be external to the patient, or protrude through the skin of the patient.

. [071] As depicted in Fig. 1 , processing unit first portion 130a is implanted under the skin of the patient such as on top of the skull of the patient under the scalp. In a preferred embodiment, sensor 200, also implanted, is placed within the skull such that one or more electrodes are placed within a cortical layer of the brain. Wire bundle 220, a single or multi-conductor cable, is attached to sensor 200 and processing unit first portion 130a. Wire bundle 220 attaches to one or more electrodes of sensor 200 and may include other conductors or conduits such as a conductor that provides a reference signal at a location in proximity to the electrodes of sensor 200. In a preferred embodiment, multiple individual electrodes of sensor 200 are attached each to individual conductors of wire bundle 220, and wire bundle 220 includes at least two conductors that do not attach to electrodes that are placed to provide relevant reference signals for one or more signal processing functions. In a preferred embodiment, the conductive wires of wire bundle 220 have a diameter of approximately 25 μm and comprise a blend of gold and palladium. Wire bundle 220 conductors are attached at their other end to processing unit first portion 130a and the conductors and housing of processing unit first portion 130a are sealed such that the signals, conductive surfaces, and other internal components of wire bundle 220 and processing unit first portion 130a are appropriately protected from contamination by body fluids and other contaminants.

[072] Processing unit first portion 130a includes means of amplifying the cellular signals, amplifier 131 , which is preferably an amplifier with a gain of approximately one hundred, a working frequency range of 0.001 Hz to 7.2 kHz, a power requirement of approximately 1.6V and a power dissipation of approximately 3OmW. Processing unit first portion 130a further includes additional signal processing means, signal processing element 132a. Various signal processing techniques can be utilized including but not limited to: filtering, sorting, conditioning, translating, interpreting, encoding, decoding, combining, extracting, sampling, multiplexing, analog to digital converting, digital to analog converting, mathematically transforming, and/or otherwise processing multicellular signals to generate a processed signal such as a control signal for transmission to a controlled device. In a preferred embodiment, signal processing element 132a includes a multiplexor function, such as a thirty-two to one multiplexor with a 1 MHz switching frequency. In another preferred embodiment, signal processing element 132a includes an analog to digital converter with twelve-bit resolution that can process 1 megasample per second data for thirty-two channels.

[073] It is desirable that all implanted components avoid the need to protrude through the skin of the patient, such as for cosmetics and reduced infection risk. In order for processing unit first portion 130a to transmit one or more signals to an external component, IR transmitter 133 is incorporated into the implant. IR transmitter 133 is preferably one or more infrared (IR) light emitting diodes (LEDs), such IR transmissions able to penetrate through a finite amount of tissue, such as the scalp. In a preferred embodiment, IR transmitter 133 transmits data at 40 megabits per second utilizing direct modulation. IR transmitter 133 receives information from signal processing element 132a, and transmits the information to processing unit second portion 130b by way of its integrated receiver, IR receiver 181. Both IR transmitter 133 and IR receiver 181 can include lenses, filters and other optical components to focus, filter, collect, capture, or otherwise improve the IR transmission and receiving performance.

[074] Processing unit second portion 130b, a component external to the body of the patient, is affixed or otherwise placed at a location in close proximity to the location of processing unit first portion 130a's transmitter, IR transmitter 133. In a preferred embodiment, processing unit first portion 130a is placed in a recess made in the skull, during a surgical procedure, at a location near to and above the ear of the patient. Processing unit second portion 130b is placed on the head just above the ear such that IR receiver 181 is at a location near aligned with IR transmitter 133, such as a line of site distance of approximately 4mm. Information transfer takes place such as that using various error detection schemes, handshaking functions and other communication and error checking protocols such as ANSI X3.230 protocol and other protocols well known to those of skill in the art and applicable to digital, analog and combined digital/analog critical use communications. In an alternate embodiment, processing unit first portion 130a may be implanted in the torso of the patient, and processing unit second portion 130b may be located on the skin proximate processing unit first portion 130a, such as at a location that is normally covered by clothing, thus providing patient privacy. In another alternate embodiment, processing unit 130a may be implanted in the back of the neck of the patient, and processing unit second portion may be located in a wheel chair seatback.

[075] Processing unit first portion 130a may include one or more additional elements, not shown, but included within, on the surface of, or attached to processing unit first portion 130a. Such elements may include but are not limited to: a temperature sensor, a pressure sensor, a strain gauge, an accelerometer, a volume sensor, an electrode, an array of electrodes, an audio transducer, a mechanical vibrator, a drug delivery device, a magnetic field generator, a photo detector element, a camera or other visualization apparatus, a wireless communication element, a light producing element, an electrical stimulator, a physiologic sensor, a heating element and a cooling element. In a preferred embodiment, processing unit first portion 130a may include a stimulator element, not shown but configured to provide electrical current and/or voltage to one or more electrodes of sensor 200. Processing unit first portion 130a may include an integrated power supply, not shown, to provide power to amplifier 131 , signal processing element 132a, IR transmitter 133 (or other high-frequency transmitting device such as, for example, and RF or microwave transmitter), or another component, not shown, of processing unit first portion 130a. In addition, power may be supplied to a power requiring component of sensor 200 such as by way of one or more conductors of wire bundle 220. Depicted in Fig. 1 , processing unit first portion 130a includes a coil, implanted coil assembly 134, the assembly being configured to receive and convert electromagnetic signals from a device external to the body of the patient, preferably processing unit second portion 130b. Processing unit second portion 130b, also includes a coil, coil assembly 182, which is oriented within a housing of processing unit second portion 130b such that when IR receiver 181 is near aligned with IR transmitter 133, coil assembly 182 can be near aligned with implanted coil assembly 134. The coil in implanted coil assembly 134 is preferably approximately 1 inch in diameter.

[076] Through inductive coupling, power can be transferred from processing unit second portion 130b to processing unit first portion 130a by supplying a driving signal to coil assembly 182 that generates an electromagnetic field that, through inductive coupling, generates power in implanted coil assembly 134. This captured energy is converted to usable power by circuitry incorporated into implanted coil assembly 134 and can be used to power one or more elements of processing unit first portion 130a and/or recharge an integrated power supply, not shown. In the preferred embodiment shown in Fig. 1 , no implanted component includes an integrated power supply such that, when coil assembly 182 is not properly energized and/or when processing unit second portion 130b is not in relative proximity to the patient, no implanted component has power. In another preferred embodiment, information can be transferred from processing unit second portion 130b to processing unit first portion 130a by modulating the power transfer waveform, such as with modulation circuitry included in coil assembly 182 or another component of processing unit second portion 130b. The transmission is received and decoded by the coil and circuitry of implanted coil assembly 134. This modulation pattern can easily be encoded and decoded to provide means of sending information to the implant, such as in a configuration procedure, embedding of a unique identifier, or other procedure.

[077] Processing unit second portion 130b also includes signal processing element 132b. Signal processing can include one or more of the processes listed above in reference to signal processing element 132a and preferably includes at least a signal decoding function or a multiplexing function. These signal processing means, in combination with signal processing element 132a of processing unit first portion 130a may complete the processing unit function of the system of the present invention such that the two signal processing means in combination produce the processed signals that will be used to control first controlled device 300a, second controlled device 300b, or both. Processing unit second portion 130b may include wireless communication means or wired communication means (e.g. cables 301a and 301b), to transmit the processed signals to the controlled devices of the system. The various embodiments and elements utilizing wireless communication means can utilize radiofrequency (RF), infrared, ultrasound, microwave and/or other data transmission technologies that do not require a physical conductor or combinations of the preceding technologies. The various embodiments and elements utilizing wired communication means can comprise electrical conductors, optical fibers, sound wave guiding conduits, other physical cables and conductors or combinations of the preceding.

[078] Also depicted in Fig. 1 is selector module 400, a component of an embodiment of the system of the present invention that is used by an operator to select one or more devices to be controlled by system 100. System 100 can have one or more operators including but not limited to: the patient; a technician; a clinician; a caregiver and a family member of the patient. In a preferred embodiment, selector module 400 can select more than one controlled device, such that processed signals control multiple controlled devices simultaneously. When multiple controlled devices are controlled simultaneously, the processed signals sent to each controlled device may be identical or different. Selector module 400 at least sends information to processing unit second portion 130b via cable 183 (e.g., a multi-conductor physical cable). It should be appreciated that various communication means could be used including but not limited to: wired electrical connection, optical fiber connection, other physical cable communication means, wireless communication, or combinations of the preceding. At a minimum, in either wireless or physical conductor communications, processing unit second portion 130b includes data receiving means, and selector module 400 includes data transmission means, both not shown. In an alternative preferred embodiment, both processing unit second portion 130b and selector module 400 each include a transceiver element, such as a wireless transceiver element, which can both transmit and receive data.

[079] Selector module 400 may also include signal processing means, signal processing element 132c, such that selector module 400 can perform signal processing for various purposes including contributing to the processing unit function of the system of the present invention, such as the neural signal classification function. Signal processing can include one or more of the processes listed above in reference to signal processing element 132a. In an alternative embodiment, signal processing element 132c completes the requirements of the processing unit, in combination with signal processing element 132a of processing unit first portion 130a, and signal processing element 132b of processing unit second portion 130b, such that processed signals can be sent to the controlled devices by a data transmission element, such as information transmission means 410. In a preferred embodiment, selector module 400 performs a signal processing function, and processed signals are transmitted from selector module 400 to the controlled devices. In an alternative preferred embodiment, processing unit second portion 130b completes the signal processing of the multicellular signals, and selector module 400 transmits a selection signal to processing unit second portion 130b. This selection signal identifies which specific device is to be controlled by the processed signals.

[080] A method of controlling one or more specific controlled devices can be accomplished by a unique identifier contained in the processed signals transmitted to the controlled devices wherein the controlled devices includes means of identifying and/or differentiating the appropriate identifier. This identification confirming means may be a part of each controlled device, or a separate discrete component in communication with one or more controlled devices. When a controlled device receives the proper unique identifier, control will commence. The transmission of the identifier can be at the outset of control, or may be required on a continuous basis, such as by being included with individual packets of transmitted information. A limited transmission or one-time sending of the identifier can be accompanied by an initiate command to start control. Similar approaches can be performed to cease control of one or more controlled devices. In continuous identifier transmission, cessation of control is accomplished by discontinuation of transmission of the identifier with the individual packets. In limited or one-time transmission of the identifier, the identifier can be resent and accompanied by a cessation command.

[081] The unique controlled device identifier approach is a preferred method when processed signals are transmitted to controlled devices with wireless communication means, such that when two or more controlled devices may both be in proximity to receive the processed signals but only the appropriate one or more controlled devices will be controlled by the processed signals. An alternative method of controlling one or more specific controlled devices involves directing the processed signals to one or more specific conductors connected to one or more specific controlled devices. Referring again to Fig. 1 , processing unit second portion 130b connects to first controlled device 300a with cable 301a, and processing unit second portion 130b connects to second controlled device 300b with cable 301 b. Both cable 301a and cable 301 b receive processed signals as determined by conductor selection circuitry 186. Conductor selection circuitry 186 may include solid state relays, transistor switches, or other signal switching or controlling circuitry well known to those of skill in the art. Based on the information received from selector module 400, processed signals are sent to first controlled device 300a and/or second controlled device 300b as the appropriate connections are made in conductor selection circuitry 186.

[082] Referring again to Fig. 1 , a wireless method of controlled device selection is illustrated. Selector module 400 includes an element to transmit the processed signal wirelessly, such as information transfer means 410, preferably RF wireless technology. Information transfer means 410 receives processed signals from signal processing element 132c via power and data bus 420. Power and data bus 420 is a series of conductors that include power and data signals, such as a series of conductive traces integral to a printed circuit board that connect multiple circuit board mounted components to similar conductors, such bus architecture well known to those of skill in the art.

Information transfer means 410 receives power from an integrated power supply, integrated battery 401 , preferably a replaceable or rechargeable battery. Numerous battery technologies, including rechargeable chemistries, can be incorporated into integrated battery 401 such as nickel cadmium or lithium iodide technologies. As depicted in Fig. 1 , integrated battery 401 also provides power, via power cable 184, to processing unit second portion 130b such as to IR receiver 181 , coil assembly 182 and signal processing element 132b. In a preferred embodiment, selector module 400 includes a redundant power supply (e.g., backup battery 408). Backup battery 408 may provide power to components of selector module 400 at specific times only, such as during a power failure or during an alarm condition. In another preferred embodiment, selector module 400 attaches to a standard household outlet for access to 120VAC power (or similar AC line power) through a standard plug and power cord, not shown, attached to a power converter integral to selector module 400, power converter also not shown. The power converter supplies power to the various elements of selector module 400 via bus 420 and also may recharge either or both integrated battery 401 and backup battery 408.

[083] Information transfer means 410 transmits wireless information sent to both third controlled device 300c and fourth controlled device 30Od. Utilizing an embedded unique identifier transmission, and unique identifiers incorporated into third controlled device 300c and fourth controlled device 30Od, each controlled device can be uniquely controlled or controlled simultaneously. The embodiment of Fig. 1 describes a system 100 that allows first controlled device 300a and second controlled device 300b to be independently controlled by processed signals received from processing unit second portion 130b as determined by inputs made to selector module 400. The system also allows third controlled device 300c and fourth controlled device 30Od to be independently controlled as determined by inputs made to selector module 400, except that the processed signals are received from selector module 400. Any of the processed signals, including processed signals transmitted via a wired connection, may include the embedded unique identifier, described above, to facilitate or ensure the selection of the device to be controlled.

[084] Selector module 400 includes a data input device, input element 402 that enables a selection of a specific controlled device to receive the processed signals of the system. Input element 402 is connected to power and data bus 420 to receive power from integrated battery 401 , as are all elements attached to bus 420, and to transmit and receive signals from one or more elements of selector module 400 such as an integrated central processing unit, CPU 405 and signal processing element 132c. CPU 405 can perform numerous processing functions well known to those of skill in the art of computers and computer controlled devices. The processing functions performed by CPU 405 can work in conjunction with the various elements of selector module 400 such as those connected to bus 420. CPU 405 receives power via power and data bus 420.

[085] Input element 402 may comprise one or more of: a keyboard, a keypad, a data entry mechanical switch or button, a mouse, a digitizing tablet, a touch screen, or other data entry element. Mechanical switches are available in various forms for persons with limited movement such as from a spinal cord injury, these patients being an applicable receiver of the system of the present invention. These forms of switches and other data entry devices include but are not limited to: a sip and puff device; an eye gaze device; a hand, tongue or other muscle activated joystick or switch; an electromyogram (EMG) activated switch; and an electro-oculogram (EOG) activated switch. Input element 402 may additionally or alternatively include a voice recognition or voice activation element to select the controlled device and/or perform a different function. Alternatively or additionally, input element 402 may include a biological signal input element. Biological signals may include one or more processed signals of the system of the present invention, or a different biological signal such as one that is under voluntary control of the patient. Neural signals can be used to accomplish the selection of the device to be controlled. These neural signals may include one or more of: neural spikes; electrocorticogram signals; local field potential signals, and electroencephalogram signals. Other signals determining the selection may include signals derived from one or more of: eye motion; eyelid motion; facial muscle; or other electromyographic activity. Signals such as EKG, respiration, and blood glucose can also be used to trigger the selection process, such as to cease control of one or more devices when an abnormal heart rate is detected.

[086] Input element 402 may provide functions in addition to the selection of the controlled device to be controlled. Input element 402 may include a physical port such as a mechanical jack attached to a power line or other power receiving means such that power can be delivered to selector module 400. Wireless power receiving means may be included to allow power transfer such as through inductive coupling between mating coils. The received power may be used to power one or more elements of selector module 400 or to recharge an internal power supply such as integrated battery 401. Input element 402 may include a physical port for a different purpose, such as to provide a connection between selector module 400 and a computer network. The computer network can be one or more of: a local area network (LAN); a wide area network (WAN); a wireless fidelity network (WlFI) and the Internet. Access via a computer network such as the Internet allows selector module 400 to be accessed from a location remote to the patient of system 100 such as to retrieve information, select a controlled device or perform another function involving two-way data communication.

[087] Input element 402 may be a switch attachment port, such that a switch can be attached to selector module 400 to perform one or more tasks; initiate, cease or modify one or more processes or functions; or enter data, such as system parameter data. Applicable patient activated switches include but are not limited to: a sip and puff device; an eye gaze device; a hand, tongue or other muscle joystick; an electromyogram (EMG) activated switch; and an electro-oculogram (EOG) activated switch. Input element 402 may include a tilt switch (e.g. a mercury switch), such that if selector module 400 is in an unacceptable orientation, an alert signal is provided via bus 420 to one or more elements. In a preferred embodiment, selector module 400 is mounted to a wheel chair, and a tilt switch would indicate when the wheelchair had fallen over. The tilt switch signal could be processed, such as by CPU 405 and selector module 400 or another component of system 100 to cause system 100 to enter an alarm condition. An audible alert can alert a nearby party, or wireless transmission of information can alert a remote party of the emergency situation. Input element 402 may include one or more sensors. A power failure sensor can be incorporated to monitor various power levels including the battery level of integrated battery 401 or the voltage level of an attached AC power line. Other applicable sensors include but are not limited to: a physiological sensor including a neural sensor; an EKG sensor; a glucose sensor; a respiratory sensor; an activity or motion sensor; an environmental sensor; a temperature sensor; a strain gauge, an implanted sensor; a position sensor; an accelerometer; an audio sensor such as a microphone; and a visual sensor such as a phototfansistor.

[088] As depicted in Fig. 1 , selector module 400 includes an output element 403. In a preferred embodiment, output element 403 is used in the controlled device selection process, such as to provide output device selection means, output device information, or other system information. Output element 403 may include a visual display, such as a touch screen display, and the visual display may display selectable icons representing one or more controlled devices. Output element 403 may include a transducer, such as an audio transducer, a tactile transducer, an olfactory transducer or a visual transducer. These transducers can be used to confirm an event, such as by sounding an audible beep when a controlled device is selected or deselected, or to alert the user of an alarm or warning condition.

[089] As depicted in Fig. 1 , selector module 400 includes multiple other functional elements such as sensors, transducers, and other functional elements, input devices, and output devices. Memory storage element 407 utilizes one or more electronic memory circuitry such as random access memory (RAM), read-only memory (ROM) or other volatile and non-volatile memory storage devices. Various pieces of information can be stored including but not limited to: integrated parameter status and history of change of values; controlled device information; system change information and other historic system information; synchronization information that can be used to restore or backup information such as information that is lost due to a system or component failure, power outage, or other cause; patient information, and other information. All or part of the information stored in memory storage element 407 may also be included in a storage element of another discrete component of system 100, such as processing unit second portion 130b. In a preferred embodiment, system 100 includes a system synchronization function, such that redundant information is placed in one or more storage elements such as memory storage element 407 of selector module 400. The system synchronization function is similar to synchronization functions utilized in commercial personal data assistants (PDAs) to synchronize data between the PDA and a personal computer database of information. In system 100, the system synchronization function can place information redundantly in one or more storage modules such that if one or more components fail such as by losing a value for an integrated parameter or other system information, is replaced or otherwise is unavailable, all parameters can be reloaded utilizing the redundant data.

[090] System 100 of Fig. 1 further includes geographic location means 406, which provides geographic position location of selector module 400 such as via a global positioning system (GPS) transducer. This geographic information can be provided to a user, such as a remote user during an alarm condition. Notification to a remote user of an alarm condition can be accomplished via an Internet connection described above, or through use of wireless communication means such as cellular telephone communications. Various alarm conditions may require assistance to the patient such as a tipped wheelchair, failed controlled device, power failure, system malfunction, undesired patient condition or other adverse events. In a preferred embodiment, system 100 includes an alarm detection element to detect one or more alarm conditions, such as system malfunction conditions or patient adverse conditions.

[091] Selector module 400 of Fig. 1 further includes a second wireless communication element, such as redundant information transfer means 409. Information transfer means 409 provides a separate capability of communicating with a separate device such as a remote controlled device, data communication, transfer or retrieval device, or other device incorporating a wireless receiver, a wireless transmitter or a wireless transceiver. Redundant information transfer means 409 may be powered by either integrated battery 401 , backup battery 408 or both. In emergency situations such as system 100 entering an alarm state, either or both information transfer means 410 and redundant information transfer means 409 may generate and/or transmit an alert or distress signal to a remote location or a remote communication device. The alert signal may include one or more of: system condition; patient condition; patient identification; system location; and patient location. Numerous events can trigger an alarm state and are described throughout this application. System 100 may typically enter an alarm state during one or more of: power failure; system malfunction; controlled device malfunction; controlled device in unacceptable orientation or position; and unacceptable environment encountered. [092] Selector module 400 further includes functional module 404, an element that can perform various functions valuable to a patient, operator or other user of system 100. The functions performed by functional module 404 may include but are not limited to: personal data assistant; phone; cellular phone; pager; calculator; electronic game; glucometer; computer; device remote control; universal remote control; and environmental control device. In a preferred embodiment, functional module 404 includes a cellular phone, and this phone can automatically dial one or more predetermined phone numbers during an alarm state or condition.

[093] In a preferred embodiment, selector module 400 includes patient feedback means. The patient feedback means can be used to improve device control and/or to assist in patient training and system configuration. Feedback can be provided by output element 403, such as incorporating one or more of a visual display, an audible transducer, a tactile transducer or other transducer. Each transducer of output element 403 may be incorporated into or on a housing of selector module 400 or one or more transducers or displays may operably connect to a jack provided on selector module 400. In a preferred embodiment, the patient feedback function utilizes, at a minimum, audio feedback.

[094] In another preferred embodiment, selector module 400 includes a separate device control function. Examples of separate devices to be controlled, such as via input element 402, include a universal remote or a medical device such as a therapeutic device, a diagnostic device, a restorative device, and an implanted device.

