CN110392549B

CN110392549B - Systems, methods and media for determining brain stimulation that elicits a desired behavior

Info

Publication number: CN110392549B
Application number: CN201880016157.XA
Authority: CN
Inventors: P·K·皮利; M·D·霍华德; H·霍夫曼; T-C·卢; 倪康宇; D·W·佩顿
Original assignee: HRL Laboratories LLC
Current assignee: HRL Laboratories LLC
Priority date: 2017-05-03
Filing date: 2018-04-23
Publication date: 2022-02-11
Anticipated expiration: 2038-04-23
Also published as: WO2018204119A1; CN110392549A; EP3618707A4; EP3618707A1

Abstract

The present disclosure relates to systems, methods, and media for determining brain stimulation that elicits a desired behavior. A system for inducing a desired behavioral effect using current stimulation is described. The brain monitoring subsystem includes monitoring electrodes for sensing brain activity, and the brain stimulation subsystem includes stimulation electrodes for applying current stimulation. Registering the multi-scale distributed data into a graphical representation. The system identifies a sub-graph in the graphical representation and maps the sub-graph to a conceptual feature, generating a conceptual lattice that relates the conceptual feature to a behavioral effect. Finally, the current stimulus to be applied to produce the behavioral effect is determined.

Description

Systems, methods and media for determining brain stimulation that elicits a desired behavior

Cross Reference to Related Applications

This is a partial continuation of U.S. non-provisional application No.14747407 entitled "System and Method for Determining a compliance of States to Assign to reach of a compliance of actions" filed on 23.6.2015, application No.14747407 is a non-provisional application No.62/015871 entitled "System and Method for Determining a compliance of States to Assign to reach of a compliance of actions" filed 23.2014, which is hereby incorporated by reference in its entirety.

This is also a provisional application of U.S. provisional application No.62/500500 entitled "Method and Apparatus to determination optical Brain Stimulation to indication determined Behavior," filed on 3.5.2017, which is hereby incorporated by reference in its entirety.

Technical Field

The present invention relates to a system for eliciting a desired behavior, and more particularly to a system for eliciting a desired behavior using determined brain stimulation.

Background

The latest techniques for finding invasive and non-invasive stimulation patterns that enhance specific behavioral functions are essentially based on trial and error, an expensive experimental procedure. For example, if there are 8 scalp electrodes and 20 possible locations with 10 intensity levels and 5 frequency levels for non-invasive stimulation, then in a brute force fashion, 6298500 trials would be required to determine the best stimulation lead (montage) for the subject.

A system for rapid analysis of large-scale neural data, called THUNDER (see list of incorporated references, reference 8), is likely only able to find a subset of cells that are regulated by some aspect of stimulation/behavior. Simulated annealing, which is the latest technique for many locally very small difficult optimization problems, requires lengthy adjustment of the annealing parameters. It is slow and must be restarted from the beginning as the optimization target moves.

Behavioral enhancement has been attempted by weak current stimulation guided by a crude understanding of underlying neural processes such as described in reference bibliographies 14 and 19, and typically involves directing stimulation generally only to the prefrontal cortex (PFC) (see reference bibliographies 5, 11, 12 and 13), or one or two brain regions known to be specialized for a particular behavior (see reference bibliography 17) or identified by a clinician (see reference bibliography 7).

There is no prior art method to produce any desired behavioral effect. Thus, there is a continuing need for a method of calculating optimal stimulus application based on available data from experiments conducted without requiring trial and error procedures.

Disclosure of Invention

The present invention relates to a system for eliciting a desired behavior, and more particularly to a system for eliciting a desired behavior using determined brain stimulation. The system comprises: a brain monitoring subsystem comprising a set of monitoring electrodes for sensing brain activity; and a brain stimulation subsystem including a set of stimulation electrodes for applying current stimulation. The system also includes one or more processors and a non-transitory computer-readable medium having executable instructions encoded thereon such that, when executed, the one or more processors perform a plurality of operations. Registering a set of multi-scale distributed data into the graphical representation, wherein at least a subset of the set of multi-scale distributed data is sensed brain activity. A sub-graph is identified in the graphical representation and the sub-graph is mapped to a set of conceptual features, generating a conceptual lattice that relates the set of conceptual features to behavioral effects. The system then determines a current stimulus to be applied to produce a behavioral effect, and causes the current stimulus to be applied via the set of stimulation electrodes.