[095] Selector module 400 includes one or more integrated parameters used to perform a function. These types of integrated parameters are incorporated into multiple discrete components of system 100. Examples of integrated parameters and the functions dependent on their use are described in detail throughout this application. A typical function requiring one or more integrated parameters is classification of neural spikes as well as production of the processed signals, both of the present invention. The integrated parameters of selector module 400 can be stored in memory storage element 407. When the integrated parameters of selector module 400 are modified, a permission routine may be invoked. [096] Other functions incorporated into selector module 400 include an information retrieval function, used to retrieve current or historic information from one or more discrete components of system 100 such as selector module 400; an interrogation function used to query the current or historic status of one or more discrete components of system 100; a system diagnostic function, used to diagnose one or more conditions, occurrences or states of system 100; a patient diagnostic function, used to perform or assist in the performance of a patient diagnostic event; and a configuration function, such as a calibration or other configuration process performed on system 100 to improve system performance and safety. In a preferred embodiment, the configuration function may be performed at least one time during the use of system 100, and in another preferred embodiment, the configuration function may be successfully completed prior to initiation of control of the controlled devices of system 100.

[097] Alternative embodiments of selector module 400 should also be considered within the spirit and scope of this application. Selector module 400 may comprise two or more discrete components, such as a wheelchair mounted component and a bed mounted component, and each discrete component may be able to operate independently with full functionality. Selector module 400 may include an embedded identifier, such as to confirm compatibility of selector module 400 with other components of system 100. Selector module 400 may be implanted within the patient. Selector module 400 may be a controlled device of the system of the present invention.

[098] Referring now to Fig. 2, a brain implant apparatus consistent with an embodiment of the present invention is illustrated. As shown in Fig. 2, the system includes a sensor (e.g., electrode array 210) that may be inserted into a brain 250 of patient 500, through an opening surgically created in skull 260. Array 210 includes a plurality of electrodes 212 for detecting electrical brain signals or impulses. Array 210 may be placed in any location of a patient's brain allowing for electrodes 212 to detect these brain signals or impulses. In a preferred embodiment, electrodes 212 can be inserted into a part of brain 250 such as a portion of the motor cortex associated with control of a patient's limb. Other locations for array 210, such as those outside of the cranium, can record cellular signals as well, such as locations neighboring spinal cord cells and locations neighboring peripheral nerve cells. Non-penetrating electrode configurations, such as subdural grids, cuff electrodes and scalp electrodes are applicable both inside the cranium such as to record local field potentials (LFPs), in, on, or near peripheral nerves, and on the surface of the scalp such as to record electroencephalogram signals (EEGs). Though Fig. 2 depicts the sensor as a single discrete component, in alternative embodiments the sensor comprises multiple discrete components, such as multiple arrays of electrodes implanted in portions of the motor cortex associated with multiple limbs . Multiple discrete components of the sensor can be implanted entirely in the brain or at an extracranial location, or the multiple discrete sensor components can be placed in any combination of locations.

[099] Electrode array 210 serves as the sensor for the biological interface system of embodiments of the present invention. While Fig. 2 shows electrode array 210 as seven aligned and similar length electrodes 212, array 210 may include one or more electrodes having a variety of sizes, lengths, shapes, forms, and arrangements, and preferably is a ten by ten array of electrodes. Moreover, array 210 may be a linear array (e.g., a row of electrodes) or a two-dimensional array (e.g., a matrix of rows and columns of electrodes), or wire or wire bundle electrodes. An individual wire lead may include a plurality of electrodes. Electrodes may have the same materials of construction and geometry, or there may be varied materials and/or geometries used in one or more electrodes, such as to create varied impedances for two or more electrodes. Each electrode 212 of Fig. 2 extends into brain 250 to detect one or more cellular signals such as those generated from the neurons located in proximity to each electrode 212's placement within the brain. Neurons may generate such signals when, for example, the brain instructs, or attempts to instruct a particular limb to move in a particular way. In a preferred embodiment, the electrodes reside within an arm or leg portion of the motor cortex of the brain.

[0100] In the embodiment shown in Fig. 2, array 210 includes a sensor substrate 213 that includes multiple projections 211 emanating from a surface of the substrate 213. At the end of each projection 211 is an electrode 212. Multiple electrodes, not shown, may be included along the length of one or more of the projections 211. Projections 211 may be rigid, semi-flexible, or flexible, the flexibility of which are such that each projection 211 can still penetrate into neural tissue, potentially with an assisting device or with projections that temporarily exist in a rigid condition. One or more projections 211 may be void of any electrode, such projections potentially including anchoring means such as bulbous tips or barbs, not shown. Two or more projections may have different lengths, tapers and/or diameters, differences not shown. Array 210 has previously been passed through a hole cut into skull 260, during a procedure known as a craniotomy, and inserted into brain 250 such that the projections pierce into brain 250 to a desired depth and sensor substrate 213 remains in close proximity to or in light contact with the surface of brain 250. The processing unit of the present invention includes processing unit first portion 130a, placed in a surgically created recess in skull 260 at a location near patient 500's ear 280. Processing unit first portion 130a receives cellular signals from array 210 via wire bundle 220, such as a multi-conductor cable of electrical wires. Processed signals are produced by processing unit first portion 130a and other processing unit components, such as processing unit second portion 130b located on the external skin surface of patient 500 near ear 280. The multicellular signals received from one or more electrodes

212 of array 210 include a time code of brain activity, including both single cellular and integrated multi-cellular activity. Processing unit first portion 130a and processing unit second portion 130b have similar elements and functionality to the identical referenced items of Fig. 1.

[0101] In the preferred embodiment depicted in Fig. 2, bone flap 261 , preferably the original bone portion removed in the craniotomy, has been used to close the hole made in the skull 260 during the craniotomy, obviating the need for a prosthetic closure implant. Bone flap 261 is attached to skull 260 with one or more straps or bands 263, preferably made of titanium or stainless steel. Band 263 is secured to bone flap 261 and skull 260 with bone screws 262. Wire bundle 220 passes between bone flap 260 and the hole cut into skull 260, potentially through a groove a recess also created in the surgical implantation procedure. During the surgical procedure, a recess was made in the top of skull 260 such that processing unit first portion 130a could be placed in the recess, allowing scalp 270 to be relatively flat in the area proximal to processing unit first portion 130a. A long incision in the scalp between the craniotomy site and the recess can be made to place processing unit first portion 130a in the recess. Alternatively, an incision can be made to perform the craniotomy, and a separate incision made to form the recess, and the processing unit first portion 130a and wire bundle 220 can be tunneled under ^"the scalp. to the desired location. Processing unit first portion 130a is attached to skull 260 with one or more bone screws and/or a biocompatible adhesive, not shown.

[0102] In an alternative embodiment, processing unit first portion 130a may be placed entirely within skull 260 or be shaped and placed to fill the craniotomy hole instead of bone flap 261. Processing unit first portion 130a can be placed in close proximity to array 210, or a distance of typically 5-20 cm can separate the two components. Processing unit second portion 130b, placed at a location proximate to implanted processing unit first portion 130a but external to patient 500, receives information from processing unit first portion 130a via wireless communication through the skin. Processing unit second portion 130b can include means of securing to patient 500 including but not limited to: an ear attachment mechanism; a holding strap; temporary adhesives; magnets; or other means. Processing unit second portion 130b, includes, in addition to wireless information receiving means, power transfer means, signal processing circuitry, an embedded power supply such as a battery, and information transfer means. The information transfer means of processing unit second portion 130b may include means to transfer information to one or more of: implanted processing unit first portion 130a; a different implanted device; and an external device such as an additional component of the processing unit of the present invention, a controlled device of the present invention, or a computer device such as a computer with Internet access.

[0103] Referring back to Fig. 2, electrodes 212 transfer the detected cellular signals to processing unit first portion 130a via array wires 221 and wire bundle 220. Wire bundle 220 includes multiple conductive elements, array wires 221 , which preferably include an individual conductor for each electrode of array 210. Also included in wire bundle 220 are two conductors, first reference wire 221 and second reference wire 222 each of which is placed in an area in relative proximity to array 210. First reference wire 221 and second reference wire 222 may be redundant and provide reference signals used by one or more signal processing elements of the processing unit of the present invention to process the cellular information detected by one or more electrodes. In an alternative or additional embodiment, a reference electrode, not shown, is integral to array 210, this reference electrode attached to an individual conductor of wire bundle 220. [0104] Each projection 211 of electrode array 210 may include a single electrode, such as an electrode at the tip of the projection 211 , or multiple electrodes along the length of each projection. Each electrode 212 may be used to detect the firing of one or more neurons, as well as to detect other types of cellular signals such as those integrated multicellular signals from clusters of neurons. Additional electrodes, not shown, such as those integrated into subdural grids, scalp electrodes, cuff electrodes, and other electrodes, can also detect cellular signals emanating from the central or peripheral nervous system, or other part of the body generating cellular signals, such that the processing unit uses these signals to produce the processed signals to send to a diagnostic, therapeutic or controlled device, all not shown. Examples of detected signals include but are not limited to: neural spikes, electrocorticogram signals, local field potential signals, electroencephalogram signals, and other signals integrating the sum of tens to millions of neuronal spikes or cellular potential changes. The processing unit may assign one or more specific cellular signals to a specific use, such as a specific use correlated to a patient imagined event. In a preferred embodiment, the one or more cellular signals assigned to a specific use are under voluntary control of the patient. In an alternative embodiment, cellular signals are transmitted to processing unit first portion 130a via wireless technologies, such as infrared communication, such transmissions penetrating the skull of the patient, and obviating the need for wire bundle 220, array wires 221 and any physical conduit passing through skull 260 after the surgical implantation procedure is completed.

[0105] Referring back to Fig. 2, processing unit first portion 130a and processing unit second portion 130b may independently or in combination preprocess the received cellular signals (e.g., impedance matching, noise filtering, or amplifying), digitize them, and further process the cellular signals to extract neural information. Processing unit second portion 130b may then transmit the neural information to an implanted or external device (both not shown), such as a further processing device and/or any device to be controlled by or otherwise utilize the processed multicellular signals. For example, the external device may decode the received neural information into control signals for controlling a prosthetic limb or limb assist device for controlling a computer cursor, or the external device may analyze the neural information for a variety of other purposes.

[0106] Processing unit first portion 130a and processing unit second portion 130b may independently or in combination also conduct adaptive processing of the received cellular signals by changing one or more parameters of the system to achieve acceptable or improved performance. Examples of adaptive processing include, but are not limited to, changing a parameter during a system configuration, changing a method of encoding neural information such as by changing a neural signal classification parameter, changing the type, subset, or amount of neural information that is processed, or changing a method of decoding neural information. Changing an encoding method may include changing neural spike sorting methodology, calculations, thresholds, or pattern recognition. Changing a decoding methodology may include changing variables, coefficients, algorithms, and/or filter selections. Other examples of adaptive processing may include changing over time the type or combination of types of signals processed, such as EEG, LFP, neural spikes, DC levels, or other signal types.

[0107] Processing unit first portion 130a and processing unit first portion 130b may independently or in combination also transmit signals to one or more electrodes 212 such as to stimulate the neighboring nerves or other cells. Stimulating electrodes in various locations can be used by processing unit 130 to transmit signals to the central nervous system, peripheral nervous system, other body systems, body organs, muscles, and other tissue or cells. The transmission of these signals is used to perform one or more functions including but not limited to: neural signal modulation enhancement, pain therapy, muscle stimulation, seizure disruption, and patient feedback.

[0108] Processing unit first portion 130a and processing unit second portion 130b independently or in combination include signal processing circuitry to perform one or more functions including but not limited to: amplification, filtering, sorting, conditioning, translating, interpreting, encoding, decoding, combining, extracting, sampling, multiplexing, analog to digital converting, digital to analog converting, mathematically transforming, and otherwise processing cellular signals to generate a control signal for transmission to a controlled device. Processing unit first portion 130a transmits raw or processed cellular information to processing unit second portion 130b through integrated wireless communication means, such as radiofrequency communications, infrared communications, inductive communications, ultrasound communications, and microwave communications. This wireless transfer allows the array 210 and processing unit first portion 130a to be completely implanted under the skin of the patient, avoiding the need for implanted devices that require protrusion of a portion of the device through the skin surface. Processing unit first portion 130a may further include a coil, not shown, which can receive power, such as through inductive coupling, on a continual or intermittent basis from an external power transmitting device as has been described in detail in reference to Fig. 1. In addition to or in place of power transmission, this integrated coil and its associated circuitry may receive information from an external coil whose signal is modulated in correlation to a specific information signal. The power and information can be delivered to processing unit first portion 130a simultaneously such as through simple modulation schemes in the power transfer that are decoded into information for processing unit first portion 130 to use, store, or facilitate another function. A second information transfer means, in addition to a wireless means such as an infrared led, can be accomplished by modulating a signal in the coil of processing unit first portion 130a that information is transmitted from the implant to an external device including a coil and decoding elements.

[0109] In an alternative embodiment, not shown, processing unit first portion 130a, and potentially additional signal processing functions are integrated into array 210, such as through the use of a bonded electronic microchip. In another alternative embodiment, processing unit first portion 130a may also receive non-neural cellular signals and/or other biologic signals, such as from an implanted sensor. These signals may be in addition to the neural multicellular signals received from sensor 210, and they may include but are not limited to: EKG signals, respiration signals, blood pressure signals, electromyographic activity signals, and glucose level signals. Such biological signals may be used to turn the biological interface system of the present invention, or one of its discrete components, on or off, to begin a configuration routine, or to start or stop another system function. In another alternative embodiment, processing unit first portion 130a and processing unit second portion 130b independently or in combination produce one or more additional processed signals, to additionally be transmitted to a diagnostic, therapeutic or controllable device of the present invention or to be transmitted to a separate device.

[0110] In an alternative embodiment, a discrete component such as a sensor of the present invention, is implanted within the cranium of the patient, such as array 210 of Fig. 2, a processing unit, or a portion of a processing unit of the present invention is implanted in the torso of the patient, and one or more discrete components are external to the body of the patient. The processing unit may receive multicellular signals from the sensor via wired communication, including conductive wires and optic fibers, or wireless communication. An external processing unit component can be in close proximity to the implanted processing unit component, yet remain hidden under the patient's clothes during use.

[0111] Each sensor discrete component of the present invention can have as few as a single electrode, with the sensor including multiple sensor discrete components that collectively contain a plurality of electrodes. Each electrode is capable of recording single neuron activity, a plurality of neurons, and/or other electrical activity. In an alternative embodiment, one or more electrodes are included in the sensor to deliver electrical signals or other energy to the tissue neighboring the electrode, such as to stimulate, polarize, hyperpolarize, or otherwise cause an effect on one or more cells of neighboring tissue. Specific electrodes may record cellular signals only, or deliver energy only, and specific electrodes may provide both functions.

[0112] Referring now to Fig. 3, a biological interface system 100' comprises implanted components, not shown, and components external to the body of a patient 500. A sensor for detecting multicellular signals, preferably a two dimensional array of multiple protruding electrodes, may be implanted in the brain of patient 500 in an area such as the motor cortex. In a preferred embodiment, the sensor is placed in an area to record cellular signals that are under voluntary control of the patient or in an area to record cellular signals indicative of a patient's disease state or condition. Alternatively or additionally to the two dimensional array, the sensor may include one or more wires or wire bundles which include a plurality of electrodes. Patient 500 of Fig. 3 is shown as a human being, but other mammals and life forms that produce recordable cellular signals would also be applicable. Patient 500 may be a patient with a spinal cord injury or afflicted with a neurological disease that has resulted in a loss of voluntary control of various muscles within the patient's body. Alternatively or additionally, patient 500 may have lost a limb, and system 100' will include a prosthetic limb as its controlled device. Alternatively or additionally, patient 500 may be afflicted with a neurological or psychological disorder or condition and system 100' will include one or more diagnostic or therapeutic devices that receive processed multicellular signals to diagnose and or provide a therapeutic benefit.

[0113] The sensor electrodes of system 100' can be used to detect various multicellular signals including neural spikes, electrocorticogram signals (ECoG), local field potential (LFP) signals, electroencephalogram (EEG) signals, and other cellular and multicellular signals. The electrodes can detect multicellular signals representing clusters of neurons and provide signals midway between single neuron and electroencephalogram recordings. Each electrode may be capable of recording a combination of signals, including a plurality of neural spikes. Alternatively or additionally the sensor may be placed on the surface of the brain without penetrating, such as to detect local field potential (LFP) signals, or on the scalp to detect electroencephalogram (EEG) signals.

[0114] A portion of the processing unit, such as processing unit second portion 130b receives signals from an implanted processing unit component, such as has been described in reference to Fig. 1 and Fig. 2. Processing unit second portion 130b is located just above the ear of patient 500, such that the data transmitting implanted component is located under the scalp in close proximity to the location of processing unit second portion 130b, as depicted in Fig. 3. Signals are transmitted from the implanted processing unit component to processing unit second portion 130b using wireless transmission means. The processing unit components of system 100' perform various signal processing functions including but not limited to: amplification, filtering, sorting, conditioning, translating, interpreting, encoding, decoding, combining, extracting, sampling, multiplexing, analog to digital converting, digital to analog converting, mathematically transforming, and/or otherwise processing cellular signals to generate a control signal for transmission to a controllable device.

The processing unit may process signals that are mathematically combined, such as the combining of neural spikes that are first identified using manual, semi-automated and/or automated neural spike discrimination methods, such as the method of the present invention. In alternative embodiments, the processing unit may comprise three or more components or a single component, and each of the processing unit components can be fully implanted in patient 500, be external to the body, or be implanted with a portion of the component exiting through the skin.

[0115] In Fig. 3, one controlled device is a computer, such as CPU 305 which is attached to monitor 302. In a preferred embodiment, patient 500 can control cursor 303 of CPU 305 and potentially other functions of the computer such as turning it on and off, keyboard entry, joystick control, or control of another input device, each function individually or in combination. System 100' includes another controlled device, such as wheelchair 310. Numerous other controlled devices can be included in the systems of this application, individually or in combination, including but not limited to: a computer; a computer display; a mouse; a cursor; a joystick; a personal data assistant; a robot or robotic component; a computer controlled device; a teleoperated device; a communication device or system; a vehicle such as a wheelchair; an adjustable bed; an adjustable chair; a remote controlled device; a Functional Electrical Stimulator device or system; a muscle stimulator; an exoskeletal robotic brace; an artificial or prosthetic limb; a vision enhancing device; a vision restoring device; a hearing enhancing device; a hearing restoring device; a movement assist device; medical therapeutic equipment such as a drug delivery apparatus; medical diagnostic equipment such as epilepsy monitoring apparatus; other medical equipment such as a bladder or bowel control device; closed loop medical equipment and other controllable devices applicable to patients with some form of paralysis or diminished function as well as any device that may be utilized under direct brain or thought control in either a healthy or unhealthy patient.

[0116] The sensor is connected via a multi-conductor cable implanted in patient 500 to an implanted portion of the processing unit which includes some signal processing elements as well as wireless communication means as has been described in detail in reference to Fig. 1 and Fig. 2. The implanted multi- conductor cable preferably includes a separate conductor for each electrode, as well as additional conductors to serve other purposes, such as providing reference signals and ground.

[0117] Processing unit second portion 130b includes various signal processing elements including but not limited to: amplification, filtering, sorting, conditioning, translating, interpreting, encoding, decoding, combining, extracting, sampling, multiplexing, analog to digital converting, digital to analog converting, mathematically transforming, and/or otherwise processing cellular signals to generate a control signal for transmission to a controllable device. Processing unit second portion 130b includes a unique electronic identifier, such as a unique serial number or any alphanumeric or other retrievable, identifiable code associated uniquely with the system 100' of patient 500. The unique electronic identifier may take many different forms in processing unit second portion 130b, such as a piece of electronic information stored in a memory module; a semiconductor element or chip that can be read electronically via serial, parallel, or telemetric communication; pins or other conductive parts that can be shorted or otherwise connected to each other or to a controlled impedance, voltage or ground, to create a unique code; pins or other parts that can be masked to create a binary or serial code; combinations of different impedances used to create a serial code that can be read or measured from contacts, features that can be optically scanned and read by patterns and/or colors; mechanical patterns that can be read by mechanical or electrical detection means or by mechanical fit, a radio frequency identifier or other frequency spectral codes sensed by radiofrequency or electromagnetic fields, pads and/or other marking features that may be masked to be included or excluded to represent a serial code, or any other digital or analog code that can be retrieved from the discrete component.

[0118] Alternatively or in addition to embedding the unique electronic identifier in processing unit second portion 130b, the unique electronic identifier can be embedded in one or more implanted discrete components. Under certain circumstances, processing unit second portion 130b or another external or implanted component may need to be replaced, temporarily or permanently. Under these circumstances, a system compatibility check between the new component and the remaining system components can be confirmed at the time of the repair or replacement surgery through the use of the embedded unique electronic identifier. [0119] The unique electronic identifier can be embedded in one or more of the discrete components at the time of manufacture, or at a later date such as at the time of any clinical procedure involving the system, such as a surgery to implant the sensor electrodes into the brain of patient 500. Alternatively, the unique electronic identifier may be embedded in one or more of the discrete components at an even later date such as during a system configuration such as a calibration procedure.

[0120] Referring again to Fig. 3, processing unit second portion 130b communicates with one or more discrete components of system 100' via wireless communication means. Processing unit second portion 130b communicates with selector module 400, a component utilized to select the specific device to be controlled by the processed signals of system 100'. Selector module 400 includes an input element 402, such as a set of buttons, used to perform the selection process. The functionality of selector module 400 has been described in detail in reference to Fig. 1. Processing unit second portion 130b also communicates with controlled device CPU 305, such as to control cursor 303 or another function of CPU 305. Processing unit second portion 130b also communicates with processing unit third portion 130c. Processing unit third portion 130c provides additional signal processing functions, as have been described above, to control wheelchair 310. System 100' of Fig. 3 utilizes selector module 400 to select one or more of CPU 305, wheelchair 310, or another controlled device, not shown, to be controlled by the processed signals produced by the processing unit of the present invention. System 100' also includes a modality wherein one set of processed signals emanate from one portion of the processing unit, such as processing unit second portion 130b, and a different set of processed signals emanate from a different portion of the processing unit, such as processing unit third portion 130c.

[0121] The various components of system 100' communicate with wireless transmission means, however it should be appreciated that physical cables can be used to transfer information alternatively or in addition to wireless means. These physical cables may include electrical wires, optical fibers, sound wave guide conduits, and other physical means of transmitting data and/or power, and any combination of those means. [0122] A qualified individual, such as operator 110, may perform a configuration of system 100' at some time during the use of system 100, preferably soon after implantation of the sensor. In a preferred embodiment, at least one configuration routine is performed and successfully completed by operator 110 prior to use of system 100' by patient 500. As depicted in Fig. 3, operator 110 utilizes configuration apparatus 120 which includes first configuration monitor 122a, second configuration monitor 122b, configuration keyboard 123, and configuration CPU 125, to perform a calibration routine or other system configuration process such as system training, algorithm and algorithm parameter selection, and output device setup. The software programs and hardware required to perform the configuration can be included in the processing unit, such as processing unit second portion 130b, be included in selector module 400, or be incorporated into configuration apparatus 120. Configuration apparatus 120 may include additional input devices, such as a mouse or joystick, not shown. Configuration apparatus 120 may include various elements, functions and data including but not limited to: memory storage for future recall of configuration activities, operator qualification routines, template or standard human data, template or standard synthesized or artificial data, neural spike discrimination software, operator security and access control, controlled device data, wireless communication means, remote (such as via the Internet) configuration communication means, and other elements, functions, and data used to provide an effective and efficient configuration on a broad base of applicable patients and a broad base of applicable controlled devices. The unique electronic identifier can be embedded in one or more of the discrete components at the time of system configuration, including the act of identifying a code that was embedded into a particular discrete component at its time of manufacture, and embedding that code in a different discrete component. In an alternative embodiment, all or part of the functionality of configuration apparatus 120 is integrated into selector module 400 such that system 100' can perform one or more configuration processes such as a calibration procedure utilizing selector module 400 without the availability of configuration apparatus 120.

[0123] In a preferred embodiment, an automatic or semi-automatic configuration function or routine is embedded in system 100'. This embedded configuration routine can be used in place of a configuration routine performed manually by operator 110 as is described hereabove, or can be used in conjunction with one or more manual configurations. Automatic and/or semiautomatic configuration events can take many forms including but not limited to: monitoring of cellular activity, wherein the system automatically changes which particular signals are chosen to produce the processed signals; running parallel algorithms in the background of the one or more algorithms currently used to create the processed signals, and changing one or more algorithms when improved performance is identified in the background event; monitoring of one or more system functions, such as alarm or warning condition events or frequency of events, wherein the automated system shuts down one or more functions and/or improves performance by changing a relevant variable; and other methods that monitor one or more pieces of system data, identify an issue or potential improvement, and determine new parameters that would reduce the issue or achieve an improvement. In a preferred embodiment of the disclosed invention, when specific integrated parameters are identified, by an automated or semi-automated calibration or other configuration routine, to be modified for the reasons described above, an integral permission routine of the system requires approval of a specific operator when one or more of the integrated parameters are modified.

[0124] Operator 110 may be a clinician, technician, caregiver, patient family member, or even the patient themselves in some circumstances. Multiple operators may be needed or required to perform a configuration or approve a modification of an integrated parameter, and each operator may be limited by system 100', via passwords and other control configurations, to only perform or access specific functions. For example, only the clinician may be able to change specific critical parameters, or set upper and lower limits on other parameters, while a caregiver, or the patient, may not be able to access those portions of the configuration procedure or the permission procedure. The configuration procedure includes the setting of numerous parameters needed by system 100' to properly control one or more controlled devices. The parameters include but are not limited to various signal conditioning parameters as well as selection and de-selection of specific multicellular signals for processing to generate the device control creating a subset of signals received from the sensor to be processed. The various signal conditioning parameters include, but are not limited to, threshold levels for amplitude sorting, other sorting and pattern recognition parameters, amplification parameters, filter parameters, signal conditioning parameters, signal translating parameters, signal interpreting parameters, signal encoding and decoding parameters, signal combining parameters, signal extracting parameters, mathematical parameters including transformation coefficients, and other signal processing parameters used to generate a control signal for transmission to a diagnostic, therapeutic and/or patient thought-controlled device.

[0125] The configuration routine will result in the setting of various configuration output parameters, all such parameters to be considered integrated parameters of the system of the present invention. Configuration output parameters may comprise but are not limited to: electrode selection, cellular signal selection, neural spike classification, electrocorticogram signal selection, local field potential signal classification, electroencephalogram signal classification, sampling rate by signal, sampling rate by group of signals, amplification by signal, amplification by group of signals, filter parameters by signal, and filter parameters by group of signals. In a preferred embodiment, the configuration output parameters are stored in memory in one or more discrete components, and the parameters are linked to the system's unique electronic identifier.

[0126] Calibration and other configuration routines, including manual, automatic, and semi-automatic routines, may be performed on a periodic basis, and may include the selection and deselection of specific cellular signals over time. The initial configuration routine may include initial values, or starting points, for one or more of the configuration output parameters. Setting initial values of specific parameters, may invoke a permission routine. Subsequent configuration routines may involve utilizing previous configuration output parameters that have been stored in a memory storage element of system 100'. Subsequent configuration routines may be shorter in duration than an initial configuration and may require less patient involvement. Subsequent configuration routine results may be compared to previous configuration results, and system 100' may require a repeat of configuration if certain comparative performance is not achieved.

[0127] The configuration routine may include the steps of (a) setting a preliminary set of configuration output parameters; (b) generating processed signals to transmit to a diagnostic, therapeutic and/or patient thought-controlled device; (c) measuring the performance of the system; and (d) modifying the configuration output parameters. The configuration routine may further include the steps of repeating steps (b) through (d). The configuration routine may also require invoking the permission routine of the present invention.

[0128] In the performance of the configuration routine, the operator 110 may involve patient 500 or perform steps that do not involve the patient. The operator 110 may have patient 500 imagine one or more particular movements, imagined states, or other imagined events, such as a memory, an emotion, the thought of being hot or cold, or other imagined event not necessarily associated with movement. In an alternative embodiment, a specific patient condition is monitored, such as an epileptic seizure or state of depression, during a configuration routine. The patient participation may include the use of one or more cues such as audio cues, visual cues, olfactory cues, and tactile cues. The patient 500 may be asked to imagine multiple movements, and the output parameters selected during each movement may be compared to determine an optimal set of output parameters. The imagined movements may include the movement of a part of the body, such as a limb, arm, wrist, finger, shoulder, neck, leg, angle, and toe, and imagining moving to a location, moving at a velocity or moving at an acceleration. The patient may imagine the movement while viewing a video or animation of a person performing the specific movement pattern. In a preferred embodiment, this visual feedback is shown from the patient's perspective, such as a video taken from the person performing the motion's own eye level and directional view. Multiple motion patterns and multiple corresponding videos may be available to improve or otherwise enhance the configuration process. The configuration routine correlates the selected movement with modulations in the multicellular signals received from the sensor, such as by correlating the periodicity of the movement with a periodicity found in one or more cellular signals. Correlations can be based on numerous variables of the motion including but not limited to position, velocity, and acceleration. In an alternative embodiment, the patient may receive a medication, or have a cessation of medication delivery, prior to and/or during a configuration process.

[0129] The configuration routine will utilize one or more configuration input parameters to determine the configuration output parameters. In addition to the cellular signals themselves, system or controlled device performance criteria can be utilized. Other configuration input parameters include various properties associated with the cellular signals including one or more of: signal to noise ratio, frequency of signal, amplitude of signal, neuron firing rate, average neuron firing rate, standard deviation in neuron firing rate, modulation of neuron firing rate as well as a mathematical analysis of any signal property including but not limited to modulation of any signal property. Additional configuration input parameters include but are not limited to: system performance criteria, controlled device electrical time constants, controlled device mechanical time constants, other controlled device criteria, types of electrodes, number of electrodes, patient activity during configuration, target number of signals required, patient disease state, patient condition, patient age, and other patient parameters and event based (such as a patient imagined movement event) variations in signal properties including neuron firing rate activity. In a preferred embodiment, one or more configuration input parameters are stored in memory and linked to the embedded, specific, unique electronic identifier. All configuration input parameters shall be considered an integrated parameter of the system of the present invention.

[0130] It may be desirous for the configuration routine to exclude one or more cellular signals based on a desire to avoid signals that respond to certain patient active functions, such as non-paralyzed functions, or even certain imagined states. The configuration routine may include having the patient imagine a particular movement or state, and based on sufficient signal activity such as neuron firing rate or modulation of firing rate, exclude that signal from the signal processing based on that particular undesired imagined movement or imagined state. Alternatively, real movement accomplished by the patient may also be utilized to exclude certain cellular signals emanating from specific electrodes of the sensor. In a preferred embodiment, an automated or semi- automated calibration or other configuration routine may include through addition, or exclude through deletion, a signal based on insufficient activity during known patient movements.

[0131] Patient 500 of Fig. 3 can be a quadriplegic, a paraplegic, an amputee, a spinal cord injury victim, or a physically impaired person. Alternatively or in addition, patient 500 may have been diagnosed with one or more of: obesity, an eating disorder, a neurological disorder, a psychiatric disorder, a cardiovascular disorder, an endocrine disorder, sexual dysfunction, incontinence, a hearing disorder, a visual disorder, sleeping disorder, a movement disorder, a speech disorder, physical injury, migraine headaches, or chronic pain. System 100' can be used to diagnose or treat one or more medical conditions of patient 500, or to restore, partially restore, replace, or partially replace a lost function of patient 500.

[0132] Alternatively, system 100 can be utilized by patient 500 to enhance performance, such as if patient 500 did not have a disease or condition from which a therapy or restorative device could provide benefit, but did have an occupation wherein thought control of a device provided an otherwise unachieved advancement in healthcare, crisis management, and national defense.

[0133] The systems of the present invention, such as system 100' of Fig. 3, include a processing unit that processes multicellular signals received from patient 500. Processing unit second portion 130b and other processing unit components, singly or in combination, perform one or more functions. The functions performed by the processing unit include but are not limited to: producing the processed signals; transferring information to a separate device; receiving information from a separate device; producing processed signals for a second device; activating an alarm, alert or warning; shutting down a part of or the entire system; ceasing transmission of processed signals to a device; storing information, and performing a configuration.

[0134] In order for the processing unit of system 100' to perform one or more functions, one or more integrated parameters are utilized. These parameters include pieces of information stored in, sent to, or received from, any component of system 100, including but not limited to: the sensor; a processing unit component; processing unit second portion 130b; or a controlled device. Parameters can be received from devices outside of system 100' as well, such as configuration apparatus 120, a separate medical therapeutic or diagnostic device, a separate Internet based device, or a separate wireless device. These parameters can be numeric or alphanumeric information, and can change over time, either automatically or through an operator involved configuration or other procedure. [0135] In order to change an integrated parameter, system 100' includes a permission routine, such as an embedded software routine or software driven interface that allows the operator to view information and enter data into one or more components of system 100. The data entered must signify an approval of the parameter modification in order for the modification to take place. Alternatively, the permission routine may be partially or fully located in a separate device such as configuration apparatus 120 of Fig. 3, or a remote computer such as a computer that accesses system 100' via the Internet or utilizing wireless technologies. In order to access the permission routine and/or approve the modification of the integrated parameters, a password or security key, such as a mechanical, electrical, electromechanical, or software based security key, may be required of the operator. Multiple operators may be needed or required to approve a parameter modification. Each specific operator or operator type may be limited by system 100', via passwords and other control configurations, to approve the modification of only a portion of the total set of modifiable parameters of the system. Additionally or alternatively, a specific operator or operator type may be limited to only approve a modification to a parameter within a specific range of values, such as a range of values set by a clinician when the operator is a family member. Operator or operator types, hereinafter operator, include but are not limited to: a clinician, primary care clinician, surgeon, hospital technician, system 100' supplier or manufacturer technician, computer technician, family member, immediate family member, caregiver, and patient.

[0136] Referring now to Fig. 4, a biological interface system 100" comprises implanted components and components external to the body of patient 500. System 100" includes multiple controlled devices, such as controlled computer 305, first controlled device 300a, and second controlled device 300b. While three controlled devices are depicted, this particular embodiment includes any configuration of two or more controlled devices for a single patient. Each controlled device may be a diagnostic, therapeutic and/or patient thought-controlled device. First controlled device 300a and second controlled device 300b can include various types of devices such as prosthetic limbs or limb assist devices, robots or robotic devices, communication devices, computers, and other diagnostic, therapeutic and/or controllable devices as have been described in more detail hereabove. The multiple controlled devices can include two or more joysticks or simulated joystick interfaces, two or more computers, a robot and another device, and many other combinations and multiples of devices as have been described in detail hereabove. Each controlled device includes one or more discrete components or is a portion of a discrete component.

[0137] A sensor 200 for detecting multicellular signals, preferably a two dimensional array of multiple protruding electrodes, has been implanted in the brain of patient 500 in an area such as the motor cortex. In a preferred embodiment, the sensor 200 is placed in an area to record cellular signals that are under voluntary control of the patient. Alternatively or additionally to the two dimensional array, the sensor may include: an additional array; one or more wires or wire bundles which include a plurality of electrodes; subdural grids; cuff electrodes; scalp electrodes; or other single or multiple electrode configurations. Sensor 200 is attached to transcutaneous connector 165 via wiring 216, a multi-conductor cable that preferably, though not necessarily, includes a separate conductor for each electrode of sensor 200. Transcutaneous connector 165 includes a pedestal which is attached to the skull of the patient such as with glues and/or bone screws, preferably in the same surgical procedure in which sensor 200 is implanted in the brain of patient 500. Electronic module 170 attaches to transcutaneous connector 165 via threads, bayonet lock, magnetic coupling, velcro, or other engagement means. Transcutaneous connector 165 and/or electronic module 170 may include integrated electronics including but not limited to signal amplifier circuitry, signal filtration circuitry, signal multiplexing circuitry, and other signal processing circuitry, such that transcutaneous connector 165 and/or electronic module 170 provide at least a portion of the processing unit of the disclosed invention. Transcutaneous connector 165 preferably includes electrostatic discharge protection circuitry. Electronic module 170 includes wireless information transfer circuitry, utilizing one or more of radiof requency, infrared, ultrasound, microwave, or other wireless communication means. In an alternative embodiment, transcutaneous connector 165 includes all the appropriate electronic signal processing, electrostatic discharge protection circuitry, and other circuitry, and also includes wireless transmission means, such that the need for electronic module 170 is obviated. [0138] In a preferred embodiment, electronic module 170 includes wireless transmission means and a power supply, not shown, such that, as the power supply is depleted or electronic module 170 has a malfunction, it can be easily replaced. In another preferred embodiment, electronic module 170 is a disposable component of system 100". Electronic module 170 transmits information to processing unit transceiver 135 which is integrated into a portion of system 100"s processing unit, such as processing unit first portion 130a. In a preferred embodiment, processing unit transceiver 135 is a two-way wireless communication device, and electronic module 170 is also a two-way wireless communication device such that information can be sent to or from electronic module 170.

[0139] All of the physical cables of Fig. 4, as well as all the other figures of this disclosure, can be in a permanently attached, or in a detachable form. In addition, all of the physical cables included in system 100" of Fig. 4 as well as the systems of the other included figures can be eliminated with the inclusion of wireless transceiver means incorporated into the applicable, communicating discrete components. Processing unit first portion 130a, a discrete component as defined in this disclosure, includes various signal processing functions as has been described in detail in relation to separate figures hereabove. Processing unit first portion 130a preferably includes a unique system identifier, the makeup and applicability of the unique identifier also described in detail hereabove. Processing unit first portion 130a electrically connects to processing unit second portion 130b via intra- processing unit cable 140. Cable 140 is detachable from processing unit second portion 130b via female plug 153 which is attached to processing unit second portion 130b at its input port, male receptacle 152. Cable 140 may be constructed of electrical wires and/or fiber optic cables. In a preferred embodiment, data is transmitted from processing unit first portion 130a to processing unit second portion 130b via a fiber optic cable. Information and other signals transmitted between processing unit first portion 130a and processing unit second portion 130b may be in analog format, digital format, or a combination of both. In addition, wireless transmission of information can be provided, not shown, to replace intraprocessing unit cable 140 or work in conjunction with intraprocessing unit cable 140. [0140] Processing unit second portion 130b includes further signal processing means which in combination with the signal processing of processing unit first portion 130a produces processed signals, such as to be transmitted to multiple diagnostic, therapeutic and/or patient thought-controlled devices. Processing unit first portion 130a and/or processing unit second portion 130b include various functions including but not limited to: a neural spike classifier function, such as a threshold based neural spike sorting function or the neural signal classifier of the present invention; an amplifier function; a signal filtering function; a neural net software function; a mathematical signal combination function; and a database storage and retrieval function such as a database including a list of acceptable neural information or a database of unacceptable neural information each of which can be used to perform a system diagnostic. In another preferred embodiment, the processing unit assigns one or more cellular signals to a specific use, such as a specific use that is correlated to a patient imagined event.

[0141] The processed signals emanating from processed unit second portion 130b can be analog signals, digital signals, or a combination of analog and digital signals. The processing unit of the present invention may include digital to analog conversion means as well as analog to digital conversion means. The processed signals can be transmitted to one or more devices with a hardwired connection, a wireless connection or a combination of both technologies. As depicted in Fig. 4, controlled computer 305, first controlled device 300a, and second controlled device 300b are controlled by the processed signals produced by processing unit first portion 130a and processing unit second portion 130b. Similar to processing unit first portion 130a, processing unit second portion 130b preferably includes the system unique electronic identifier, which can be embedded in processing unit second portion 130b at the time of manufacture, during installation procedures, during calibration or other post-surgical configuration procedures, or at a later date.

[0142] The three controlled devices are shown permanently attached to physical cables, with each physical cable including a removable connection at the other end. Controlled computer 305 is attached to cable 311 that has female plug 155 at its end. First controlled device 300a is attached to first controlled device cable 301a which has female plug 159 at its end. Second controlled device 300b is attached to second controlled device cable 301b which has female plug 157 at its end. Each physical cable can be attached and detached from processing unit second portion 130b. Female plug 159 attaches to male receptacle 158; female plug 157 attaches to male receptacle 156, and female plug 155 attaches to male receptacle 154.

[0143] Each of controlled computer 305, first controlled device 30Oa₁ and second controlled device 300b preferably has embedded within it a unique identifier of the particular device. Additional codes, such as the unique system identifier, may also be embedded. When any of the physical cables are first attached, such as controlled computer cable 311 being attached via female plug 157 to male receptacle 156, a compatibility check is performed by system 100" to assure that the unique system identifier embedded in controlled computer 305 is identical or otherwise compatible with a unique electronic identifier embedded in any and all other discrete components of system 100", such as the unique electronic identifier embedded in processing unit second portion 130b. Similar system compatibility checks can be performed with the attachment of first controlled device 300a or second controlled device 300b. If improper compatibility is determined by system 100", various actions that can be taken include but are not limited to: entering an alarm state, displaying incompatibility information, transmitting incompatibility information, deactivation of controlled device control, limiting controlled device control, and other actions.

[0144] Also depicted in Fig. 4 is selector module 400 which can be used by the patient or a different operator, such as a clinician, to select one or more specific devices to be controlled by the processed signals of system 100". Selector module 400 includes numerous elements and functional capability as has been described in detail in relation to Fig. 1. Selector module 400 is shown with a data entry keypad, input element 402, and an output element 403, such as a visual display. Input element 402 is used by an operator to select the specific controlled device, and to perform other data entry. Output element 403 provides information to the operator such as selectable controlled device icons, controlled device information, and other system information. Selector module 400 communicates with processing unit first portion 130a via wireless technology, information transfer means 410. After selection of the one or more controlled devices to be controlled by the processed signals, these processed signals include one or more unique codes identifying the selected controlled device or devices, and may additionally include the unique system identifier. These codes can be sent at the initiation or cessation of control or on a periodic or continuous basis in order to assure that only the selected devices are controlled or are otherwise influenced by the processed signals. A selection event can either cause a controlled device to begin to be controlled or stop the control of a controlled device that is already being controlled. In a preferred embodiment, specific operators can select specific equipment, such conditional matrix stored in a memory module of selector module 400 or other discrete component of system 100".

[0145] Selector module 400 may include access passwords or require mechanical or electronic keys to prevent unauthorized use, and may also include a function, such as a permission routine function, to select a controlled device to modify its control. Selector module 400 may have other integrated functions such as information recall functions, system configuration, or calibration functions, as well as a calculator, cellular telephone, pager, or personal data assistant (PDA) functions. Clinician control unit 400 may be a PDA that has been modified to access system 100" to select one or more controlled device to modify its control, such as through the use of a permission routine.

[0146] Selector module 400 of Fig. 4 includes an integrated monitor for displaying the information, however in an alternative embodiment, the selector module 400 can cause the information to be displayed on a separate visualization apparatus such as the monitor of controlled computer 305. Alternatively or additionally, one or more of the functions of the selector module 400 can be integrated into one or more discrete components of system 100".

[0147] Numerous configurations and types of controlled devices can be used with system 100" of Fig. 4. Numerous types of controlled devices have been described in detail in relation to the systems of Fig. 1 and Fig. 3 and are applicable to system 100" of Fig. 4 as well. System 100" works with a single patient 500 who can control multiple controlled devices such as controlled computer 305, first controlled device 300a, and second controlled device 300b. In a preferred embodiment, patient 500 can select and/or control more than one controlled device simultaneously. While each controlled device is connected to the same discrete component, such as processing unit second portion 130b, in an alternative embodiment, the multiple controlled devices can be connected to multiple processing unit discrete components. In that embodiment, the selector module 400 is used to start or stop the transmission of the individual processing units to their corresponding controlled device.