In another aspect, the graphical representation includes a plurality of nodes, each node representing a data item in the set of multi-scale distributed data, and edges between the plurality of nodes representing relationships between the data items, wherein the relationships are topological, statistical, and/or causal relationships.

In another aspect, the set of multi-scale distributed data includes electroencephalographic data recorded from the set of monitoring electrodes as a result of stimulation leads applied via the set of stimulation electrodes.

In another aspect, the set of multi-scale distributed data is transformed into currents in voxels of the brain volume by means of forward simulation, and the currents are integrated into the graphical representation.

In another aspect, a set of stimulation electrode placement locations and parameters are identified that can reproduce behavioral effects.

In another aspect, the identified set of stimulation electrode placement locations and parameters are applied via a brain stimulation subsystem, and wherein the brain monitoring subsystem monitors behavioral effects, wherein if the behavioral effects are not satisfactory, the brain monitoring subsystem updates the graphical representation and the one or more processors perform operations to determine new current stimulation to apply to produce satisfactory behavioral effects.

In another aspect, self-organizing thresholds (SOCs) are used to search for electrode locations or settings for brain stimulation.

In another aspect, forward simulations are used to predict behavioral outcomes for various brain stimuli.

Finally, the present invention also includes a computer program product and a computer implemented method. The computer program product includes computer-readable instructions stored on a non-transitory computer-readable medium that are executable by a computer having one or more processors such that, when the instructions are executed, the one or more processors perform the operations listed herein. Alternatively, a computer-implemented method includes acts of causing a computer to execute such instructions and perform the resulting operations.

Drawings

The objects, features and advantages of the present invention will become apparent from the following detailed description of various aspects of the invention, taken in conjunction with the following drawings in which:

FIG. 1 is a block diagram illustrating components of a system for inducing desired behavior in accordance with some embodiments of the present disclosure;

FIG. 2 is a diagram of a computer program product according to some embodiments of the present disclosure;

FIG. 3 is a diagram of an overview of stimulus patterns found to produce desired behavior according to an embodiment of the present disclosure;

FIG. 4 is a diagram of a knowledge representation framework according to an embodiment of the present disclosure;

FIG. 5 is a diagram of generation of assumptions about a mesh and a Bayesian (Bayesian) network, according to an embodiment of the present disclosure;

FIG. 6 is a flow diagram illustrating a system for causing a desired behavior according to an embodiment of the present disclosure;

FIG. 7 is a diagram relating neurosciences data features to the behavior of correlating to finding nerves behind memory enhancement, according to an embodiment of the present disclosure;

FIG. 8 is a diagram of a signal diagram according to an embodiment of the present disclosure;

FIG. 9 is a diagram of using a sparse Support Vector Machine (SVM) and Sparse Canonical Correlation Analysis (SCCA) for dimensionality reduction, according to an embodiment of the present disclosure;

FIG. 10 is a diagram of using a self-organizing criticality (SOC) search according to an embodiment of the present disclosure; and

fig. 11 is a diagram of a subject undergoing monitoring and stimulation via electrodes, according to an embodiment of the present disclosure.

Detailed Description

The following description is presented to enable any person skilled in the art to make and use the invention, and is incorporated in the context of a particular application. Various modifications and uses will become apparent to those skilled in the art in light of the teachings herein, and the generic principles defined herein may be applied in a wide variety of ways. Thus, the present invention is not intended to be limited to the aspects shown, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

In the following detailed description, numerous specific details are set forth in order to provide a more thorough understanding of the present invention. It will be apparent, however, to one skilled in the art that the present invention may be practiced without necessarily being limited to these specific details. In other instances, well-known structures and devices are shown in block diagram form, rather than in detail, in order to avoid obscuring the present invention.

The reader's attention is directed to all papers and documents which are filed concurrently with this specification and which are open to public inspection with this specification, and the contents of all such papers and documents are incorporated herein by reference. All the features disclosed in this specification (including any accompanying claims, abstract and drawings) may be replaced by alternative features serving the same, equivalent or similar purpose, unless expressly stated otherwise. Thus, unless expressly stated otherwise, each feature disclosed is one example only of a generic series of equivalent or similar features.

Furthermore, any element in the claims that does not explicitly recite "means" or "step" for performing a specified function is not to be construed as an "means" or "step" clause as specified in 35 u.s.c. section 112, paragraph 6. In particular, the use of "step" or "action" in the claims herein is not intended to invoke the provisions in section 6 of section 112 of 35 u.s.c.

Before describing the present invention in detail, a list of cited references is first provided. Next, a description of various principal aspects of the present invention is provided. Finally, specific details of various embodiments of the invention are provided to give an understanding of the specific aspects.

(1) List of incorporated references and patent bibliography

The following references are cited throughout this application and incorporated herein by reference. For clarity and convenience, the references are listed herein as a central resource for the reader. The following references are hereby incorporated by reference as if fully set forth herein. The references are incorporated by reference in this application to the corresponding reference headings.

1.Ardekani B.A.，Guckemus S.，Bachman A.，Hoptman M.J.，Wojtaszek M.，&Nierenberg J.，Quantitative comparison of algorithms for inter-subject registration of 3D volumetric brain MRI scans.Journal of neuroscience methods，142(1)，67-76，2005.

2.Avants B.B.，Cook P.A.，Ungar L.，Gee J.C.，&Grossman M.，Dementia induces correlated reductions in white matter integrity and cortical thickness：a multivariate neuroimaging study with sparse canonical correlation analysis.Neuroimage，50(3)，1004-1016，2010.

3.Bak P.，Tang C.，&Wiesenfeld K.，Self-organized criticality：An explanation of the l/f noise.Physical review letters，59(4)，381，1987.

4.Barzel B.，&Barabási A.L.，Network Link Prediction by Global Silencing of Indirect Correlations.Nature biotechnology，2013.

5.Chrysikou EG，Hamilton RH，Coslett HB，Datta A，Bikson M，Thompson-Schill SL.Noninvasive transcranial direct current stimulation over the left prefrontal cortex facilitates cognitive flexibility in tool use.Cogn Neurosci.4(2)：81-9，2013.

6.Correa N.M.，et al.Multi-set Canonical Correlation Analysis for the Fusion of Concurrent Single Trial ERP and Functional MRI，Neuroimage 50.4，1438-1445，2010.

7.Dmochowski JP，Datta A，Bikson M，Su Y，Parra LC.Optimized multi-electrode stimulation increases focality and intensity at target.J Neural Eng.8(4)：046011，2011.

8.Freeman J.，Vladimirov N.，Kawashima T.，Mu Y.，Sofroniew N.J.，Bennett D.V.，&Ahrens M.B.，Mapping brain activity at scale with cluster computing.Nature Methods，11(9)，941-950，2014.

9.Hardoon D.R.，and Shawe-Taylor J.，Sparse Canonical Correlation Analysis，Machine Learning 83.3，331-353，2011.

10.Hinton G.E.，Salakhutdinov R.R.，Reducing the Dimensionality of Data with Neural Networks，Science，313：5786，2006.

11.Javadi AH，Cheng P.Transcranial direct current stimulation(tDCS)enhances reconsolidation of long-term memory.Brain Stimulation，2012.

12.Javadi AH，Walsh V.Transcranial direct current stimulation(tDCS)of the left dorsolateral prefrontal cortex modulates declarative memory.Brain Stimulat.，5(3)：231-41，2012.

13.Marshall L，Helgadóttir H，

M，Bom J.Boosting slow oscillations during sleep potentiates memory.Nature.444(7119)：6103，2006.

14.Miniussi C，Cappa SF，Cohen LG，Floel A，Fregni F，Nitsche MA，et al.Efficacy of repetitive transcranial magnetic stimulation/transcranial direct current stimulation in cognitive neurorehabilitation.Brain Stimulat.1(4)：326-36，2008.

15.Ni，C.and Lu T-C.，Information Dynamic Spectrum Characterizes System Instability toward Critical Transitions，EPJ Data Science，In Press，2014.