[0148] While patient 500 has been implanted with a sensor 200 including a single discrete component, sensor 200 may comprise multiple discrete components, not shown, such as multiple electrode arrays, implanted in different parts of the brain, or in other various patient locations to detect cellular signals. Cellular signals from the individual sensor discrete components, such as a single electrode component, may be sent to individual processing units, or to a single processing unit. Separate processed signals can be created from each individual discrete component of the sensor, and those particular signals tied to a specific controlled device. Thus, each controlled device can be controlled by processed signals from a different sensor discrete assembly, such as discrete components at different locations in the brain or other parts of the body. It should be appreciated that any combination of discrete component cellular signals can be used in any combination of multiple controlled devices. Alternatively, whether the sensor is embodied in a single discrete component or multiple discrete components, the processed signals for individual controlled devices may be based on specific cellular signals or signals from specific electrodes, such that individual device control is driven by specific cellular signals. Any combination of exclusively assigned cellular signals and shared cellular signals used to create processed signals for multiple controlled devices are to be considered within the scope of this application. In an alternative, preferred embodiment, the system includes multiple patients, these patients collectively selecting and/or controlling one or more controlled devices.

[0149] The system 100" of Fig. 4 may include two or more separate configuration routines, such as a separate calibration routine for each controlled device. Any and all discrete components of system 100" may have a unique electronic identifier embedded in it. The processing unit of system 100", comprising processing unit first portion 130a and processing unit second portion 130b, may conduct adaptive processing as has been described hereabove.

[0150] The unique electronic identifier of the system is a unique code used to differentiate one system, such as the system of a single patient, from another system, as well as to differentiate all discrete components of a system, especially detachable components, from discrete components of a separate, potentially incompatible system. The unique electronic identifier may be a random alphanumeric code or may include information including but not limited to: patient name, other patient information, system information, implant information, number of electrodes implanted, implant location or locations, software revisions of one or more discrete components, clinician name, date of implant, date of calibration, calibration information, manufacturing codes, and hospital name. In a preferred embodiment, the unique electronic identifier is stored in more than one discrete component such as a sensor discrete component and a processing unit discrete component. The unique electronic identifier may be programmable, such as one time programmable, or allow modifications for multiple time programming, such programming performed in the manufacturing of the particular discrete component, or by a user at a later date. The unique electronic identifier may be configured to be changed over time, such as after a calibration procedure. The unique electronic identifier can be permanent or semi-permanent, or hard wired, such as a hard wired configuration in a transcutaneous connector of the system. The unique electronic identifier can be used in wireless communications between discrete components, or in wireless communications between one or more discrete components and a device outside of the system. The unique electronic identifier can represent or be linked to system status. System status can include but not be limited to: output signal characteristics, level of accuracy of output signal, output signal requirements, level of control needed, patient login settings, such as customized computer configuration information, one or more software revisions, one or more hardware revisions, controlled device compatibility list, patient permissions lists, and calibration status. In a preferred embodiment, the unique identifier includes information to identify the system as a whole, as well as information identifying each discrete component, such as each controlled device applicable to the system. The unique portion identifying each controlled device can be used in wireless communication, after a selection has been made via the selector module, such that the selected controlled devices are properly controlled.

[0151] The system 100" of Fig. 4 may include a library of various integrated parameters, such integrated parameters utilized by the processing units, processing unit first portion 130a and processing unit second portion 130b to perform a function including but not limited to the creation of the processed signals to control one or more controlled devices. Integrated parameters include various pieces of system data, such as data stored in electronic memory. In a preferred embodiment, the data is electronically linked with the unique electronic identifier of system 100". The integrated parameter data may be stored in memory of one or more discrete components, such as processing unit second portion 130b, or alternatively or additionally the integrated parameter data may be stored in a computer based network platform, separate from system 100' such as a local area network (LAN), a wide area network (WAN) or the Internet. The integrated parameter data can contain numerous categories of information related to the system including but not limited to: patient information such as patient name and disease state; discrete component information such as type of sensor and electrode configuration; system configuration information such as calibration dates, calibration output parameters, calibration input parameters, patient training data, signal processing methods, algorithms and associated variables, controlled device information such as controlled device use parameters and lists of controlled devices configured for use with or otherwise compatible with the system; and other system parameters useful in using, configuring and assuring safe and efficacious performance of system 100".

[0152] In an alternative embodiment, system 100" of Fig. 4 further comprises a patient feedback module. The feedback module may include one or more of an audio transducer, a tactile transducer, and a visual display. This patient feedback module may be used during patient or other system training, or at all times that the patient is controlling an external device. Feedback can be used to enhance signal quality and power of the processed signals, as well as to avoid unsafe or undesirable conditions. The feedback module may utilize one or more discrete components of system 100" such as sensor 200. In another preferred embodiment, one or more electrodes of sensor 200 can be stimulated, such as via a stimulation circuit provided by one or more of transcutaneous connector 165 or electronic module 170. The stimulation can evoke a variety of responses including but not limited to the twitching of a patient's finger. The feedback signal sent to the patient can take on a variety of forms, but is preferably a derivative of a modulating variable of the controlled device. For example, feedback can be a derivative of cursor position of controlled computer 305. If audio feedback is implemented, a signal representing horizontal position and a signal representing vertical position can be combined and sent to a standard speaker. Other audio feedback, such as specific discrete sounds, can be incorporated to represent proximity to an icon, etc. Parameters of the feedback module should be considered integrated parameters of the systems of this invention, such that one or more feedback parameters require approval of an operator via the system's permission routine. In a preferred embodiment, the patient feedback function is incorporated into selector module 400 such as via a visual display or audio transducer.

[0153] Patient 500 of Fig. 4 is at a specific location, Location 1. An operator such as a clinician operator 111 is at a location remote from patient 500, Location 2. Also at Location 2 is configuration system 120 which can access system 100" via the Internet as has been described in reference to previous embodiments. Configuration system 120 can be used to perform various configuration procedures such as calibration procedures as has been described in reference to a similar configuration system of Fig. 3. In a preferred embodiment, configuration system 120 can perform the functions of the selector module such that clinician operator 111 can select a specific device to modify its control via configuration apparatus 120 and the Internet.

[0154] Numerous methods are provided in the multiple embodiments of the disclosed invention. A preferred method embodiment includes a method of selecting a specific device to be controlled by the processed signals of a biological interface system. The method comprises: providing a biological interface system for collecting multicellular signals emanating from one or more living cells of a patient and for transmitting processed signals to control a device such as a diagnostic, therapeutic and/or patient thought-controlled device. The biological interface system comprises: a sensor for detecting the multicellular signals, the sensor comprising a plurality of electrodes to allow for detection of the multicellular signals; a processing unit for receiving the multicellular signals from the sensor, for processing the multicellular signals to produce processed signals, and for transmitting the processed signals; a first controlled device for receiving the processed signals; a second controlled device for receiving the processed signals; and a selector module that is used to select the specific device to be controlled by the processed signals. [0155] It should be understood that numerous other configurations of the systems, devices, and methods described herein can be employed without departing from the spirit or scope of this application. It should be understood that the system includes multiple functional components, such as a sensor for detecting multicellular signals, a processing unit for processing the multicellular signals to produce processed signals, and the controlled device that is controlled by the processed signals. Different from the logical components are physical or discrete components, which may include a portion of a logical component, an entire logical component, and combinations of portions of logical components and entire logical components. These discrete components may communicate or transfer information to or from each other, or communicate with devices outside the system. In each system, physical wires, such as electrical wires or optical fibers, can be used to transfer information between discrete components, or wireless communication means can be utilized. Each physical cable can be permanently attached to a discrete component, or can include attachment means to allow attachment and potentially allow, but not necessarily permit, detachment. Physical cables can be permanently attached at one end, and include attachment means at the other.

[0156] The sensors of the systems of this application can take various forms, including multiple discrete component forms, such as multiple penetrating arrays that can be placed at different locations within the body of a patient. The processing unit of the systems of this application can also be contained in a single discrete component or multiple discrete components, such as a system with one portion of the processing unit implanted in the patient, and a separate portion of the processing unit external to the body of the patient. The sensors and other system components may be utilized for short term applications, such as applications less than twenty four hours, sub-chronic applications such as applications less than thirty days, and chronic applications. Processing units may include various signal conditioning elements such as amplifiers, filters, signal multiplexing circuitry, signal transformation circuitry, and numerous other signal processing elements. In a preferred embodiment, an integrated neural signal classification function is included. The processing units perform various signal processing functions including but not limited to: amplification, filtering, sorting, conditioning, translating, interpreting, encoding, decoding, combining, extracting, sampling, multiplexing, analog to digital converting, digital to analog converting, mathematically transforming and/or otherwise processing cellular signals to generate a control signal for transmission to a controllable device. Numerous algorithms and/or mathematical and software techniques can be utilized by the processing unit to create the desired control signal. The processing unit may utilize neural net software routines to map cellular signals into desired device control signals. ' Individual cellular signals may be assigned to a specific use in the system. The specific use may be determined by having the patient attempt an imagined movement or other imagined state, or from patient-attempted control using a partially built decoding mechanism. For restoration of lost function applications such as an ALS patient operating a communication device or a spinal cord injury patient controlling a wheelchair, it is preferred that that the cellular signals be under the voluntary control of the patient. The processing unit may mathematically combine various cellular signals to create a processed signal for device control.

[0157] According to an exemplary embodiment, the disclosed biological interface systems may include a neural signal classification system 501 for processing one or more cellular signals. As explained, cellular signals may include, among other things, a neural spike indicative of voluntary or involuntary neural activity of patient 500, local field potential signals associated with extracellular activity, and noise signals caused by one or more environmental noise sources, injected by one or more components of the biological interface system, and/or related to one or more environmental variables associated with patient 500. Thus, according to one aspect, the present disclosure is directed toward efficiently isolating the neural signal information from the sensor-recorded signals and extracting the neural spike occurrences without requiring large amounts of technician (manual) activity, computer processing power and/or computer memory capabilities. The neural spike occurrences may be associated with a particular voluntary activity or desired activity of the patient, or may be indicative of a present medical condition.

[0158] Furthermore, systems and methods associated with the disclosed embodiments provide a biological interface system that efficiently adapts to changes in one or more cellular signals. For example, a neural signal classification system may employ adaptive filtering techniques to suppress noise levels injected by one or more components associated with a biological interface system based on real-time noise information collected from the one or more components.

[0159] In addition to adaptive filtering, neural signal classification system 501 may provide adaptive techniques for identifying and classifying neural spike activity. For example, neural signal classification system 501 may correlate incoming neural spike signals with a benchmark signal obtained through experimentation and/or historic neural spike data. The correlated samples may be projected onto a feature space that may include one or more previous correlation samples. These samples may be grouped into clusters, each cluster representing a particular neural spike (i.e. represent a specific neuron's firing characteristic). As the feature space becomes more populated with correlated data, neural signal classification system 501 may modify one or more parameters associated with the biological interface system until the samples within the feature space converge into well defined clusters. Each cluster may then be identified and classified as a unique neuron's firing characteristic.

[0160] Referring now to Fig. 5, a flow chart of the performance of sequential system configuration routines for the biological interface system is depicted. Step 20 includes the step a providing a system configuration plan, this plan including the performance of one or more automated system configurations, such as a configuration performed with or without the assistance of an operator. Step 21 includes the performance of a first automated configuration. Step 22 includes the performance of an analysis of data collected during step 21. Step 23 is an analysis that determines whether the analysis of step 22 was successful. If the analysis of Step 22 was not successful, the first automated configuration is modified, such as the modification of a parameter measured or the modification of a test parameter, and step 21 is repeated. If the analysis of step 22 was successful, step 24 is performed. Step 24 consists of allowing the patient to control one or more controlled devices. Step 25 is performed at a predetermined time interval, when system performance falls below a threshold, or when another system event occurs. Step 25 includes the performance of a second automated configuration. Step 26 includes the performance of an analysis of data collected during step 25. Step 27 is an analysis that determines whether the analysis of step 26 was successful. If the analysis of Step 26 was not successful, the second automated configuration is modified, such as the modification of a parameter measured or the modification of a test parameter, and step 25 is repeated. If the analysis of step 26 was successful, step 28 is performed, and a similar sequence of steps is cyclically repeated. These steps illustrate a preferred embodiment of an adaptive system configuration routine of a biological interface system. Adaptation includes modification of rules, coefficients, thresholds, starting levels, and other parameters of the system configuration routine.

[0161] Fig. 6 illustrates a preferred embodiment of a biological interface system with a configuration routine that includes automated neural spike sorting and a system diagnosis. The results of the automated spike sorting and system diagnosis are summarized, and the output of the summary is used to modify one or more system parameter used in the run mode of the system.

[0162] Fig. 7 illustrates a series of neural signals (left side of drawing) and two graphs illustrating a comparison of a preferred automated spike sorting routine as compared to manual spike sorting.

[0163] Fig. 8 illustrates sets of manual spike sorting outputs versus automated spike sorting outputs as seen in a human patient model. The top pair of spike graphics show similar results with an automated spike sorting routine (left) with a manual spike sorting method (right). The middle pair of spike graphics show an instance in which the automated spike sorting routine differentiated two different neural spikes classified by the manual spike sorting method (right) as a single neuron. The bottom pair of spike graphics show similar results with an automated spike sorting routine (left) with a manual spike sorting method (right). The graph at the right of the pairs of spikes shows the associated correlations between the automated and manual spike sorting methods. Demonstration that patterns are repeated over time is shown.

[0164] Fig. 9 shows cross correlograms of manual and automated spike sorting methods of Fig. 8.

[0165] Referring now to Fig. 10, a flow chart of a preferred automated spike sorting routine is depicted. Fig. 10 shows the system components for processing neural signals to extract spikes and generate parameters suitable for classification, also known as sorting into units. Not all details are illustrated; only broad sections of processing. The front of the system (working from left to right) contains processing elements for removing power line artifacts. Any DC offset present is removed using a high pass filter with very low cut-on frequency. Power line interference is also abated before the signal is passed to further stages.

[0166] The upper half of the first branch in the system diagram simply captures the minimally processed raw data for off-line review. Since noise below 250 Hz is extremely strong, to the point of totally obscuring any spikes, the lower half of the first branch splits the neural signal into two bands using an efficient, combined filter. Above 250 Hz filtered signal is passed onto further spike extraction processing. Below 250 Hz filtered signal is reserved for processing of local field potential (LFP). LFP has frequency content similar to electroencephalogram (EEG) data. Hence, after decimation for sake of efficiency, LFP data is further filtered into bands of conventional acceptance.

[0167] For neural spike extraction (middle branch of diagram), the signal is low pass filtered to 5 kHz because there is little information above 5 kHz to distinguish a spike of one unit (class) from that of another unit (class). This is especially true in electrically noisy environs where a patient may be situated. Having been limited to that frequency range, the signal may be decimated to bring its sampling frequency down to 15 kHz, well above the Nyquist rate.

[0168] Triggering the detection of a spike is very sensitive, and it is desirable to not consume CPU cycles for excessive false positives. Consequently, the frequency content is further band limited from 1 kHz to 5 kHz because electrical noise below 1 kHz can be of high enough energy to easily be mistaken as a trigger for detection. However, that band from 250 kHz to 1 kHz has low enough noise energy that it can serve to classify spikes once they have been detected. As shown in the diagram, this lower frequency portion is set aside using another efficient splitting filter for later use.

[0169] The upper portion of the last stage of processing (right side of figure) is dedicated to detecting spikes. Delay is for alignment purposes. The absolute. value effectively extracts the signal's envelope. A 1.5 kHz low pass filter smoothens that envelope to rid spurious multiple detection of a single spike. Again, decimation for CPU savings is warranted. When said smoothed signal crosses a threshold, a trigger occurs and the spike waveform (tens of samples) is recorded. The threshold is derived from a very low pass filtered envelope, a function of signal energy. In other words, only sharp upward transitions in the envelope are of interest and are accepted for consideration as a spike.

[0170] Upon capture of the signal, its 250 Hz to 1 kHz and 1 kHz to 5 kHz bands are processed into two component parameters to create a feature space for classification. These elements of the system diagram just before the label "feature space" correlate (multiply) the captured waveforms against predefined waveforms chosen to best represent a feature space. By choosing predefined waveforms, the feature space is consistently in one quadrant of a two dimensional space.

[0171] Note an essential element of good clustering in the feature space, without the complicating appearance of vestiges of clusters, is the spike alignment just before correlation in this final stage. Its details are difficult to illustrate. The algorithm searches alignment of a few typically higher strength samples of the waveform by trial and error, alignment position declared as that for highest correlation. By not performing a localized cross-correlation of the complete waveform against the prototype waveforms, CPU cycles are conserved.

[0172] Feature space data is passed on to a sophisticated classification routine, not shown here. The automated spike sorting routine shown can work with much lower signal to noise ratios than manual methods and other methods.

[0173] Fig. 11 shows a flow chart of a preferred embodiment of a configuration routine of a biological interface system. The configuration routine includes the performance of a diagnostic, creating quantitative and/or qualitative measurements of ground and reference signals of the system. Measurements may include an analysis of signal amplitude or impedance. A diagnostic of channel performance (signals received from each electrode) is performed. This channel diagnostic may include a signal to noise ratio measurement, an impedance measurement, a crosstalk (between other channels or signals) measurement, an assessment of quality of channels, an assessment of the number of cellular signals received for each channels, other performance measurements, and combinations thereof. All of the data from these steps is saved, and a summarization of the data is made. The summarization may be provided to an operator, and/or used by the system in an automated configuration. One or more of these diagnostic summaries is here by termed a "fig. of merit" and can be used by the operator and/or an automated system routine, to modify device control such as to limit which devices are appropriate to be used based on the level of the fig. of merit. Alternatively or additionally, channels of data can be turned on or off based on the measurement. Alternatively or additionally, a prediction of future performance, such as time to failure or other inadequate performance, can be made via a trending or other analysis. The described method allows separation of these parameters from signal decoding methods and issues. The described method allows quantification of multiple parameters to a single value.

[0174] Fig. 12 is a schematic of an exemplary embodiment of an impedance diagnostic device 1100. As illustrated in Fig. 12, the impedance diagnostic device may include a system 1105 for measuring impedance of one or more electrodes associated with a biological interface system. The system may include a plurality of channel amplifiers 1110, a reference amplifier 1120, a signal generator 1130, and a processing device 1140. It should be noted that, in addition to being implemented as part of the biological interface system, impedance diagnostic device 1100 and/or impedance measuring system 1105 may be implemented in any system that may require fast impedance diagnostics and/or measurement.

[0175] Channel amplifiers 1110 and reference amplifier 1120 may each include a buffer circuit 1112 and a protective circuit 1114. Buffer circuit 1112 may include any type of circuit with high input impedance and near-unity gain such as, for example, a voltage follower. As such, buffer circuit 1112 may provide an output signal substantially similar to the input signal without overloading the relatively low impedance processing device. Protective circuit 1114 may include any type of device adapted to limit the transmission of low frequency (DC or low frequency AC) signals to processing device 1140 and a feed back to the sensor (212) to avoid charge accumulation. As illustrated in Fig. 12, protective circuit 1114 may include a high pass filter coupled to the input of buffer circuit 1112. Alternatively and/or additionally, protective circuit 1114 may include a bandpass filter, a blocking capacitor, or any other type of device for limiting the transmission of low frequency signals into processing device 1140.

[0176] Channel amplifiers 1110 may each be communicatively coupled to an electrode 212 associated with a sensor 200 that is implanted within a biological system. The term "biological system," as used herein, refers to any type of biological environment wherein one or more cells emanate electrical signals in response to a stimulus. For example, a biological system may include, but not be limited to, a brain, central nervous system, or any other part of the human body, an in vitro cellular sample, cellular tissue of any living organism, or any other type of biological system. Each channel amplifier 1110 may be adapted to receive multicellular signals associated with the biological system and provide the multicellular signals to processing device 1140. According to an exemplary embodiment, one or more channel amplifiers may be mounted on chip or printed circuit board, with each channel amplifier 1110 being coupled to a particular electrode of a multi-electrode array. Furthermore, channel amplifiers 1110 may be connected in parallel with each other to simultaneously measure signals associated with the biological system.

[0177] Reference amplifier 1120 may be communicatively coupled to a reference electrode 1121 that is implanted within the biological system. Reference electrode 1120 may include any type of electrode or wire configured to measure a noise level associated with the biological system. Reference electrode 1121 may be disposed at or near a portion of the brain where a limited amount of neural activity occurs. As such, reference electrode 1121 may collect noise data associated with the biological system. The noise data may be subtracted from the collected multicellular signals to correct for noise in the biological system, thereby reducing and/or eliminating the noise in the multicellular signal.

[0178] Signal generator 1130 may include a signal source for providing impedance testing signals to the biological system. For example, signal generator 1130 may include an AC signal source, such as, for example, a sinusoidal or square wave generator. According to an exemplary embodiment, signal generator 1130 is a pulsed, square wave voltage generator for providing pulsed test signals to the biological system. [0179] Signal generator 1130 may be electrically coupled to the biological system and adapted to periodically provide test signals to the biological system for determining an impedance associated with one or more electrodes 212. For example, signal generator 1130 may be selectively coupled to the biological system by a switching device 1131 arranged in parallel with signal generator 1130. A switch controller (not shown) may actuate switching device 1131 to selectively couple signal generator 1130 to the biological system. Alternatively and/or additionally, signal generator 1130 may periodically provide the test signals in response to a command received from a controller (e.g., processing device 1140). Because signal generator 1130 may be selectively coupled and de-coupled from the biological system, system 1110 may include both a multicellular signal measuring device and an impedance measuring device in a single, integrated unit.

[0180] As illustrated in Fig. 12, signal generator may be communicatively coupled to the biological system via a pedestal 1135. Pedestal 1135 may include any device suitable for mounting on a solid surface, so that pedestal 1135 may be mounted, for example, on the skull of a patient during a surgical procedure. Pedestal 1135 may include one or more electrical connectors or connector conduits for providing a connection interface between one or more devices external to the skull and one or more devices disposed within the skull. For instance, pedestal 1135 may provide a connection interface between system 1105 and at least one of sensor 200 and reference electrode 1121. Alternatively, pedestal 1135 may be configured to perform measurements in vitro, where the pedestal may not be mounted to a patient. Pedestal holder may also be used when performing "real-time" impedance measurements before mounting the pedestal on the skull of a patient in order to determine optimum placement of sensor 200 within the brain of the patient.