16.Ramirez RR，Makeig S.，Neuroelectromagnetic source imaging of spatiotemporal brain dynamical patterns using frequency-domain independent vector analysis(IVA)and geodesic sparse Bayesian learning(gSBL).13th Annual Meeting of the Organization for Human Brain Mapping.Chicago，USA，2007.

17.Reis J，Schambra HM，Cohen LG，Buch ER，Fritsch B，Zarahn E，et al.Noninvasive cortical stimulation enhances motor skill acquisition over multiple days through an effect on consolidation.Proc NatlAcad Sci USA.106(5)：1590-5，2009.

18.Sui，Jing，et al.A Review of Multivariate Methods for Multimodal Fusion of Brain Imaging Data.，Journal of neuroscience methods 204.1，68-81，2012.

19.Utz KS，Dimova V，

K&Kerkhoff G.Electrified minds：transcranial direct current stimulation(tDCS)and galvanic vestibular stimulation(GVS)as methods of noninvasive brain stimulation in neuropsychology-a review of current data.and future implications.Neuropsychologia.48(10)：2789-810，2010.

20.Wolters CH，Anwander A，Tricoche X，Weinstein D，Koch MA，MacLeod RS.，Influence of tissue conductivity anisotropy on EEG/MEG field and return current computation in a realistic head model：A simulation and visualization study using high-resolution finite element modeling.Neuroimage 30(3)：813-826，2006.

21.Hoffmann H，Payton DW，Optimization by Self-Organized Criticality，Scientific Reports，Feb.5，2018.

22.Dmochowski JP，Koessler L，Norcia AM，Bikson M，Parra LC，Optimal use of EEG recordings to target active brain areas with transcranial electrical stimulation.Neuroimage 15；157：69-80，2017.

23.Rodrigo，QQ，Spike sorting.Scholarpedia，2(12)：3583，2007.

24.Fletcher R，Powell MJ.A rapidly convergent descent method for minimization.The Computer Journal 1；6(2)：163-8，1963.

(2) Main aspects of the invention

Various embodiments of the present invention include three "primary" aspects. A first aspect is a system for inducing a desired behavior. The system is typically in the form of a computer system running software or in the form of a "hard-coded" instruction set. The system may be incorporated into a wide variety of devices that provide different functions. The second main aspect is a method, usually in the form of software, run using a data processing system (computer). A third broad aspect is a computer program product. The computer program product generally represents computer readable instructions stored on a non-transitory computer readable medium, such as an optical storage device (e.g., a Compact Disc (CD) or a Digital Versatile Disc (DVD)) or a magnetic storage device (such as a floppy disk or a magnetic tape). Other non-limiting examples of computer readable media include hard disks, Read Only Memories (ROMs), and flash-type memories. These aspects will be described in more detail below.

A block diagram illustrating an example of the system of the present invention, namely computer system 100, is provided in fig. 1. The computer system 100 is configured to perform calculations, processes, operations, and/or functions associated with a program or algorithm. In one aspect, the specific processes and steps discussed herein are implemented as a series of instructions (e.g., a software program) that are resident in a computer readable storage unit and executed by one or more processors of the computer system 100. When executed, the instructions cause the computer system 100 to perform specific actions such as those described herein and to exhibit specific behaviors.

Computer system 100 may include an address/data bus 102 configured to communicate information. In addition, one or more data processing units, such as a processor 104, are coupled to the address/data bus 102. The processor 104 is configured to process information and instructions. In one aspect, the processor 104 is a microprocessor. Alternatively, the processor 104 may be a different type of processor, such as a parallel processor, an Application Specific Integrated Circuit (ASIC), a Programmable Logic Array (PLA), a Complex Programmable Logic Device (CPLD), or a Field Programmable Gate Array (FPGA).

Computer system 100 is configured to use one or more data storage units. The computer system 100 may include a volatile memory unit 106 (e.g., random access memory ("RAM"), static RAM, dynamic RAM, etc.) coupled to the address/data bus 102, wherein the volatile memory unit 106 is configured to store information and instructions for the processor 104. The computer system 100 may also include a non-volatile memory unit 108 (e.g., read only memory ("ROM"), programmable ROM ("PROM"), erasable programmable ROM ("EPROM"), electrically erasable programmable ROM ("EEPROM"), flash memory, etc.) coupled to the address/data bus 102, wherein the non-volatile memory unit 108 is configured to store static information and instructions for the processor 104. Alternatively, the computer system 100 may execute instructions retrieved from an online data storage unit (such as in "cloud" computing). In one aspect, computer system 100 may also include one or more interfaces (such as interface 110) coupled to address/data bus 102. The one or more interfaces are configured to enable the computer system 100 to interface with other electronic devices and computer systems. The communication interfaces implemented by the one or more interfaces may include wired communication techniques (e.g., serial cable, modem, network adapter, etc.) and/or wireless communication techniques (e.g., wireless modem, wireless network adapter, etc.).

In one aspect, computer system 100 may include an input device 112 coupled to address/data bus 102, wherein input device 112 is configured to communicate information and command selections to processor 100. According to one aspect, input device 112 is an alphanumeric input device (such as a keyboard) that may include alphanumeric and/or function keys. Alternatively, the input device 112 may be an input device other than an alphanumeric input device. In one aspect, the computer system 100 may include a cursor control device 114 coupled with the address/data bus 102, wherein the cursor control device 114 is configured to communicate user input information and/or command selections to the processor 100. In one aspect, cursor control device 114 is implemented using a device such as a mouse, trackball, trackpad, optical tracking device, or touch screen. Notwithstanding the foregoing, in one aspect, cursor control device 114 is also directed and/or actuated via input from input device 112 (such as in response to the use of specific keys and key sequence commands associated with input device 112). In an alternative aspect, cursor control device 114 is configured to be guided by voice commands.

In one aspect, computer system 100 may also include one or more optional computer usable data storage devices (such as storage device 116) coupled to address/data bus 102. Storage 116 is configured to store information and/or computer-executable instructions. In one aspect, storage device 116 is a storage device such as a magnetic or optical disk drive (e.g., hard disk drive ("HDD"), floppy disk, compact disk read only memory ("CD-ROM"), digital versatile disk ("DVD")). In accordance with one aspect, a display device 118 is coupled with the address/data bus 102, wherein the display device 118 is configured to display video and/or graphics. In one aspect, the display device 118 may include a cathode ray tube ("CRT"), a liquid crystal display ("LCD"), a field emission display ("FED"), a plasma display, or any other display device suitable for displaying video and/or graphical images recognizable to a user as well as alphanumeric characters.

Computer system 100 presented herein is an example computing environment in accordance with an aspect. However, non-limiting examples of computer system 100 are not strictly limited to computer systems. For example, one aspect provides that computer system 100 represents a data processing analysis that may be used in accordance with various aspects described herein. Moreover, other computing systems may also be implemented. Indeed, the spirit and scope of the present technology is not limited to any single data processing environment. Thus, in one aspect, one or more operations of various aspects of the technology are controlled or implemented using computer-executable instructions, such as program modules, executed by a computer. In one implementation, such program modules include routines, programs, objects, components, and/or data structures that are configured to perform particular tasks or implement particular abstract data types. In addition, one aspect provides that one or more aspects of the present technology are implemented using one or more distributed computing environments (such as environments where tasks are performed by remote processing devices that are linked through a communications network or such as environments where various program modules are located in both local and remote computer storage media including memory-storage devices).

A diagram of a computer program product, i.e., a storage device, embodying the present invention is shown in fig. 2. The computer program product is depicted as a floppy disk 200 or an optical disk 202 (such as a CD or DVD). However, as mentioned previously, the computer program product generally represents computer readable instructions stored on any compatible non-transitory computer readable medium. The term "instructions," as used with respect to the present invention, generally refers to a set of operations performed on a computer, and may represent an entire program or a separate, stand-alone software module. Non-limiting examples of "instructions" include computer program code (source or object code) and "hard-coded" electronic devices (i.e., computer operations encoded into a computer chip). "instructions" are stored on any non-transitory computer readable medium (such as in the memory of a computer or on floppy disks, CD-ROMs, and flash drives). In either case, the instructions are encoded on a non-transitory computer readable medium.