[0181] Processing device 1140 may include any type of electronic system for analyzing signals received from channel amplifiers 1110 and reference amplifier 1120. According to one embodiment, processing device 1140 may include a processing unit associated with the biological interface system 110, such as processing unit first and/or second portion 130c or 13Od (see Fig. 3). It is contemplated, however, that processing device 1140 may embody a separate, standalone processing system adapted to analyze impedances associated with one or more electrodes, as part of an impedance diagnostic system.

[0182] Processing device 1140 may be configured to receive electrical signals from channel amplifiers 1110 and reference amplifier 1120. When signal generator 1130 is de-coupled from the biological system and the continuous ground signal is reestablished, processing device 1140 may be configured to receive multicellular signals and noise data from channel amplifiers 1110 and reference amplifier 1120, respectively. When signal generator 1130 is coupled to the biological system, processing device 1140 may receive feedback corresponding to the test signals provided by signal generator 1130. As explained, these signals may be corrected by adjusting each signal respective of noise data collected by reference amplifier 1120.

[0183] According to one exemplary embodiment, processing device 1140 may include a neural signal amplifier comprising a plurality of differential amplifiers for measuring the voltage difference between two inputs. The output of each channel amplifier 1110 may be provided as one input of a corresponding differential amplifier and the output of reference amplifier 1120 may be provided as the other input for each differential amplifier. Accordingly, a differential amplifier may be provided for each channel amplifier 1110. The output of the differential amplifier (i.e., the difference between a signal measured by each electrode and the noise measured by the reference electrode) may correspond to the signal measured by the electrode adjusted to compensate for noise associated with the biological system.

[0184] System 1105 may use a built-in C-R protection circuit of the channel and reference amplifiers to measure the impedance of electrodes 212. According to one exemplary embodiment, the values of RC are 1 nF and 500MOhm. These values are exemplary only and not intended to be limiting. This protection circuit creates a negligible voltage loss for the signal, but can be used as a voltage divider for calculating additional impedances (like the electrode) connected to it. For measuring these small signal changes, a low noise, high gain amplifier system may be required together with a sharp bandpass filter around the generator frequency, what will increase the signal- noise ration for impedance measurement. The band pass filter is implemented by software solution. The low noise processing unit 1140 may detect small signal changes in this magnitude.

[0185] Fig. 13 illustrates an exemplary test circuit associated with a single electrode. As illustrated in Fig. 13, during testing (e.g., when signal generator is coupled to the biological system either via pedestal 1135 or by some other means), the test signal may be injected into the biological system in series with the electrode. The signal may be passed through a protection circuit (such as a high pass RC filter) and provided to an input of a voltage follower. Because R and C have known values, any signal loss in the system is typically associated with the electrode. This signal loss may be measured and used to determine the impedance of the electrode.

[0186] In some exemplary aspects, processes and methods consistent with the disclosed embodiments may provide a multi-purpose system for measuring multicellular signals corresponding to a biological system and measuring the impedance and crosstalk associated with a sensor array to analyze the reliability of the measured neurological activity. A real-time impedance measurement system adapted to quickly provide impedance and crosstalk data may enable technicians to diagnose potential problems with a sensor device, without requiring the technician to perform manual impedance measurements, which can be time consuming. Fig. 14 illustrates a flowchart 1300 depicting an exemplary method of determining the impedance of a plurality of electrodes, in accordance with the present disclosure.

[0187] System 1105 may initiate a sequence for measuring the impedance of a plurality of electrodes (Step 1310). For example, system 1105 may switch the mode of system operation from a multicellular signal measurement mode to an impedance measurement mode by actuating switching device 1131 associated with signal generator 1130. As a result, signal generator 1130 is coupled to the biological system for testing.

[0188] Once system 1105 has been configured for impedance testing, signal generator 1130 may provide a test signal to the biological system (Step 1320). The test signal to the biological system may include a low amplitude, pulsed voltage signal between the pedestal ground of the array assembly and the system ground. As explained, the pedestal ground may be directly coupled to the biological system (as in the case where the pedestal is mounted to the skull of the patient). In certain cases where the pedestal ground is coupled directly to the biological system, a pedestal ground wire may be provided from the pedestal to inject the test signal into the biological system. According to one exemplary embodiment, signal generator 1130 may be adapted to provide a substantially low voltage pulse (e.g., -400 microvolts in single input measurement mode) in order to limit the current generated by the test.

[0189] System 1105 may receive/collect signals in response to the test signals (Step 1330). For example, electrodes 212 and reference electrode 1121 may each collect signals in response to the test signal provided by signal generator 1130 . To correct for noise, processing device 1140 may subtract noise data collected by reference electrode 1121 from the test signal data collected by each channel amplifier (Step 1340).

[0190] Processing device 1140 may determine a signal loss between the amplitude of the test signal provided by signal generator 1130 and the measured test signal data collected by each electrode 212 of sensor array 200 (Step 1350). The magnitude of the signal loss depends on the ratio of the electrode impedance and the impedance of the protective circuit at the input of the 1X amplifiers inside the patient cable. Because the values of R and C are known, the electrode impedance may be calculated using the magnitude of the signal loss (Step 1360).

[0191] Processing device 1140 may include computer software adapted to collect the impedance data on each channel and display impedance values corresponding to an electrode map of the sensor array. As a result, a technician or surgeon may analyze and diagnose particular problems associated with an electrode or sensor array. For example, if the impedance data indicates that the impedance associated with one or more electrodes is above a predetermined upper threshold level (e.g., an open circuit condition) or below a predetermined lower threshold level (e.g., a short circuit condition), a technician may determine that a potential problem exists with a connection associated with the electrode.

[0192] Fig. 15 includes an exemplary view of an impedance map provided by the computer software of processing device 1140. As illustrated in Fig. 15, the impedance map may provide a graphical representation of each electrode based on its relative position in the sensor array (e.g., top-left electrode may correspond to an electrode located at the top-left of the sensor array). After impedance testing, the computer software may display the impedance value determined by processing device 1140. The computer software may include one or more predetermined threshold levels. If the impedance deviates from these predetermined threshold levels, the software may display an identification signal, notifying the technician that the predetermined threshold level has bee.n tripped. For example, if the impedance of a particular electrode is greater than a predetermined threshold level, the impedance map may display that electrode as red, notifying the technician that a potential open circuit condition (or other potential problem) may exist. Alternatively, if the impedance of a particular electrode is less than a predetermined threshold level, the impedance map may display that electrode as yellow, notifying the technician that a short-circuit condition (or other potential problem) may exist.

[0193] In another exemplary embodiment, an integrated crosstalk diagnostic method is included. The crosstalk method may calculate the correlation of two channels in two frequency bands, for example one between 500 and 1000 Hz and one between 1750 and 2250 Hz. By way of example only, if the correlation in the first range is above 0.5 and the correlation in the second range is above 0.35, the two channels are said to crosstalk.

[0194] Crosstalk may be determined using multiple techniques. First, impedance data may be gathered using a conventional impedance measuring device (which may utilize passive test methods for measuring electrode impedance). Because these measurements exclude the protective circuit associated with the channel amplifier, the impedance may differ from the impedance measured by the automated methods described herein. However, in most cases this difference may be predictable. In situations where the impedance measurements differ by greater than a threshold amount, the electrode may be identified as containing crosstalk with one or more other channels.

[0195] Alternatively, crosstalk may also be determined by injecting multiple frequencies into the biological system and determining how well the impedance measurements at the respective frequencies compare with one another. If the correlation of the impedance measurements for each frequency band is less than a predetermined range, the signal is said to crosstalk. By using multiple frequency bands, the likelihood that similar noise existing on both bands may be reduced, thereby improving the accuracy of the crosstalk determination method.

[0196] Systems and methods consistent with the disclosed biological interface systems provide a system for collecting cellular and multicellular signals associated with a biological system of a patient, identifying neural spikes, such as those corresponding to an imagined movement initiated by a patient, and converting these signals into processed signals for transmission to one or more controlled devices 300a-d. In an alternative or additional embodiment, the biological interface system of the present invention identifies neural spikes associated with an involuntary event, such as an epileptic seizure or other medical condition, and converts the signals into a diagnostic signal and/or a control signal for transmission to a medical diagnostic and/or therapeutic device(s). Fig. 17 provides a block diagram illustration of an exemplary neural signal classification system 501 associated with one or more signal processing systems (e.g., processing unit first portion, processing unit second portion, or one or more computer systems) associated with biological interface system 100. As illustrated in Fig. 17, neural signal classification system 501 may include one or more components for identifying and classifying signals associated with biological interface system 100. For example, neural signal classifying system 501 may include, among other things, a preprocessing device 510 coupled to sensor 200, neural spike processing module 520, a local field potential processing module 530, and a data bus 502 for providing communication among one or more additional devices associated with neural signal classification system 501. Neural signal classification system 501 may also include a data processor 550 for monitoring, analyzing, and/or processing data associated with neural signals; one or more controlled devices 300a-d; a selector module 400 for selecting a particular controlled device from among the one or more controlled devices 300a-d; and storage 503 for storing raw and/or processed neural signals. Although neural signal classification system 501 is illustrated as a plurality of discrete components, it is contemplated that each element associated with neural signal classification system 501 may be implemented in software and/or digital logic, such that the functionality of neural signal classification system 501 may be realized as an integrated system included as part of one or more components associated with biological interface system 100. For example, neural signal classification system 501 may be implemented, in whole or in part, substantially within one or more components associated with implanted processing device first portion 130a and/or external processing unit second portion 130b, as shown in Figs. 1- 2.

[0197] Pre-processing device 510 may include one or more components that cooperate to receive multicellular signals from sensor 200, filter out extraneous noise signals, and separate the neural spike information (e.g., high- frequency single neural signal content including but not limited to a "hash" signal which refers to low amplitude, high rate, random neural spikes often undetectable from noise) from the local field potential information (e.g., low- frequency multi-neural signal content). As illustrated in Fig. 17, pre-processing device 510 may include a DC suppression device 511 for removing extremely low-frequency noise, an adaptive filter 512 for canceling additive noise injected by one or more components associated with biological interface system 100, and a signal separator 513 for separating the neural spike signal from the local field potential signal. Signal separator 513 may be operatively coupled to each of neural spike processing device 520 and local field potential processing device 530. Neural spike processing device 520 and local field potential processing device 530 may each be operatively coupled to data bus 502 for communication with data processing device 550.

[0198] DC suppression device 511 may be operatively coupled to the input channel and configured to suppress, filter, or otherwise limit the transmission of DC and/or extremely low-frequency signals to the remainder of neural signal classification system 501. For example, DC suppression device 511 may include a DC blocking capacitor, a high-pass filter with an extremely low cutoff frequency, or any other suitable device for limiting the transmission of DC signals. In one embodiment, DC suppression device may include a 0.3 Hz high-pass filter. Although DC signal suppression device 511 is illustrated and described as being associated with neural signal classification system 501 , it is contemplated that DC signal suppression may be performed at any point prior to analysis and/or processing of the multicellular signals. [0199] Adaptive filter 512 may include any hardware, software, and/or combination hardware/software devices operatively coupled to one or more input channels and configured to suppress extraneous signals associated with biological interface system 100. For purposes of the present disclosure, extraneous signals may include a noise signal injected by one or more electrical and/or mechanical components of biological interface system 100 (e.g., thermal and other noise associated with an electronic device, vibration noise associated with a mechanical device, power supply noise associated with a power supply, etc.), an environmental noise (e.g., spurious signal noise from the environment surrounding patient 501), and/or any other type of noise signal. According to one embodiment, adaptive filter 512 may embody a software filter that receives reference signals from one or more noise sources and generates a mathematical model indicative of the noise signal. Accordingly, mathematical models may include analysis logic that allows the mathematical function to be periodically updated to account for variations in the reference signal's interference relationship over time. For example, one or more amplifiers may include semiconductor devices that, when heated, generate a frequency and/or time varying thermal drift in electronic charge, thereby injecting a variable amount of noise signal into the system. Adaptive filter 512 may be configured to model the environmental and/or electromagnetic noise, along with other noise signals, to actively suppress the noise injected into the input channel.

[0200] Adaptive filter 512 may also be coupled to one or more electrode channels associated with sensor 200 that have been identified as inactive (i.e., not located in a region of significant neural activity and, therefore, not containing significant neural spike data). Signals collected from the one or more inactive channels may be input to adaptive filter 512 for use as reference signals to model the noise floor associated with biological interface system 200. Adaptive filter 512 may subsequently model the inactive channels to create a noise algorithm associated with sensor 200 that subtracts, cancels, and/or suppresses the noise corresponding to sensor 200 from the cellular and/or multicellular signals, thereby substantially removing all but desired neural activity. It is contemplated that adaptive filter may be initially supplied with noise algorithms associated with sensor 200 generated from lab tests of sensor 200. Alternatively and/or additionally, adaptive filter 512 may be coupled to a reference output associated with sensor 200 and may monitor sensor noise in real-time, periodically or continuously updating noise algorithms. This updating of noise algorithms may provide a filtering means that adapts with changing environmental or other additive noise conditions.

[0201 ] Adaptive filter 512 may also be coupled to average multiple inactive (or minimally active) electrode channels in order to more appropriately model system noise. For example, adaptive filter 512 may include one or more algorithms that averages and/or correlates signals from multiple electrode channels in order to model noise signals injected by sensor 200 and/or other components associated with sensor 200. According to one embodiment, adaptive filter 512 may identify and model the additive noise injected by the sensor array and/or other biological interface system components as well as additional broadband noise in the vicinity of sensor 200. In contrast, the modeling of a signal received from a single electrode may include system noise, spurious environmental noise, and local non-informative brain activity (neural noise). Adaptive filter 512 may be configured to identify, model, and/or extract each particular type of noise and limit the transmission of these signals to neural signal classification system 501.

[0202] Signal separator 513 may be operatively coupled to adaptive filter 512 and may include one or more devices configured to separate the neural spike information (generally high-frequency information, i.e. above 250 Hz), and the local field potential information (generally low-frequency information, i.e. in the range of 30-250 Hz). Signal separation devices may include frequency splitters, filters, circulators, or any other type of frequency separating device. For example, according to one embodiment, a 250 Hz high-pass filter may be placed in parallel with a 250 Hz low-pass filter to separate respective high and low-frequency components of the neural signal. Signal separator 513 may provide the neural spike information and the local field potential information to neural spike processing device 520 and local field potential processing device 530, respectively. The frequency threshold of 250 Hz is intended to be exemplary only and not intended to be limiting. It is understood that one of ordinary skill would recognize that deviations (e.g., +/- 30%) in frequency selection may result in substantially similar results. Adaptive filter 512 may be applied after signal separator 513. One or more adaptive filters may be implemented, i.e. one for the neural spike and one for the local field potential data.

[0203] Data bus 502 may include any interface that provides a communication platform between one or more components associated with neural signal classification system. For example, data bus 502 may provide a communication interface between and/or among neural spike processing device 520, local field potential processing device 530, adaptive filter 512, and/or data processing device 550.

[0204] Storage 503 may be coupled to data bus 502 and may include any device for storing data. For example, storage 503 may include a magnetic, electronic, and/or optical storage device, such as a hard drive, CD-ROM, DVD- ROM, EEPROM, or any other type of storage device. In certain embodiments, storage 502 may be included as part of processing unit first portion 130a and/or processing unit second portion 130b, in which case storage 502 may include flash media devices, ROM devices, or other small footprint and/or low-profile memory devices.

[0205] As illustrated in Fig. 17, data processing device 550 may be operatively coupled to neural spike processing device 520 and local field potential processing device 530. Data processing device 520 may include one or more components configured to execute software for performing methods consistent with the disclosed embodiments. For example, data processing device 550 may include one or more hardware and/or software components configured to collect, monitor, store, analyze, evaluate, distribute, report, process, record, and/or sort data information associated with one or more neural signals. One or more hardware components may include, for example, a central processing unit (CPU) 551 , a random access memory (RAM) module 552, a read-only memory (ROM) module 553, a storage 554, a database 555, one or more input/output (I/O) devices 556, and an interface 557. One or more software components may include, for example, a computer-readable medium including computer-executable instructions for performing a method associated with a neural signal classification system 501. It is contemplated that one or more of the hardware components listed above may be implemented in software and, similarly, one or more software programs may be implemented in hardware (e.g., digital logic, etc.). For instance, storage 554 may include a software partition associated with one or more other hardware components of processing device 550. Processing device 550 may include additional, fewer, and/or different components than those listed above. It is understood that the components listed above are exemplary only and not intended to be limiting.

[0206] CPU 551 may include one or more processors, each configured to execute instructions and process data to perform one or more functions associated with processing device 550. For instance, CPU "551 may execute software that enables processing device 550 to detect and/or receive one or more cellular signals from a sensor associated with a biological system of a patient. CPU 551 may execute software that filters each of the detected signals to produce a neural signal associated with each of the detected signals. CPU 551 may also execute software that projects one or more neural signals upon a feature space according to at least one distinct component of a generally typical wave shape associated with the neural signals. CPU 551 may also execute software that performs a statistical analysis on the feature space according to one or more distribution models, to determine clusters of neural spike activity. CPU 551 may also execute software that identifies one or more neural spikes from among the clusters of neural spike activity. According to one embodiment, CPU 551 may receive cellular data, extract the neural spike information from the cellular data, and update the feature space in real-time, such that the latest neural spike information is considered in the identification process. Accordingly, neural signal classification system 501 may be adaptable to appropriately respond to changes in environment and neural activity of a patient.

[0207] For purposes of the present disclosure, a feature space refers to any numerical or algorithmic matrix where samples of data represent projections of a set of signals into its defining elements or adequately separable components. For example, a feature space may include a graphical representation where each point in the region corresponds to a particular sample of data, i.e. a channel's brief time sample. The arrangement of points within the feature space may provide information as to the distribution of statistical information of a signal or event. For example, in pattern recognition systems, a feature space may include a multi-dimensional graphical space where individual samples of data are converted and compared based on different criteria. Analysis within the feature space may provide alternative comparison techniques to characterize information that may be otherwise overlooked. For instance, in voice recognition systems, waveform and frequency analysis between two signals may not identify minute differences between the two signals. However, these signals may be compared by piecewise correlating the signals and projecting the correlated sample onto a two-dimensional feature space (e.g., corresponding to the correlation between the signals' magnitude and phase, for example). In this manner each signal may be analyzed according to a particular feature or characteristic of the signal.

[0208] The feature space may include any multi-dimensional space for representing analysis data. According to an exemplary embodiment, this analysis typically includes some method of correlation between a neural spike and at least one benchmark signal. The feature space may include a two- dimensional space for the graphical representation of the strength of correlation between two characteristics of the signal such as, for example, a high- frequency component associated with the leading rise or fall at the beginning of a neuronal firing and a low-frequency component associated with the decaying tail of that neural spike. Ensemble projections of neural signals of similar shapes will be mapped into various regions of the feature space forming statistical clusters.

[0209] In one embodiment, the feature space may include a simple two- dimensional graph, where each point in the graph corresponds to a correlation value between a sample of a neural spike signal and the corresponding sample of the benchmark signal. It is contemplated that the feature space may include two-dimensions representative of a two-component correlation (i.e. two- characteristics) for each sample. Alternatively and/or additionally, the feature space may include additional dimensions corresponding to additional characteristics associated with the signals in the case where more distinction between classes is required that can't be had with lower dimension projections. Higher dimensional feature spaces are difficult to represent graphically but are mathematically sound.

[0210] RAM 552 and ROM 553 may each include one or more devices for storing information associated with an operation of processing device 550 and/or CPU 551. For example, ROM 553 may include any non-volatile memory device configured to access and store information associated with processing device 550 including information for identifying, initializing, and monitoring the operation of one or more components and subsystems of processing device 550 or storing and accessing information for use by CPU 551. RAM 552 may include a memory device for storing data associated with one or more operations of CPU 551. For example, ROM 553 may load instructions into RAM 552 for execution by CPU 551.

[0211 ] Storage 554 may include any type of mass storage device configured to store any type of information that CPU 551 may need to perform processes consistent with the disclosed embodiments. For example, storage 554 may include one or more magnetic and/or optical disk devices, such as hard drives, CD-ROMs, DVD-ROMs or any other type of mass media device.

[0212] Database 555 may include one or more software and/or hardware components that store, organize, sort, filter, and/or arrange data used by processing device 550 and/or CPU 551. For example, database 555 may store historical information such as previous neural spike activity and/or noise data, benchmark data such as signals used to classify particular neural events, classification data associated with neural spike and local field potential data. According to one embodiment, database 555 may store raw data at various time intervals for future comparison and/or analysis. This data may be sorted and retrieved by CPU 551 automatically or at the request of a user of the system. Database 555 may also store project parameters associated with one or more neural signals such as, for example, threshold levels and waveforms associated with respective neural spikes. It is contemplated that database 555 may store additional and/or different information than that listed above.

[0213] I/O devices 556 may include one or more components configured to communicate information with a user associated with processing device 550. For example, I/O devices 556 may include a console with an integrated keyboard and mouse to allow a user to input parameters associated with processing device 550. I/O devices 556 may also include a display including a graphical user interface (GUI) for displaying information on a display monitor. I/O devices may also include peripheral devices such as, for example, a printer for printing information associated with processing device 550, a user- accessible disk drive (e.g., a USB port, a floppy, CD-ROM, or DVD-ROM drive, etc.) to allow a user to input data stored on a portable media device, a microphone, a speaker system, or any other suitable type of interface device.

[0214] Interface 557 may include one or more components configured to transmit and receive data via any appropriate communication network, such as the Internet, a local area network, a workstation peer-to-peer network, a direct link network, a wireless network, or any other suitable communication platform. For example, interface 557 may include one or more modulators, demodulators, multiplexors, demultiplexers, network communication devices, wireless devices, antennas, modems, and any other type of device configured to enable data communication via any suitable communication network.

[0215] According to one embodiment, data processing device 550 may be configured to execute software that analyzes multicellular signals in order to provide updated noise models, mathematical algorithms, adapted benchmark signals, or any other information that may allow neural signal classification system 501 and/or its components and subsystems to adapt over time. For instance, data processing device 550 may execute software that analyzes and updates neural benchmark signals, based on real-time information received during a system training session. This adaptive capability may be performed separate or in parallel with the neural spike detection and separation capabilities. Accordingly, data processing device 550 may be communicatively coupled to one or more of the adaptive systems of neural signal classification system 501 , such as adaptive filter 512, neural spike processing device 520, and/or local field potential processing device 530, in order to provide adaptive feedback, including updated mathematical models and/or algorithms, to these components.

[0216] Neural spike processing device 520 may include one or more components configured to analyze, sort, identify, and/or classify neural spikes associated with neural activity of patient 500. For example, as shown in Fig. 18, neural spike processing device 520 may include a spike capture component 526, in parallel a low-pass filter 521 for extracting low-frequency signal content from the neural signal information, a down-sampling device 522 for smoothing the filtered signal, a signal separator 523 (illustrated in Fig. 18 as a high-pass filter 523a and a low-pass filter 523b coupled in parallel with the input channel) to further separate the high-frequency (e.g., generally between 1 kHz and 5 kHz, +/- 30%), neural signal information (e.g., the neural spike portion of the signal) from the low-frequency (generally between 250 Hz and 1000 Hz, +/- 30%) neural signal information (e.g., the neural tail portion of the signal), and high-frequency and low-frequency component detection modules 524 and 525, respectively. It is contemplated that neural spike processing device 520 may include additional, fewer, and/or different components than those listed above. It is understood that the components listed above are exemplary only and not limiting. It should be noted that the certain components may be excluded and/or modified, such as certain filters and down-sampling devices, which serve in some embodiments to reduce the processing and/or memory requirements of neural signal classification system 501.

[0217] Fig. 19 illustrates neural spike component detection modules 524- 526 according to an exemplary disclosed embodiment. Neural spike detection modules 524-526 may include a high-frequency component detector module 524, a low-frequency component detector module 525 and a spike capture module 526. High-frequency component detection module 524 may be configured to identify a neural spike and extract the significant neural spike information by means of memory and time efficient signal processing. It is advantageous to use the higher frequency portion for detection because of less biological and environmental noise contained in that frequency range. Low- frequency component detection module 525 may be configured to detect certain waveform characteristics embedded within the low-frequency portion of the neural spike, which may be used for classification of the signal once high- frequency component detection module 524 has identified the neural spike. While it is contemplated that the high-frequency and low-frequency analysis may be performed using the same signal, separating these processes provides a mechanism to manipulate the signal independently in order to extract the relevant information more efficiently. Moreover, decomposition into a feature space may be performed more efficiently by utilizing the signal already filtered into its high-frequency and low-frequency components for purposes of detection. Spike capture module 526 may be configured to capture the unprocessed neural spike signals that are detected by detection modules 524 and 525.

[0218] As illustrated in Fig. 19, high-frequency component detection module 524 may include two separate signal paths, an upper path and a lower path. The upper path may be configured to detect neural spikes in a processed version of the high frequency signal, i.e. the signal envelope, by use of an adaptive base threshold. The base threshold may adapt by averaging certain signal characteristics associated with the signal envelope (components 603- 606). The signal exiting signal separator 513 may include actual neural spikes, "spike-like" events and large amplitude, non-"spike-like" events. By performing the detection on a signal in a narrow frequency range, these non- "spike-like" events which may be of lower or higher frequency content are filtered out. By adequately averaging the remaining neural and non-neural "spike-like" characteristics (which occur at relatively low rate) with the remainder of the information contained in the envelope, a reasonable estimate of the base noise floor is obtained and a base threshold calculated from it. Trigger algorithm 607 tentatively identifies neural spikes in the envelope signal, i.e. whether it surpasses the base threshold in amplitude. This detection may happen somewhere within a typical timeframe of at most fifteen samples of data output from downsampler 522). The lower path may include one or more devices for determining a more accurate timing for a neural spike and transforming it into one or more dimensions of the feature space. In both the upper and lower paths, measures may be taken to further reduce the processing and memory requirements of the neural spike processing module 520. These measures may include additional frequency selection and downsampling, which cooperate to reduce the number of digital samples systematically, without removing significant amounts of signal content.

[0219] High-frequency component detection module 524 may include one or more components that cooperate to detect a neural spike associated with one or more cellular signals detected by sensor 200. High-frequency component detection module 524 may include, among other things, an envelope detector 600 operatively coupled in series to a low-pass filter 601 and a down-sampler 602, the output of which is fed into a trigger algorithm detection device 607. High-frequency component detection module may also include a feedforward path (lower path) that includes a boxcar smoother 603 coupled to the envelope detector 600. A down-sampler 604, an exponential smoother 605, and an amplifier 606 may be operatively coupled in series, the output of which may be fed into trigger algorithm 607 to complete the feedforward path. In a preferred embodiment, the parameter controlling the threshold level is a function of the noise floor estimate.

[0220] Envelope detector 600 may include any device and/or software component configured to determine the envelope associated with a neural signal. For example, envelope detector 600 may include any method for extracting a relative waveshape, based on the amplitude and/or spectral content associated with each signal. According to one embodiment, envelope detector 600 may include a digital detector configured to measure a relative amplitude associated with a digital sample and rectify this amplitude to generate the waveform envelope of the digitally sampled signal. Envelope detector 600 may include an algorithm that calculates the inverse z-transform of a digital signal and extracts the absolute value of this inverse z-transformed signal. According to another embodiment, envelope detector 600 may be configured to provide some smoothing between samples, such that the envelope detected represents a substantially averaged signal that limits the effects of spurious and/or extraneous samples. Envelope detector 600 may also include or be substituted with a matched filter detector which correlates the time-reversed version of the typical ensemble high-frequency neural spike waveform in order to further reduce the effects of noise on the system (e.g. in accordance with Wiener filter theory). It is contemplated that additional and/or different methods of envelope detection may be employed, and that those methods described above are exemplary only and not intended to be limiting.

[0221] Low-pass filter 601 and downsampler 602 may be provided to further isolate and smooth the detected envelope. Low-pass filter 601 may include any device configured to isolate the portion of the high-frequency component of the neural spike signal that contains the actual neural spike transition, generally between 1 kHz and 1.5 kHz (+/- 30%). However, it is contemplated that additional and/or different frequency bands may be included to provide the inclusion of signal sideband data, depending on the processing and/or memory requirements of the system. The incoming signal contains neural spikes, "spike-like" and noise artifact waveforms of significant amplitude that are not of neural nature and, hence, generally are of different frequency content. Hence, the envelope detector 600 and/or low-pass filter 601 decrease the occurrence of false-positive spike detections. Downsampler 602 may be provided to limit the number of samples of the envelope signal, for more efficient processing and faster response times during the training period. A sampling rate associated with downsampler 602 may include any desired sampling rate but, ideally, only sampling rates above the Nyquist frequency should be considered in order to prevent aliasing of the envelope signal. Although low-pass filter 601 and/or downsampler 602 are illustrated as separate components, it is contemplated that they may be integrated within a single system and/or implemented by a single function. It is also contemplated that low-pass filter 601 and/or downsampler 602 may be optional components to high-frequency component detection module 524. Accordingly, a signal output from envelope detector 600 may be fed directly into trigger algorithm 607, without any intermediate processing.

[0222] Additionally, in the lower path of high-frequency component detection module 524 following envelope detector 600, boxcar smoother 603 may be provided to average the envelope signal and provide a weighted step function. This weighted function provides a signal that allows for easy identification of the signal threshold, which is provided as an averaged function of the relative energy of the envelope. Boxcar smoother 603 may average the energy of the channels to establish a good estimate of the noise floor of the signal while averaging out neural spikes and larger, more spurious noise signals existing only in a portion of the signal. Accordingly, this ensures that trigger algorithm 607 will detect neural spikes with very small amplitudes by adjusting the estimated noise amplitude that spikes are compared to frequently.

[0223] Because boxcar smoother 603 essentially establishes a step function, downsampler 604 and exponential smoother 605 may be provided to reduce the sampling rate and further smooth the signal. Thus, sampling and smoothing of the boxcar-smoother signal may further conserve processing and/or memory resources. The resulting signal produced by the lower path (603-606) includes the base threshold information associated with the envelope signal, essentially representing the weighted magnitude of the waveform at various time intervals along the envelope signal. The noise estimation signal may then be scaled by system parameter 606 to appropriately control sensitivity of detection by trigger algorithm 607. A larger parameter means less sensitive, i.e. more true negatives; a smaller parameter means more sensitive, i.e. more false positives. It is contemplated that, as an alternative or in addition to boxcar smoother 603, other methods and devices may be employed to smooth the signal. For example, an exponential smoother or a low-pass filter may be implemented instead of boxcar smoother 603. Accordingly, the use of boxcar smoother 603 embodies one exemplary smoothing technique and is not intended to be limiting. Trigger algorithm 607 may includes any component and/or function that compares the envelope signal to one or more thresholds to determine if the envelope signal is indicative of a prospective neural spike. Threshold signals may include, for example, processed signals, such as the base threshold signal associated with the lower path of the high-frequency threshold detection module 524 following envelope detector 600. Threshold signals may also include fixed values chosen manually, reflecting empirical studies, or other noise-floor estimation procedures involving, for example, multi- electrode (i.e., cross channel) methods. The envelope/threshold mechanism of the upper path 524 of high frequency processing may include a means to address the polarity of a neural signal, its nominal shape exhibiting non- symmetry in terms of amplitude. Whereas neural signals most often are of one polarity, occasional inversion results from the physical location of the sensor tip with respect to a cell. For example, a rectifier type of envelope detection disregards the polarity, and hence trigger mechanism 607 can capture using a single threshold the typical neural signal and/or its inversion. The alignment process 608 must then account for either polarity with its own absolute comparison. An alternative is to use a non-rectified type of envelope (e.g., just level crossing or matched filtering) in which case trigger 607 must have a duo- threshold (one positive, one negative). Determining polarity via duo- thresholding in 607 means a moderately simplified alignment process 608. It is contemplated that trigger algorithm 607 may include an adaptive portion, which allows a user, technician 110, data processing device 550, and/or patient 500 to modify one or more parameters associated with the trigger algorithm 607, either manually and/or automatically, in response to one or more particular criteria. For example, technician 100 may modify trigger algorithm 607 to adjust a sensitivity (e.g. parameter 606) of biological interface system 100 associated with a particular activity of patient 500. Included in this adaptive mechanism may be techniques refining allowable spikes, such as refractory times (disallowing rapid successions of spikes) and thresholds tracking envelope levels aimed at avoiding false positive detection due to abnormal large "ringing" tails of some neural signals. The ringing effect can arise from high-pass filtering large low-frequency waveforms. The output of trigger algorithm 607 may be a timestamp A that identifies when the neural spike occurred.

[0224] According to one embodiment, high-frequency component detection module 524 may also include one or more components configured to analyze the processed neural spike signal to ensure that any processing abnormalities that may have been introduced, such as delay, dispersion, and inadvertent neural spike content removal, can be corrected prior to further characterization and identification of the neural spike data. For example, high- frequency component detection module 524 may include signal alignment device 608 and artifact detector 609. These devices may cooperate to correct any residual noise, delay, improper sampling errors, bit-errors, or any other abnormality may have been introduced into timestamp A by envelope detector 600, low-pass filter 601 and/or downsampler 602. The effect of signal misalignment in the feature space is very problematic vestigial clusters attributed to some other valid cluster, i.e., a spatial pattern of successively smaller quantity clusters streaking across the feature space.

[0225] Signal alignment device 608 may include any device or function that is configured to align the waveforms of detected neural spikes to later simplify spike classification in the feature space. Alignment is necessary because characteristic spike peaks may shift in time in the processed signal and timestamp A of a detected event might be off by up to a millisecond This shift may arise because the processed neural spike contained in the envelope signal has been significantly averaged and smoothed to remove noise and reduce the processing requirements, while still retaining substantially all of its neural signal content. Therefore, the alignment may involve certain processes that incorporate delay resolution with sample feature alignments. Signal alignment device 608 may be configured to align the processed neural spike appropriately against the benchmark signal to insure that it is properly identified and that unit classification may be accomplished. Signal alignment device 608 may include a mathematical algorithm that first correlates the two signals and finds a single time value where maximum correlation occurs. Once the time value for maximum correlation is selected, signal alignment device 608 may average points near this selected time value, in order to adjust the alignment according to the averaged time value where maximum correlation occurs. It is contemplated that additional components, processes, and subsystems may be used to perform this signal alignment, such as using a matched filter function. It is understood that the above methods and systems for signal alignment are exemplary only and not intended to be limiting. The output of signal alignment device 608 is a corrected timestamp A, now called B, and a shifted neural spike waveform.

[0226] Artifact detector 609 may include any device or function which limits the removal of certain signal characteristics of an original signal during processing. Thus, artifact detector 609 may include one or more components configured to observe the original (i.e., unprojected) neural spike signal to identify signal artifacts that may have been mistaken as a neural spike, e.g. environmental noise of large amplitude that may or may not saturate analog electronics. This may be done by checking whether an event on one channel was also detected on other channels (typically a minimum of 30+10) within a defined window of time (typically ±5 samples) or discarding events of larger amplitude than possible given the electrodes location and properties (typically ±2000 μV). The existence of such artifact-related false positives is noted, but otherwise they are discarded from further processing and/or classification.

[0227] Once the signal has been appropriately aligned and any artifacts have been corrected, the high-frequency neural spike content may be correlated with one or more benchmark high-frequency signals, using one or more correlation processes 610 for digital signal analysis. In addition, the two steps of correlation 610 and artifact detection 609a may be combined for efficiency reasons. The correlated results may be provided to data bus 502 for storage and further signal analysis, signal classification, characterization, and system calibration. These methods will be described in detail below.

[0228] Similarly, low-frequency component detection module 525 may be configured to detect any abnormalities of the low-frequency signal with respect to the original signal and/or correct any signal errors resulting from the processing (e.g., low-pass filtering) of the original neural spike. As in the high- frequency detection module 524 above, low-frequency component detection module 525 may include an artifact detector 609b to identify and/or correct any artifacts realized in the low-frequency component of the neural spike. Once the low-frequency signal has been appropriately processed, it may be correlated with a low-frequency benchmark, using one or more correlation processes 611. Again, there is no restriction on combining the operations of correlation 611 and artifact detection 609b for efficiency. The correlated results may be provided to data bus 502 for further analysis, characterization, and processing, which will be described in detail below.

[0229] Referring back to Fig. 16, neural signal classification system 501 may also include a local field potential processing device 530 to monitor, analyze, sort, store, and process local field potential signals associated with the collected multicellular signals. While not directly utilized in the neural signal characterization and identification processes, local field potential signals may be analyzed to prevent duplicate neural spike detection and/or one or more other purposes. They may also be further utilized to prevent false positives resulting from processing errors in the neural spike processing device 520. Fig. 20 illustrates an exemplary disclosed local field potential processing device 530 according to an exemplary embodiment. Local field potential processing device 530 may include a low-pass filter 532 disposed between one or more downsamplers 531 , 533 to isolate the local field potential frequency spectrum, remove excess noise, and smooth the signal. Local field potential processing device 530 may also include a band selector 534 to select particular frequency channels associated with a particular portion of the local field potential spectrum, for isolating and analyzing particular frequency bands.

[0230] Systems and methods consistent with the proposed neural signal classification system 501 provide users (e.g., patients 500, technicians 110, health care providers, etc.) with an adaptive system for receiving cellular signals associated with neurological activity, both voluntary and involuntary, identifying a neurological event or neurological information in the form of one or more time-based patterns of neural spikes embedded within the cellular signal, and characterizing the identified neural spike pattern (s) as being associated with a desired information and/or control signal of patient 500. Operation of the neural signals classification system 501 will now be described.

[0231] Fig. 21 illustrates flowchart 700 which provides an overview of the operation of neural signal classification system 501 , according to an exemplary disclosed embodiment. The process of neural signal classification may include receiving one or more cellular signals from a sensor 200 associated with biological interface system 100 (701), performing pre-processing on the cellular signals (710), performing neural spike signal processing (800) and local field potential signal processing (720), and performing post processing analysis (730). Although these methods may be illustrated and/or explained as being performed by discrete components, it is contemplated that these methods may be implemented in a software program, in one or more logic circuits, and/or a combination of hardware and/or software. It is also contemplated that the methods described may be performed by a single integrated device and/or software process or, alternatively, by discrete devices and/or software processes. Each of the method steps illustrated in Fig. 21 will be discussed in greater detail below. Neural signal classification system 501 may be associated with a single electrode or a substantially low number of electrodes associated with a multi-electrode array. According to this exemplary embodiment, neural signal classification system 501 may classify signals associated with one or more electrodes in parallel.

[0232] Fig. 22 illustrates flowchart 710 of an exemplary disclosed preprocessing method according to an exemplary disclosed embodiment. Once the multicellular signals have been received by sensor 200, the signal may be high-pass filtered (e.g., with a cutoff frequency of 0.03 Hz) to remove DC offsets and noise leakage associated with very low-frequency sources, such as power supply signals injected into the system and/or collected by sensor 200 (711). As explained, this DC signal suppression may be implemented using a variety of DC blocking devices and/or functions, such as a blocking capacitor, a high-pass filter, or any other type of signal filtering method.

[0233] Upon suppressing the DC noise on the input channels, neural signal classification system 501 may be configured to perform adaptive filtering to cancel line and/or other additive noise associated with biological interface system 100 (712). Adaptive filtering may include any device or function for suppressing a noise signal associated with one or more background signals, such as wireless Internet broadcast activity, thermal noise associated with one or more implanted electronic devices, harmonic noise associated with a fundamental frequency of a power supply, or any other type of noise signal. These adaptive filter techniques provide a mechanism for neural signal classification system 501 to adapt to the changes in noise signals inherent in neurological learning processes. Adaptive filtering may also include algorithms to periodically and/or continuously model and characterize noise signals to ensure accurate noise signal cancellation. Adaptive filtering algorithms may be determined using reference signals (i.e., signals from inactive channels or specific reference wires or electrodes) from one or more components associated with biological interface system 100. For example, amplifying devices, power supplies, null electrode channels, etc. may each be modeled offline (or during periods of system inactivity), and subtracted from signals containing neural spike information. This filtering may be performed iteratively, periodically (at certain time intervals), and/or continuously to update the noise cancellation algorithms, thereby providing neural signal classification system 501 with an adaptive noise filtering means to accurately cancel line additive noise.

[0234] According to an exemplary embodiment, adaptive noise filtering may include receiving a noise reference signal associated with one or more components (712a). For example, one or more electrodes of sensor 200 that may be associated with inactive neural regions (i.e., regions of the brain that experience substantially no neural activity, therefore measuring only noise floor) may be provided to adaptive filter 512. To maximize the noise reduction on every electrode in a grid, the inactive electrodes may be distributed throughout the array and different adaptive filters may be constructed for different electrodes. Adaptive filter 512 may create an adaptive noise model indicative of the reference signal (712b), which may include generating a noise transfer function including an algorithm indicative of the noise signal. This transfer function output may be subtracted from the multicellular signal (712c) to remove the additive noise associated with the reference signal. In addition to modeling a single reference signal, it is contemplated that adaptive noise filter 512 may average several reference signals received from the same or similar types of components, to further prevent removal of any valid neural signal content. If the interfering noise is common to all references, the sum of such noises combine for an enhanced reference signal, while individual neural spikes unique to a single channel are effectively averaged to insignificance.

[0235] Upon adaptively filtering the sampled channel, ideally leaving only neural signal information (e.g., neural spike information and local field potential information), the neural spike signal may be separated from the local field potential signal (713). As explained, this separation may be performed by any number of signal separation means, including filtering, circulating, demodulating, etc. In one exemplary embodiment, signal separation may be performed using a high-pass filter and a low-pass filter in parallel, the high-pass filter corresponding with the extraction of the neural spike content and the low- pass filter corresponding with the extraction of the local field potential information. .

[0236] Figures 23a and 23b illustrate a flowchart 800 describing a method of operation of neural spike processing device 520. According to one aspect, once the neural spike content has been separated from the local field potential content, neural signal classification system 501 may perform neural spike identification and characterization (800). The neural spike may be separated into high-frequency and low-frequency components for independent processing using any suitable signal separation means (801). The number of components may be higher with finer frequency division. According to one embodiment, the high-frequency component may be first analyzed to detect the neural spike (because of less noise in its frequency band) and, subsequently, both the high and the low-frequency component may be analyzed to classify the signal based on the distinctive high and low-frequency characteristics. It is also contemplated, however, that the high-frequency and low-frequency components may be analyzed in parallel, at substantially similar times.

[0237] Neural signal classification system 501 may extract a signal envelope associated with the high-frequency component of the neural spike (802). The signal envelope may be determined by any suitable envelope detection means, such as by applying a matched filter algorithm, performing discrete LTI transforms (such as the discrete or continuous Fourier transform). The signal envelope may then be averaged and/or downsampled (803) to further reduce processing cycles and memory required for the analysis, which serves to minimize system resources and battery consumption according to an exemplary embodiment.

[0238] If the amplitude of the signal envelope is not greater than a predetermined threshold value (804: No), a neural spike is not detected (805). Once a signal is determined not to be associated with neural activity, no further analysis is required. Alternatively, if the amplitude of the signal envelope is greater than a predetermined threshold value (804: Yes), a neural spike is eventually detected, upon passing below a threshold at a slightly later date or meeting some other similar criteria, and the signal is aligned with the baseline neural signal (806) to correct for delay and/or processing abnormalities that may have been introduced during the envelope and neural spike detection processes. As explained, this alignment may include correlating the processed signal and selecting the maximum point of correlation as a reference. One or more points near the maximum may be averaged and correlated with the original to align the signal in accordance with the maximum averaged correlation between the signals.

[0239] Once the processed signal has been appropriately aligned, the high and low-frequency component may each be correlated with a respective benchmark signal (807, 809). This benchmark signal may be a predetermined waveform corresponding to one or more ideal signals indicative of an established neural spike associated with patient 500. The benchmark signal may alternatively be based on historic data from patients other than patient 500. In the case of non-human animal models a similar approach applies but the signal may be more suited to the nominal signal for the genus or species. Each of high-frequency 809 and low-frequency 807 projection components, alpha 1 and alpha 2, associated with each of high-frequency and low-frequency signals may be arranged as a vector to form a feature space 810. The feature space may include any suitable multi-dimensional projection, e.g., third order or higher principle components techniques. The advantage of a fixed, predetermined benchmark is the avoidance of complex decomposition and projection algorithms that occasionally could be susceptible to noise artifact abnormalities and the fact that the feature space remains fixed and familiar.