(3) Details of various embodiments

A system is described that discovers the relationship between neural activity, applied current stimulation, and behavioral performance (such as memory enhancement) from neural data (e.g., generated by programs such as DARPA (department of defense advanced research project) RAM (restoring active memory) and submines (system-based neural technology for emerging therapies)), and calculates optimal brain stimulation leads for individual subjects or across the general population that will produce the desired behavioral effects. Non-limiting examples of desired behavioral effects include enhancement of selected memory (e.g., beneficial to task performance) and weakening of selected memory (e.g., memory that causes trauma or impairs task performance). The system operates in closed loop, monitoring the brain at each round, calculating and applying new stimulation patterns, and improving behavioral effects until the desired effect is achieved. The only element of the system described herein is a forward model that predicts the behavioral results of different stimuli, and uses the search method as an optimization technique to identify the ideal stimulus pattern that produces the desired behavioral result.

There are currently many brain stimulation systems on the market, but none of them provide a forward model that can predict the behavioral effects of specific stimulation leads. This prediction makes the system according to embodiments of the present disclosure different from the prior art and makes brain stimulation science proceed from intuitive based on a resting original understanding of brain function to educated analytical optimization. Although the invention described herein can be used as an intelligent closed-loop stimulation system, the forward model on which the invention is based can also provide answers that can clearly define queries (such as "what are the basic brain attributes underlying a given behavioral result.

Understanding brain function involves a large amount of multi-modal data ranging in spatial scale from single neuron spiking activity to local field potentials and electroencephalographic (EEG) signals. Data also ranges widely in time scale from low to high frequency spectral interactions between various neural entities. A method according to an embodiment of the present disclosure includes a technique to register multi-scale distributed neural data and induced current flow distributions in brain volumes across subjects and within subjects across trials. For optimization, a recently discovered approach to solving a high dimensional optimization problem that is approximately 8 times faster than simulated annealing, the self-organizing criticality (SOC) search (see reference bibliography 21 for description of SOC search) is suitable and used to solve the inverse problem of determining the optimal stimulation patterns that may induce currents in brain volumes for specific behavioral states across the trial and across the subject. Each of these aspects will be described in further detail below.

(3.1) overview

Fig. 3 illustrates how the system described herein may be applied to find stimulation patterns that may produce desired behavior. The datamation (i.e., domain-specific data acquisition 300) extracts and acquires relevant experimental and computational data, which includes both quantitative and qualitative elements, and integrates these elements into the Knowledge Representation (KR) 302. Datamation involves the formation of pairwise topological, statistical, and causal relationships within data, and their synthesis into multi-layer charts, each annotated with qualitative and quantitative information of the experimental design. The datamation tools are capable of extracting time-varying multi-scale correlations and causal interactions between identified neural entities to facilitate deeper discovery of behavior-based functional connections. Datamation refers to the conversion of phenomena or data into a computable format that aids in knowledge extraction.

KR 304 outputs ranking-ordering hypotheses to domain-specific discovery/digitization tool 306 (described in detail below) to further guide experimental design and computational data analysis with domain-specific computational models. Non-limiting examples of domains include complex domains with large data sets like neuroscience, climate science, gene protein disease networks, smart grids, wireless communication systems, and autonomous systems. These tools analyze and perform functional inference for rapid brain classification to predict memory behavior (element 308) and optimization of stimulation patterns for memory enhancement (element 310).

The multimodal data can be described within a signal graph such as that shown in FIG. 8 or across a large number of annotated graphs such as that shown in FIG. 4. FIG. 8 illustrates an adjacency matrix that may capture a graph representation of links within and across models. As shown in FIG. 5, after identifying local regions in these charts, the auto-encoder (element 508) computes conceptual features for the concept lattice 512 and the Bayesian network 514. An auto-encoder as known to those skilled in the art is a recursive network of unsupervised learning for efficient encoding. The purpose of the encoder is to learn the representation (encoding) of a set of data, usually for dimensionality reduction. Hypotheses for the concept lattice 512 and the bayesian network 514 may then be generated (element 516), and the hypotheses may retrieve corresponding entities, relationships, and functions from the multi-layer diagram (fig. 5).

For example, if the data domain is neuroscience, the hypothesis may be which neurological features (a large amount of available data from EEG and implanted electrodes over time and across subjects) emphasize successful memory coding. If the data fields are property predictions in the materials science, it is assumed that it is possible to synthesize which set of material properties to achieve the desired behavior for each material in order to discover new polymeric materials with desired properties from historical experimental and computational data of polymer formulations and monomer compositions.

(3.2) digitization

Fig. 4 shows source data from EEG or invasive electrodes (converted to a space-time diagram 400) compiled into a graphical representation 402 of causal (element 404), statistical (element 406) and topological (element 408) relationships between neural data. The raw brain data is processed and converted into a cross-modal spatiotemporal graph corresponding to the memory network in the brain (element 400). These networks are the sub-regions of the brain believed to be primarily responsible for the behaviors of interest, including the regions of the hippocampal structures, the temporal lower lobe, and the prefrontal cortex for memory performance.

For data recorded from invasive electrodes (iCS, used for invasive current stimulation), the nodes in the graph (depicted as circles (e.g., n 1410 and n 2412)) would be voltage traces from individual neurons and Local Field Potentials (LFPs) extracted from extracellular recordings using standard frequency filtering, spike sequencing, and source isolation techniques with implanted microelectrode arrays in the brain region of interest. Frequency filtering is a standard technique meant to derive the power spectral density of the EEG, which gives the power of the signal as a function of frequency. For the description of the spike ranking technique, reference is made to bibliography 23. For a description of the source separation technique, reference is made to bibliography 22. For data recorded from a non-invasive electrode (tCS, used for transcranial current stimulation), the nodes (elements 410 and 412) correspond to current sources from scalp EEG signals mapped to voxels in the standardized brain volume. Links between nodes within and between modalities are of different kinds related to underlying statistical relationships (element 406) and causal relationships (element 404).

The statistical linkage (element 406) is represented using time-resolved coherence spectra that capture time-varying multi-scale correlation structures for pairs of time series. The time-varying causal link (element 404) is computed using a robust version of the transfer entropy measure that substantially quantifies conditional mutual information between two random processes across multiple time scales and is invariant to their relative amplitudes. EEG sources and stimuli-induced currents are mapped to brain voxels using a forward model and labeled to provide qualitative data for building hierarchical clusters in the graph, and to enable discovery of functional interaction aspects between specific brain regions.

The instantiation of these brain networks is informed by the knowledge state of neurophysiology (i.e., by those skilled in the art) to limit the graph structure to meaningful and interpretable relationships, particularly between invasive and non-invasive modalities (e.g., LFP-EEG links and not spike-EEG links), and to inform the clusters using knowledge related to brain region function specialization. The multidimensional nodes and link characteristics are compressed into relevant features using a stack of automated encoders 508 (as described in reference bibliography 10) with data registered across the trial and subject. The auto-encoder 508 is an artificial neural network for unsupervised learning for efficient encoding. An auto-encoder learns the representation (or encoding) for a set of data. In the present invention described herein, an auto-encoder 508 is used for dimensionality reduction.

As shown in fig. 5, once the graphical representation 402 (or relationship graph) is prepared (where the topological relationships fall into the brain sub-region prefrontal cortex 500(PFC), the temporal sub-cortex 502(ITC), the hippocampal structures 504(HC) as evaluated by the EEG 506), the auto-encoder 508 runs on the local regions of the graphical representation 402, which form compact conceptual features 510 (or representations) that are arranged into a conceptual lattice 512 that is searched (i.e., by the SOC search) to find the most relevant regions to stimulate to produce the desired behavioral effect.

(3.2.1) time-resolved Power and coherence Spectroscopy (element 406)

A statistical relationship chart (graphical representation 402) is generated by using wavelet transforms in various neural and non-neural modalities to compute time-varying characteristics of the identified entities and the statistical relationships between them (element 406). For coherent spectra, first from the complex monent (Morlet) wavelet w (t, f) family (one for one frequency) as a function of time:

to convolve the time series at any node, wherein the Gaussian envelope σ_tIs inversely proportional to the frequency f and the coefficient a is set such that the total energy of each wavelet is equal to 1. The complex conjugate of one of the outputs is then multiplied element by element with the other outputs to obtain a coherent spectrum. This is equivalent to a time-resolved fourier spectrum of the cross-correlation between two time series. The time-resolved power spectrum is calculated similarly, the only difference being that it is linked to the autocorrelation.