[0240] After the system has been established through appropriate training, subsequent signals are projected onto the feature space to determine if the signal aligns with an established cluster. Should the signal not align with a cluster (i.e., correlation is less than a minimum threshold or the cluster is not within classification boundaries, i.e., go outside the hyperplane or conic section noise boundary) (812: No), the signal is flagged as a false positive and discarded (813). If the feature space point passes the threshold or is within the classification boundaries of a cluster and falls outside the noise region, it is classified as belonging to its nearest cluster and then passed onto the decoding/filtering mechanism as a statistical neural event contributing to a corresponding voluntary patient event.

[0241] Because neural signal classification system 501 may be configured to store and adapt as additional neural spikes are identified and projected onto the feature space, one or more benchmark signals and/or threshold levels may be modified based on signal migration within the feature space (817). For example, in some cases, neural signals may slightly change based on cell death, sensor movement or other physical aspect associated with patient 500. These changes may call for temporary or continual adaptation of neural signal classification system 501 to correspond to these changes. Accordingly, it is contemplated that historical averages may be utilized to ensure that residual data is preserved in case that these changes are temporary and not characteristic of a new type of neural spike. Further, postprocessing analysis (730) may be performed to adjust one or more system parameters based on the successful identification of a neural spike.

[0242] Neural signal classification system 501 may allow the user to manually combine two or more clusters of neural spikes that were separately identified by an automated method. For example, a lookup table may be defined that maps a cluster to another. This mapping may be applied after automated spike sorting is complete. Manual override may have an effect on automated spike sorting parameters, e.g. in the case when two clusters are combined. One possible cluster may be defined as noise and not a neural spike.

[0243] Neural signal classification system 501 may allow the user to partially or completely override the automated method on one or more channels. For example, the user may be able to draw circles, ellipses or freeform shapes in the feature space to define the edges of clusters. Manually defined clusters may serve as suggestions to the automated sorter and effectively adjust automated spike sorting parameters.

[0244] Neural signal classification system 501 may be configured to combine (814) the identified neural spikes with other signals, such as with neural spikes identified by multiple other neural signal classifiers, similar to or different from the neural signal classifier of the present invention. Neural signal classification system 501 may be further configured to decode (815) the combined signals to produce one or more processed signals and transmit (816) the one or more processed signals to one or more devices, such as a diagnostic, therapeutic and/or patient thought-controlled device.

[0245] Systems and methods consistent with the disclosed embodiments provide a process for analyzing the local field potential signals. Fig. 24 illustrates a flowchart 720 describing an operation of an exemplary disclosed local field potential processing device 530. Local field potential processing device 530 may low-pass filter and downsample the electrode neural signal (721), in order to smooth the signal and limit the processing and memory requirements of local field potential processing device 530. The signal may be subsequently filtered to resolve the downsampled signal to a particular frequency range of interest for local field potential processing (typically, < 250 Hz) (722). Once the signal has been filtered, it may be band-selected (723) for further low-frequency analysis and characterization. It is contemplated that local field potential data may be observed with respect to its high-frequency counterpart to extract any relevant indications of a future event such as intended patient motion embedded within the local field potential signal.

[0246] Post-processing analysis (730) may include processes for modifying one or more component parameters based on neural signal classification system training. This processing may include manual and/or automated adaptive techniques. Manual techniques may include any techniques where a user modifies a parameter, such as a control signal sensitivity, a benchmark or threshold signal level, a noise reference signal, or any other parameter associated with biological interface system 100, neural signal classification system 501 , one or more therapeutic, diagnostic and or patient thought-controlled devices 300a-d, and/or any subsystem or component associated with these systems. Automated adaptive techniques may include adjusting one or more mathematical and/or functional algorithms based on modeling and/or statistical analysis of neural signal classification system 501. For example, during a learning period associated with biological interface system, neural signal classification system 501 may operate iteratively, making adjustments to mathematical algorithms (such as noise and trigger algorithms) until the neural signal clusters begin to converge in the feature space. [0247] According to one embodiment, upon projecting signals into the appropriate feature space, the data may be grouped in clusters, each cluster corresponding to a particular type of neural signal. The data may be grouped using a two-dimensional (or higher dimensional) Gaussian distribution function in order to map distinct clusters of signals in the feature space (811). Given a Gaussian mixture model as shown in Figure 25, clusters in a feature space 1401 may have a different mean and/or a different covariance determined by some underlying multidimensional distribution 1402. Various methods may be employed within the feature space to define and/or associate each signal with an appropriate type of signal (e.g., to determine the number of clusters over time). For example, during training, signals may be projected in an empty feature space. As signal points begin to populate the feature space, distinct clusters may begin to form. These clusters may be analyzed to determine if these signals are, in fact, the same type of signal or part of a distinct cluster associated with multiple types of neural spike. Based on the analysis of individual clusters or groups of two clusters, they may either be divided or combined. Additionally, first order hyperplanes or second order conic sections may be established by a user of the system, in order to distinguish between noise and all of the other clusters and/or to manually separate clusters. As additional neural spike samples are projected onto the feature space, neural spike classification system 501 may begin adapting the cluster model either continuously, at certain time/sample intervals or until there is a stable model. From there, through high correlative yields, each type of neural signal may be associated with activity of patient 500 via a decoding/filtering mechanism.

[0248] Several different methods of spike classification may be employed in neural spike classification system 501. According to one embodiment, the algorithm may initially assume that there is one cluster as illustrated in Figure 26a. What appear as three Gaussian clusters 1501 , 1502 and 1503 may be treated first as one type of unit. Lines on Figure 15a indicate the classifier's model of the principle major 1504 and minor 1505 axes of a single Gaussian distribution (not shown). After a certain period of time or number of new samples, the single cluster may split (shown as two classes 1506 and 1507 in Figure 15a) and those subclusters may later split (shown as four classes 1508, 1509, 1510 and 1511 in Figure 26b) and so on. Because the statistical classification may model not precisely match the underlying physical statistics, a random aberrant splitting may occur resulting in too many clusters, as illustrated in Figure 26b. But a combining mechanism, which acts analogous to the division mechanism, may eventually correct that occurrence. The intent is to randomly approach the best approximation to the number of clusters from either condition of not enough or too many components of the statistical model. Figure 26b illustrates that the algorithm eventually converges to the proper number of clusters or spike unit clusters 1512, 1513 and 1514.

[0249] In one embodiment, neural spike classification system 501 may use a second order model for the classification, based upon two dimensional Gaussian distributions (method 1). The classifier may be implemented using log likelihood to avoid computations for such things like exponentials and logarithms which consume much memory. For example, second order means that boundaries between clusters may be conic sections. The model may be complete given the correlation matrix and mean vector which are easily estimated given a set of points. The covariance is calculated by dividing by the number of points rather than dividing by number of points minus one to achieve further computational savings. First order models such as Euclidean distance are inferior and not preferred. The statistics may be kept constant for an interval of time while classification is performed on a block of samples (e.g. 200±100). As new data samples are collected, the model is updated, the algorithm is inherently adaptive. To get multiple clusters to move around in a logical fashion it seems important to update the vector mean on a sample-by- sample basis for which a gradient descent approach may be used. This allows this algorithm to use little resources and the ability to be updated after each sample while not being optimal.

[0250] Because of the non-globally optimal approach, these cluster may adapt in such a way as to compete with one another, i.e. detected clusters move together and clusters with low spike rates may not be detected. To alleviate this issue, the amount of adaptation for the mean is scaled back by a factor of ^ΛA if a defined cluster adapts its mean towards the same point as its closest neighbor.

[0251] The classification method may be biased to create more clusters than are actually present because the later decoding can weight two classes equally; effectively claiming they are the same class. [0252] Clusters may be split by generating a separating hyperplane (line) from the eigendecomposition of the covariance matrix and dividing a cluster in half if the ratio of sub-cluster variance to distance between centers decreases significantly by the split. Likewise, clusters may be combined by combining a cluster with its nearest neighbor (e.g. based on Mahalanobis distance) and again checking if the sub-cluster variance falls below a certain level by the combination. The number of clusters may not change after a few minutes and, in turn, the number of units and/or cluster statistice, i.e. ellipse center and orientation, may be frozen.

[0253] In another embodiment, neural spike classification system 501 may collect a block of samples (e.g. 200±100), construct a histogram (method 2), estimate peaks and then make a decision if 1) a cluster should be split into two separate clusters because there is more than one peak and 2) a cluster and its nearest neighboring cluster should be combined into a single cluster because there is only a single peak.

[0254] Several different techniques for constructing a histogram based on feature space samples may be employed. For example, the histogram may be constructed by projecting the sample points onto an axis. The axis may be the major principle component, the Fisher linear discriminant or other statistical measures of spread or variation. The Fisher linear discriminant provides one exemplary estimate than, in certain circumstances may perform better the principle component analysis, (i.e. is more sensitive). Figure 27 shows the axis 1601 associated with the principle component of variation along which points are perpendicularly projected to construct the histogram 1602 shown in Figure 16. The same group of data is shown in Figure 27 but the direction of projection is the Fisher linear discriminant 1701. Figure 27 also illustrates the improved separation 1703 in the associated histogram 1702 compared to the separation 1603 of the principle component projection histogram 1602. Because the Fisher linear discriminant requires statistics for individual clusters but the principle component analysis does not, principle component analysis may be used when the number of clusters is unknown, i.e. when the classification model is initialized, whereas the linear discriminant may be used when the clusters are estimated. Because of the differences in sensitivity, it would be less likely initially that a group of two clusters modeled as one will be split than it would be that a group of two clusters modeled as two will be combined. For that reason, we may consider making the splitting algorithm slightly more sensitive than the combining algorithm. If a false split is followed by a correct combine, we can lessen the sensitivity of split for that cluster such that successive split tests will not be so apt to produce multiple clusters.

[0255] The advantage of constructing a histogram and determining the number of units from it is that it contains more information than a single statistic and allows decision making about cluster splitting or combining based on several different criteria rather than one. Independent of the axis chosen to project the samples onto, the histogram may be constructed either with a fixed, user-defined or performance-dependent number of samples. Feature space samples may be translated via an affine function and then cast to an integer for a fixed number of bins. The number of bins may be chosen to strike a balance between minimizing resources and computations necessary and maximizing cluster separability (e.g. 20+10 bins similar to the examples of Figure 27 and Figure 28). The number of bins may be adjusted depending on the spatial spread of clusters in the feature space. For example, the range of values included in the histogram may be limited to scaling the projected data to capture the mean +- six standard deviations but this may discard outlying, low spike rate clusters. If such small, outlying clusters are present, the range may be extended to the limits of the data to include these clusters by either keeping the number of bins or the bin width constant, i.e. increasing the number of bins. Keeping the number of bins constant may be undesirable because the resolution of individual clusters is reduced and information might be lost. Another method to avoid this may be to discard some of the samples at the edges (e.g. two or three) 1801 which would remove single outliers that are not part of clusters and limit the range to some degree, as was done in the low rate cluster example of Figure 29. This may be implemented using a three element bubble sort method.

[0256] Upon choosing a direction, i.e. projection axis, that achieves good or optimum separation amongst clusters, the hypothesis test of whether there is one cluster or are multiple clusters needs to be robust to a wide range of possible distribution characteristics. Although individual clusters do generally fall in patterns that are convex in nature (see Figure 25), the arrangements of these clusters due to multiple cells firing in the vicinity of the electrode can be rather complex and confuse statistical measures. By observing enough spike samples we may see histogram patterns that will reveal all clusters present. The histogram may be implemented to not be computationally expensive. However, it may require retention of a block of data for multiple computations. Alternate buffers may be used to perform computations on collected data while accumulating sample space data. Data buffers may be updated in a circular fashion where incoming samples eliminate the oldest samples. Furthermore, retaining the data means it may be reclassified at a later time. The histogram leaves open a variety of methods for which to decide what is and what isn't a peak or valley. Here we present one peak-valley-finding technique, but this is not meant to limit the scope of alternatives.

[0257] The easiest of arrangements is when there are two distinct clusters of equal probability. One of the more difficult situations is when two units are of nearly equal probability but are near one another in the feature space such that the clusters overlap. Another difficult situation is when two units may be relatively far apart in the feature space but one of the clusters may have a much smaller spike rate. To address the general problem and the situation where two units overlap, consider beginning from the left most point in the histogram and progressing to the right. Figure 30 shows an example histogram where the "peak trace" line tracks the maximum of all previous bin counts (i.e., those to the left in the figure) beginning from the start of the peak region. Notice how the trace remains elevated 1901 near bin location 5 in Figure 30 even though the histogram begins descending near 1901. The strategy is to declare entering a valley upon first descending past, say, 50% of the peak trace, which is labeled "valley threshold" in Figure 30. This transition 1902 happens near bin location 7 in the figure. The value of the peak 1903 occurs at the highest value the histogram achieves before reaching a new valley.

[0258] Similar to the peak, the "valley trace" tracks the minimum of all . previous bin counts beginning from the start of the valley region. Because a valley can extend to zero, a percentage of the valley as a threshold strategy needs modification. Instead, we use a "peak threshold" (Figure 30) which is, say, above the valley trace by 50% of the average of the three previous bins, or five, whichever value is larger. The value of the valley 1904 occurs at the lowest point the histogram achieves after entering the valley and before ascending past the peak threshold. [0259] The concept of percentages that the histogram must fall or rise is important because a histogram, i.e., density estimation, is inherently noisy, and it is difficult to generally choose the resolution of the histogram to smooth out noise. To illustrate how averaging previous bin counts adjusts sensitivity of the peak threshold, Figure 31 shows the histogram for a low-spike-rate cluster isolated from a high-spike-rate cluster. After traversing the first large peak 2001 and entering the valley region, the histogram settles near zero. Accordingly, the peak threshold tends toward the minimum allowed value, say, five 2002. This makes the algorithm more sensitive to catching the next small peak in the histogram 2003.

[0260] It should be noted that, the location of the valley used in the binary division of the cluster, nearest the largest peak, may be important for constructing a boundary if there is more than one peak in the histogram. In such a scenario, it may be the case that the histogram valley region achieves a minimum at more than one location. The valley is at the average bin location 1802 in this situation, as illustrated in Figure 29 where there is a region of nine bins having zero count.

[0261] Any spike classification method requires a certain number of spike samples to be able to reliably determine the number of distinct neural spikes, i.e. clusters in the feature space. Collecting feature space samples may be done over a fixed period of time or until a fixed number of samples have been with the latter being the preferred method. Collected samples may be stored in a circular buffer of fixed or adapting size. This may allow spike classification to be performed after each incoming sample or at any time afterwards always using the most recent samples and without having to wait for enough samples to be collected. The latter is particularly important when there are neural spikes that occur infrequently or with varying frequency. Spike classification may also be triggered based on statistics calculated on the collected samples. The contents of the buffer may be requested by another spike classification module or computer connected to perform spike classification and return its results when spike classification module 501 does not have sufficient resources to perform its operation.

[0262] An advantage of the histogram approach may be that a valley offers an excellent location at which to place a boundary between groups of data. Retaining data enables the reclassification of points based upon which side of the valley they are located, effectively creating a hyperplane boundary in the higher dimensional space. This allows the classification technique to be adaptive when spike signals change, e.g. by micromovement of the implant. For example, as more samples are added or replace older ones and are analyzed, the number of clusters may remain the same but the histograms constructed for the combine test allows refining the classification boundaries.

[0263] If inverted spikes, i.e. spikes that have a waveform that goes up first and then down instead of down and up, are included in the classification, samples will fall into two quadrants in the feature space, the upper right and bottom left quadrants. The spike classification system may decide to treat both quadrants separately, i.e. not allowing samples in each quadrant to be grouped into one cluster.

[0264] The spike classification system may decide to automatically split samples into two or more clusters if the range of projected values gets larger than a certain threshold implying that the spread of data points is larger than realistically possible for a single unit/spike. This may remove the need for adjustments in the range or width of histogram bins.

[0265] Another way to address outlying and/or infrequently firing spikes versus higher firing spikes, i.e. low versus high peaks in the histogram, which might not be caught by the histogram threshold method described above may be to automatically split two clusters when there is a gap between the peaks where there is a minimum number of bins with a count of zero (e.g. 5±3) and the second, smaller peak contains a minimum number of samples (e.g. 20+10).

[0266] The histogram classification technique sets itself apart from other classification algorithms by assuming a Gaussian mixture model but not relying on it when splitting or combining clusters. The fact that the user may be allowed to define noise boundaries that may distort or partially cut into the Gaussian distribution of a neighboring cluster may impair an algorithm's ability to recognize the samples as a separate cluster. The histogram may only require two peaks of any shape and a defined valley between the two to be able to separate the two into clusters. This may also allow the algorithm to operate on a smaller sample set than other algorithms that rely on statistical distributions or measures.The number of possible clusters may be limited by the system, e.g. to 5 or 10 different clusters, i.e. units. The classification algorithm may not be limited and the limit may be enforced before data is presented to the user by, for example, either disregarding all units after the first 5 or 10 as noise, or by combining clusters in some way, e.g. renaming all units after the first 5 as unit 5.

[0267] Systems and methods consistent with the disclosed embodiments provide a neural signal classification system 501 for use with the biological interface system 100 that may be configured to detect, identify, classify, sort, and analyze one or more neural spikes associated with cellular signals received from sensor 200. Neural signal classification system 501 may also include an adaptable training system that rapidly, efficiently, and accurately populates a feature space with neural spike information and efficiently groups clusters of signals to define preliminary neural spike identification. As the feature space becomes more populated, the clusters begin defining more distinct patterns, which eventually converge. These clusters may be averaged to generate and/or update a benchmark type of neural spike associated with a particular neural activity.

[0268] As explained, one or more components associated with neural signal classification system 501 may include processes, algorithms, and/or functions that include manually or automatically adjustable parameters. For example, technician 110 may determine during a training session that neural spike activity is not being properly identified. Technician 110 may manually adjust the trigger algorithm threshold to increase the sensitivity of neural signal classification system 501. Alternatively, a patient 500 may determine that the signal is generating false positives (i.e., that neural spikes are being detected when no voluntary neural activity is being coordinated). Accordingly, technician 110 may increase the trigger algorithm threshold to reduce the sensitivity of neural signal classification system 501 in order to decrease the likelihood of a spurious signal triggering a neural spike.

[0269] Other embodiments of the invention will be apparent to those skilled in the art from consideration of the specification and practice of the invention disclosed herein. It is intended that the specification and examples be considered as exemplary only, with a true scope and spirit of the invention being indicated by the following claims. In addition, where this application has listed the steps of a method or procedure in a specific order, it may be possible, or even expedient in certain circumstances, to change the order in which some steps are performed, and it is intended that the particular steps of the method or procedure claim set forth here below not be construed as being order- specific unless such order specificity is expressly stated in the claim.

Claims

WHAT IS CLAIMED IS:

1. A method for classifying neural signals for a biological interface system, comprising: receiving a plurality of multicellular signals from a sensor associated with a biological system of a patient; filtering the plurality of multicellular signals to produce a neural signal, the neural signal including a neural spike portion; extracting the neural spike portion of the neural signal; correlating the neural spike with a signal indicative of an ideal neural spike associated with the patient; projecting samples indicative of the correlation on a feature space; adaptively determining a spike sorting statistical model for the feature space samples; and classifying the neural spike based on one or more clusters of data samples observed in the feature space.

2. The method of claim 1 , wherein filtering includes: receiving a reference signal associated with at least one component of the biological interface system; and determining an algorithm indicative of noise based on the reference signal.

3. The method of claim 2, wherein filtering further includes substantially suppressing the noise based on the determination.

4. The method of claim 2, wherein at least one component includes an electrode associated with the sensor, the electrode being located in a region of substantially no voluntary neural activity.

5. The method of claim 2, wherein the reference signal includes a signal indicative of a noise level associated with the neural activity of the brain, the noise level including substantially no voluntary neural activity.

6. The method of claim 1 , wherein extracting the neural spike portion of the neural signal includes separating the neural signal into a high- frequency portion and a low-frequency portion.

7. The method of claim 6, wherein the high-frequency portion includes the neural spike portion associated with the neural signal.

8. The method of claim 6, wherein the low-frequency portion includes a local field potential portion associated with the neural signal.

9. The method of claim 6, wherein extracting the neural spike portion includes filtering the neural signal.

10. The method of claim 9, wherein filtering the neural signal includes high-pass filtering the neural signal.

11. The method of claim 10, wherein high-pass filtering includes passing the neural signal through a high-pass filter with a cutoff frequency of about 250 Hz.

12. The method of claim 9, wherein filtering the neural signal includes low-pass filtering the neural signal.

13. The method of claim 12, wherein low-pass filtering includes passing the neural signal through a law-pass filter with a cutoff frequency of about 250 Hz.

14. The method of claim 1 , wherein determining whether the neural spike portion includes a neural spike comprises: detecting an envelope associated with the neural spike portion; comparing the detected envelope with a threshold value; and determining, if the detected envelope exceeds the threshold value, that the neural spike portion contains a neural spike.

15. The method of claim 14, wherein determining whether the neural spike portion includes a neural spike further comprises: passing the neural spike portion through a boxcar smoothing function to produce a signal indicative of a weighted amplitude associated with the neural spike portion over a predetermined time period; comparing the signal indicative of the weighted amplitude with a threshold value; and determining, if the signal indicative of the weighted amplitude exceeds the threshold value, that the neural spike portion contains a neural spike.

16. The method of claim 1 , wherein the feature space includes a two- dimensional feature space.

17. The method of claim 16, wherein a definition of a neural spike is configured to adapt as additional samples are projected on the feature space,

18. The method of claim 1 , wherein classifying the neural spike includes: identifying a cluster of correlated data in the feature space; and associating the cluster with a voluntary stimulus of the patient.

19. The method of claim 1 , wherein the voluntary stimulus includes a signal corresponding to a desired limb movement.

20. The method of claim 19, wherein the desired limb movement includes a movement of one or more fingers.

21. The method of claim 19, wherein the desired limb movement includes a hand movement.

22. The method of claim 19, wherein the desired limb movement includes an arm movement.

23. A computer readable medium for use on a computer system, the computer readable medium having computer executable instructions for performing the method according to claim 1.

24. A method for classifying neural signals for a biological interface system, comprising: receiving a plurality of multicellular signals from a sensor associated with a biological system of a patient; filtering each of the plurality of multicellular signals to produce one or more neural signals; projecting the one or more neural signals on a feature space according to a characteristic associated with the one or more neural signals; and identifying one or more neural spikes associated with the one or more neural signals.

25. The method of claim 24, wherein the characteristic includes one of a size and a wave shape associated with the one or more neural signals.

26. The method of claim 25, wherein the one or more neural spikes is indicative of a voluntary stimulus initiated by the patient.

27. The method of claim 26, wherein the voluntary stimulus includes a signal corresponding to a desired limb movement.

28. The method of claim 27, wherein the desired limb movement includes a movement of one or more fingers.

29. The method of claim 27, wherein the desired limb movement includes a hand movement.

30. The method of claim 27, wherein the desired limb movement includes an arm movement.

31. The method of claim 24, wherein identifying the one or more neural spikes includes classifying each of the one or more neural spikes according to a benchmark reference signal.

32. The method of claim 24, wherein the filtering includes providing a filter configured to cancel noise associated with the sensor.

33. The method of claim 32, wherein the filter is configured to: receive a reference signal associated with a first electrode of the sensor that is not substantially exposed to neural activity; receive a neural signal associated with a second electrode of the sensor that is substantially exposed to neural activity; and determine an algorithm based on a correlation between the reference signal and the neural signal.

34. The method of claim 24, wherein the feature space includes a second order model based on two-dimensional Gaussian distributions.

35. The method of claim 24, wherein the one or more neural signals are grouped in the feature space to form clusters of neural activity, each cluster corresponding to a particular type of neural spike.

36. The method of claim 35, wherein a definition of each particular type of neural spike is configured to adapt as additional neural signals are projected on the feature space.

37. The method of claim 35, wherein identifying one or more neural spikes includes: identifying a cluster of one or more neural signals; and determining an algorithm associated with the correlation of one or more neural signals associated with the cluster, the algorithm indicative of an adapted neural spike benchmark.

38. The method of claim 24, wherein projecting one or more neural signals in a feature space includes: detecting an envelope associated with each of the one or more neural signal to produce an envelope signal corresponding to each of the one or more neural signals; and aligning the envelope signal with the one or more neural signals to remove delay associated with the envelope signal.

39. The method of claim 38, wherein detecting the envelope associated with each of the one or more neural signals includes smoothing the envelope signal.

40. The method of claim 24, wherein filtering each of the plurality of multicellular signals to produce one or more neural signals includes separating each of the one or more neural signals into a high-frequency portion and a low- frequency portion.

41. The method of claim 40, wherein the high-frequency portion includes a neural spike associated with each of the one or more neural signals.

42. The method of claim 41 , wherein the low-frequency portion includes a local field potential associated with each of the one or more neural signals.

43. The method of claim 40, wherein separating each of the one or more neural signals includes filtering each of the one or more neural signals.

44. The method of claim 43, wherein filtering each of the one or more neural signals includes high-pass filtering each of the one or more neural signals.

45. The method of claim 44, wherein high-pass filtering includes passing each of the one or more neural signals through a high-pass filter with a cutoff frequency of about 250 Hz.

46. The method of claim 43, wherein filtering each of the one or more signals includes low-pass filtering each of the one or more neural signals.

47. The method of claim 46, wherein low-pass filtering includes passing each of the one or more neural signals through a law-pass filter with a cutoff frequency of about 250 Hz.

48. The method of claim 24, wherein projecting the one or more neural signals on a feature space includes projecting a high-frequency portion of each of the one or more neural signals on a high-frequency portion of the feature space.

49. The method of claim 48, wherein projecting the one or more neural signals on a feature space includes classifying the high-frequency portion of each of the one or more neural signals based on a benchmark signal associated with the high-frequency portion of the feature space.

50. The method of claim 24, wherein projecting the one or more neural signals on a feature space includes projecting a low-frequency portion of each of the one or more neural signals on a low-frequency portion of the feature space.

51. The method of claim 50, wherein projecting the one or more neural signals on a feature space includes classifying the low-frequency portion of each of the one or more neural signals based on a benchmark signal associated with the low-frequency portion of the feature space.

52. A computer readable medium for use on a computer system, the computer readable medium having computer executable instructions containing instructions for performing the method according to claim 24.

53. A neural signal classification system for a biological interface system, comprising: an input channel for receiving at least one signal from a sensor associated with a biological system of a patient; a filter operatively coupled to the input channel and configured to substantially suppress noise associated with the signal; a signal separator operatively coupled to the filter for separating the at least one signal according to at least one predetermined frequency threshold; and a neural signal processor operatively coupled to the signal separator for identifying at least a portion of the at least one signal as a neural spike.

54. The system of claim 53, including a DC blocking device operatively coupled to the input channel for substantially suppressing one or more low-frequency signals associated with the input channel.

55. The system of claim 53, wherein the biological system includes a neurological system.

56. The system of claim 55, wherein the neurological system includes a motor cortex of a brain associated with the patient.

57. The system of claim 53, wherein the input channel includes a digital channel.

58. The system of claim 57, wherein the digital channel is sampled at a rate of 30 kilosamples/second.

59. The system of claim 54, wherein the DC blocking device includes a high-pass filter.

60. The system of claim 54, wherein the DC blocking device includes a DC blocking capacitor.

61. The system of claim 53, wherein the filter includes a mathematical filter configured to cancel one or more of a noise signal injected by the sensor and a component associated with the sensor.

62. The system of claim 53, wherein the filter is configured to: receive a reference signal associated with a portion of the sensor that is exposed to substantially no neural activity; receive an electrode signal associated with a portion of the sensor that is exposed to neural activity; and determine a noise cancellation algorithm based on a correlation between the reference signal and the electrode signal.

63. The system of claim 62, the noise cancellation algorithm includes a least-mean-square algorithm.

64. The system of claim 53, including a storage device configured to store data associated with the one or more input signals.

65. The system of claim 53, including a storage device configured to store historical data associated with neural spike activity.

66. The system of claim 53, including an operator interface system for adjusting one of the predetermined frequency threshold, an input channel sampling rate, a line noise cancellation variable, an alignment parameter, and trigger algorithm.

67. The system of claim 53, wherein the signal separator includes one or more filters.

68. The system of claim 67, wherein at least one of the one or more filters includes a high-pass filter.

69. The system of claim 68, wherein the high-pass filter includes a high-pass filter with a 250 Hz cutoff frequency.

70. The system of claim 68, wherein at least one of the one or more filters includes a low-pass filter.

71. The system of claim 70, wherein the low-pass filter includes a low-pass filter with a 250 Hz cutoff frequency.

72. The system of claim 53, wherein the signal separator includes a frequency splitting device.

73. The system of claim 53, wherein the neural signal processor is configured to: determine whether the portion of the at least one signal includes a neural spike, the neural spike indicative of a voluntary stimulus associated with the patient; project information associated with a neural spike onto a feature space, the feature space indicative of a correlation of the neural spike with a benchmark signal; and identify one or more types of voluntary stimuli based on analysis of the feature space, wherein the projected information is grouped in clusters, each cluster defining a particular type of voluntary stimuli.

74. The system of claim 53, wherein the neural signal processor includes an envelope detector for detecting an envelope associated with the neural spike.

75. The system of claim 74, wherein the neural signal processor further includes a boxcar smother operatively coupled to the envelope detector and configured to provide a signal indicative of a weighted average of an amplitude associated with the neural spike.

76. The system of claim 75, wherein the neural signal processor further includes a downsampling device operatively coupled to the boxcar smoother for reducing a number of digital samples associated with the signal indicative of the weighted average of the amplitude associated with the neural spike.

77. The system of claim 76, wherein the neural signal processor further includes: a low-pass filter operatively coupled to the envelope detector for isolating the neural signal information contained in the envelope; and a downsampling device operatively coupled to the low-pass filter for reducing the number of digital samples associated with the filtered envelope.

78. The system of claim 77, wherein the neural signal processor further includes a trigger algorithm configured to: compare at least one of the signal indicative of the weighted average of the amplitude associated with the neural spike and the filtered envelope with a threshold value; and determine, based on the comparison, whether the neural spike contains information indicative of a voluntary stimulus initiated by the patient.

79. A neural signal classification system for identification and classification of neural spike activity, comprising: a preprocessing device operatively coupled to an input channel and configured to: i receive multicellular signals collected from a sensor, at least a portion of the sensor configured to be disposed within the brain of a patient; and filter the multicellular signals to extract a neural signal portion of the multicellular signals, the neural signal portion including a neural spike portion and a local field potential portion; and a neural spike processing device operatively coupled to the preprocessing device and configured to: determine whether the neural spike portion includes a neural spike, the neural spike indicative of a voluntary stimulus associated with the patient; project information associated with a neural spike onto a feature space, the feature space indicative of a correlation of the neural spike with a benchmark signal; and identify one or more types of voluntary stimuli based on analysis of the feature space, wherein the projected information is grouped in clusters, each cluster defining a particular type of voluntary stimuli.

80. A method for classifying neural signals for a biological interface system, comprising: receiving a plurality of multicellular signals from a sensor associated with a biological system of a patient; extracting a neural spike portion of the multicellular signal, wherein the neural spike portion corresponds substantially with a high-frequency portion of the multicellular signal; determining whether the neural spike portion includes a neural spike indicative of a voluntary stimulus initiated by the patient; correlating the neural spike with a signal indicative of an ideal neural spike associated with the patient; and identifying the neural spike based on the correlation.

81. The method of claim 80, further comprising filtering the plurality of multicellular signals to produce a neural signal that includes a neural spike portion.

82. The method of claim 81 , wherein filtering includes: receiving a reference signal associated with at least one component of the biological interface system; and determining an algorithm indicative of noise based on the reference signal.

83. The method of claim 82, wherein filtering further includes substantially suppressing the noise based on the determination.

84. The method of claim 82, wherein the at least one component includes an electrode associated with the sensor, the electrode being located in a region of substantially no voluntary neural activity.

85. The method of claim 82, wherein the reference signal includes a signal indicative of a noise level associated with the neural activity of the brain, the noise level including substantially no voluntary neural activity.

86. The method of claim 80, wherein extracting the neural spike portion of the multicellular signal includes separating the neural signal into a high-frequency portion and a low-frequency portion.

87. The method of claim 80, wherein determining whether the neural spike portion includes a neural spike comprises: detecting an envelope associated with the neural spike portion; comparing the detected envelope with a threshold value; and determining, if the detected envelope exceeds the threshold value, that the neural spike portion contains a neural spike.

88. The method of claim 87, wherein determining whether the neural spike portion includes a neural spike further comprises: passing the neural spike portion through a boxcar smoothing function to produce a signal indicative of a weighted amplitude associated with the neural spike portion over a predetermined time period; comparing the signal indicative of the weighted amplitude with a threshold value; and determining, if the signal indicative of the weighted amplitude exceeds the threshold value, that the neural spike portion contains a neural spike.

89. A biological interface system for collecting multicellular signals emanating from one or more living cells of a patient and for transmitting processed signals to a controlled device, comprising: a sensor for detecting the multicellular signals, the sensor consisting of a plurality of electrodes to allow for detection of the multicellular signals; and a processing unit configured to: receive the multicellular signals from the sensor; extract a neural spike portion of the multicellular signal, wherein the neural spike portion corresponds substantially with a high-frequency portion of the multicellular signal; determine whether the neural spike portion includes a neural spike indicative of a voluntary stimulus initiated by the patient; correlate the neural spike with a signal indicative of an ideal neural spike associated with the patient; identify the neural spike based on the correlation; and transmit a control signal indicative of the neural spike to a controlled device.

90. The system of claim 89, further comprising a controlled device for receiving the processed signals.

91. The system of claim 89, wherein the processing unit includes an integrated system configuration routine that adapts over time.

92. ^"The system of claim 90, wherein the controlled device comprises one or more of the group consisting of: a computer, a computer display, a mouse, a cursor, a joystick, a personal data assistant, a robot or robotic component, a computer controlled device, a teleoperated device, a communication device, a vehicle, an adjustable bed, an adjustable chair, a remote controlled device, a Functional Electrical Stimulator device, a muscle stimulator, an exoskeletal robot brace, an artificial or prosthetic limb, a vision enhancing device, a vision restoring device, a hearing enhancing device, a hearing restoring device, a movement assist device, a medical therapeutic equipment, a drug delivery apparatus, a medical diagnostic equipment, a bladder control device, a bowel control device, a human enhancement device, and a closed loop medical equipment.

93. The system of claim 89, wherein the biological interface system includes a console configured to receive input from a user, wherein the input includes instructions for adapting one or more parameters associated with the system.

94. The system of claim 93, wherein the processing unit is further configured to modify the one or more parameters based on the received input.

95. The system of claim 89, wherein extracting the neural spike portion includes filtering the multicellular signals to produce a neural signal, the neural signal including a neural spike portion.

96. The system of claim 95, wherein the filtering includes: receiving a reference signal associated with at least one component of the biological interface system; and determining an algorithm indicative of noise based on the reference signal.

97. The system of claim 96, wherein filtering further includes substantially suppressing the noise based on the determination.

98. The system of claim 96, wherein the at least one component includes an electrode associated with the sensor, the electrode being located in a region of substantially no voluntary neural activity.

99. The system of claim 96, wherein the reference signal includes a signal indicative of a noise level associated with the neural activity of the brain, the noise level including substantially no voluntary neural activity.

100. The system of claim 89, wherein extracting the neural spike portion of the neural signal includes separating the neural signal into a high- frequency portion and a low-frequency portion.

101. The system of claim 100, wherein the high-frequency portion includes the neural spike portion associated with the neural signal.

102. The system of claim 100, wherein the low-frequency portion includes a local field potential portion associated with the neural signal.

103. The system of claim 100, wherein extracting the neural spike portion includes filtering the neural signal.

104. The system of claim 103, wherein filtering the neural signal includes high-pass filtering the neural signal.

105. The system of claim 104, wherein high-pass filtering includes passing the neural signal through a high-pass filter with a cutoff frequency of about 250 Hz.

106. The system of claim 103, wherein filtering the neural signal includes low-pass filtering the neural signal.

107. The system of claim 106, wherein low-pass filtering includes passing the neural signal through a law-pass filter with a cutoff frequency of about 250 Hz.

108. The system of claim 89, wherein determining whether the neural spike portion includes a neural spike comprises: detecting an envelope associated with the neural spike portion; comparing the detected envelope with a threshold value; and determining, if the detected envelope exceeds the threshold value, that the neural spike portion contains a neural spike.

109. The system of claim 108, wherein determining whether the neural spike portion includes a neural spike further comprises: passing the neural spike portion through a boxcar smoothing function to produce a signal indicative of a weighted amplitude associated with the neural spike portion over a predetermined time period; comparing the signal indicative of the weighted amplitude with a threshold value; and determining, if the signal indicative of the weighted amplitude exceeds the threshold value, that the neural spike portion contains a neural spike.

110. The system of claim 89, wherein identifying the neural spike includes projecting samples indicative of the correlation on a feature space.

111. The system of claim 110, wherein the feature space includes a two-dimensional feature space.

112. The system of claim 111 , wherein a definition of a neural spike is configured to adapt as additional samples are projected on the feature space.

113. The system of claim 110, wherein identifying the neural spike includes: identifying a cluster of correlated data in the feature space; and associating the cluster with a voluntary stimulus of the patient.

114. The system of claim 89, wherein the voluntary stimulus includes a signal corresponding to a desired limb movement.

115. The system of claim 114, wherein the desired limb movement includes a movement of one or more fingers.

116. The system of claim 114, wherein the desired limb movement includes a hand movement.

117. The system of claim 114, wherein the desired limb movement includes an arm movement.

118. A system for measuring impedance of an electrode associated with a biological interface system, comprising: a channel amplifier communicatively coupled to an electrode of the biological interface system, wherein the electrode is configured to collect multicellular signals associated with a biological system; a reference amplifier communicatively coupled to a reference electrode, the reference electrode configured to collect noise data associated with the biological system; a signal generator configured to periodically provide a test signal to the biological system for measuring the impedance of the electrode; and a processing device coupled to the channel amplifier and the reference amplifier, wherein the processing device is configured to: receive multicellular signals from the channel amplifier when the signal generator is not coupled to the biological system; and determine an impedance of each electrode when the signal generator is coupled to the biological system.

119. The system of claim 118, wherein at least a portion of each electrode is disposed within the brain tissue of a patient.

120. The system of claim 118, wherein the multicellular signals comprise signal indicative of neurological activity associated with a patient.

121. The system of claim 118, further comprising a software program interface communicatively coupled to the signal generator and the processing device, wherein the software program interface is configured to: provide a command to the signal generator to generate the test signal; receive data indicative of the impedance of each electrode; and display the data indicative of the impedance of each electrode on an electrode map.

122. The system of claim 118, wherein the signal generator is selectively coupled to the biological system via a switching device arranged in parallel with the signal generator.

123. The system of claim 118, wherein the processing device is configured to: collect noise data associated with the biological system when the signal generator is not coupled to the biological system; and adjust the collected multicellular signals based on the collected noise data to produce corrected multicellular signals.

124. The system of claim 118, further comprising a differential amplifier communicatively coupled to the channel amplifier and the reference amplifier, wherein the differential amplifier is configured to determine a voltage difference between the multicellular signals and the noise data to produce corrected multicellular signals.

125. The system of claim 118, wherein the processing device is configured to: collect a voltage signal associated with the channel amplifier and the reference amplifier in response to the test signal provided by the signal generator; adjust the received voltage signal associated with the channel amplifier based on the voltage signal associated with the reference amplifier; and determine the impedance of each electrode based on a difference between the test signal and the adjusted voltage signal associated with the channel amplifier.

126. The system of claim 118, further comprising a sensor array that includes a plurality of electrodes, the sensor array being disposed substantially within the skull of a patient.

127. The system of claim 118, wherein the channel amplifier includes a plurality of channel amplifiers and the electrode includes a plurality of electrodes, such that each of the plurality of channel amplifiers is communicatively coupled to one or more electrodes configured to collect multicellular signals associated with the biological system.

128. A system for measuring impedance of one or more electrodes associated with a biological interface system, comprising: a channel amplifier communicatively coupled to an electrode of the biological interface system, wherein the electrode is configured to collect multicellular signals associated with a biological system; a reference amplifier communicatively coupled to a reference electrode and configured to collect noise data associated with the biological system; a pedestal mounted to the skull of a patient and electrically coupled to the biological system, wherein the pedestal is configured to provide electrical signals to the biological system; a signal generator communicatively coupled to the pedestal and adapted to periodically provide a test signal to the pedestal; and a processing device coupled to the channel amplifier and the reference amplifier and configured to determine an impedance of the electrode when the signal generator is electrically coupled to the biological system.

129. The system of claim 128, wherein the channel amplifier comprises: a voltage follower, an output of which is electrically coupled to the processing device; a shunt resistor electrically coupled to an input of the voltage follower; and a capacitor electrically coupled in series with the electrode.

130. The system of claim 128, wherein the pedestal provides a plurality of electrical connections for communicating electrical signals through the skull of the patient.

131. The system of claim 128, wherein the pedestal includes a ground electrode communicatively coupled to the brain of the patient for delivering the test signal to the biological system.

132. The system of claim 128, wherein the biological system includes a neurological system of a patient, and the multicellular signals include signals indicative of neurological activity associated with the patient.

133. The system of claim 128, further including a software program interface communicatively coupled to the signal generator and the processing device, wherein the software program interface is configured to: provide a command to the signal generator to generate the test signal; receive data indicative of the impedance of the electrode; and display the data indicative of the impedance of the electrode on an electrode map.

134. The system of claim 128, wherein the signal generator is selectively coupled to the biological system via a switching device arranged in parallel with the signal generator.

135. The system of claim 128, wherein the processing device is further configured to: collect noise data associated with the biological system when the signal generator is not coupled to the biological system; and adjust the collected multicellular signals based on the collected noise data to produce a corrected multicellular signal.

136. The system of claim 128, further comprising a differential amplifier coupled to the channel amplifier and the reference amplifier, wherein the differential amplifier is configured to determine a voltage difference between the multicellular signals and the noise data to produce corrected multicellular signals.

137. The system of claim 128, wherein, to determine the impedance of the electrode, the processing device is configured to: collect a voltage signal associated with the channel amplifier and the reference amplifier in response to the test signal provided by the signal generator; adjust the received voltage signal associated with each of the plurality of channel amplifiers based on the voltage signal associated with the reference amplifier; and determine the impedance of each electrode based on a difference between the test signal and the adjusted voltage signal associated with the channel amplifier.

138. The system of claim 128, further comprising a sensor array comprising a plurality of electrodes, wherein the sensor array is disposed substantially within the skull of a patient such that each of the plurality of electrodes is in contact with brain tissue of the patient.

139. The system of claim 128, wherein the channel amplifier includes a plurality of channel amplifiers and the electrode includes a plurality of electrodes, such that each of the plurality of channel amplifiers is communicatively coupled to a particular electrode.

140. A method of determining an impedance of an electrode of a biological interface system comprising: periodically providing a test signal to a biological system; receiving electrical signals from an electrode disposed within the biological system; receiving noise data from a reference electrode disposed within the biological system; adjusting the electrical signals based on the noise data to compensate for the noise associated with the biological system; determining an impedance of the electrode when the test signal is provided to the biological system based on a signal loss associated with the electrical signals; and processing the electrical signals to extract multicellular signal data from the biological system.

141. The method of claim 140, wherein the electrode comprises a plurality of electrodes, and the method further comprises determining impedances of the plurality of electrodes simultaneously.