(3.2.2) transfer entropy with invariant scale (element 404)

The time-varying causal links (element 404) are quantized in the multimodal graph/graphical representation (element 402) using a Transfer Entropy (TE) metric, which is essentially a measure of the direction of information flow between a pair of time series:

where τ is the time delay of information transfer. The above equation calculates the entropy of transfer from node j to node i (across node x)_jTo x_iLinked information flow of (1). p is a bayesian operator that evaluates the likelihood that the quantity in parentheses is true. The TE at each point in time T depends on the duration T for which the information transfer is considered and on the choice of τ. As memory usage is defined by multi-modal multi-scale brain measurements, it would be challenging to specify the optimal τ values globally. However, the duration T may be assumed to be 100 milliseconds (ms) in consideration of a typical frequency of the dominant θ rhythm in LFP within the memory lines in the brain, i.e., 10 hertz (Hz). Therefore, TE is calculated at the spectrum of time delays spanning the assumed duration T of 100 ms.

The system described herein also incorporates the recently developed concept of Associative Transfer Entropy (ATE), which extends TE by decomposing the associative state to distinguish the type of information being transferred (see bibliography 15). A simple example is to distinguish between positive and negative causal effects, rather than just obtaining an overall causal effect. ATE is implemented by constraining a particular associated state S to a set of all possible states { (x)_i，t+τ，x_i，t，x_j，t) TE is decomposed in a subset of which, which allows one to quantify the amount of a particular information transfer.

TE and ATE are computed by means of a scale-invariant symbolization method to remove any time scale dependence. TE and ATE formulations are defined with respect to continuous random variables that are discretized by estimating their probability distribution using a symbolization method. The key step is to time-series { x } the continuous values_tIs transformed into a time sequence of symbols of suitable length n. For each time point t, n successive values { x ] are applied in ascending order_t，x_t+1，…，x_t+n-1And (6) sorting. The sequence of corresponding permutation indices is then the symbol time sequence for the discrete values

The t-th element of (1).

(3.2.3) Global silencing of Indirect links

A large number of pseudo-statistical and causal links (elements 406 and 404) within each brain region for non-invasive modalities will be instantiated based on the pseudo-relationships and may be suppressed. For example, even if there is no direct path from A to C, the causal link from A to C (element 404) may be erroneously inferred in the network A → B → C. To this end, a system according to an embodiment of the present disclosure is built based on the prior art (see reference bibliography 4) to address the challenges of graphs with multidimensional node and link characteristics. For each link type and feature descriptor 414, a matrix G is constructed that represents the structure of the local network in the brain region under consideration. The technique described in reference bibliography 4 operates on the correlation matrix G by the following calculation to preserve direct links while suppressing indirect links for simpler graph structures by exploiting implicit information flow in the network:

S＝(G-I+D((G-I)G))G^-1

where I is the identity matrix and d (M) sets the off-diagonal entries of M to zero.

(3.2.4) registration (element 302)

To register both measured and induced data for non-invasive modalities, the latest techniques forward and inverse models and a unique application of mapping techniques are used. Registration of EEG signals across the scalp of a subject is typically ensured by using a standard EEG cover (such as a 10-20 or 10-10 system) with a prescribed layout for electrode locations, where the spacing between adjacent electrodes is proportional to the individual head dimensions. However, in the memory use case described here, EEG electrode locations for different subjects are not registered due to the practical problem that it is not possible to implant the same electrodes in different subjects; the data is constrained by the location of the ports for the sensor array.

In addition, for the implanted array, EEG was recorded from the same electrode (tCS) used for transcranial current stimulation, so their positions were also varied between trials to optimize the specific stimulation pattern. To address this registration challenge, first, EEG signals (time series) are converted into the volumetric and temporal dynamics of current density sources within gray matter voxels in the brain, segmented according to T1 weighted MRI for individual subjects, using state-of-the-art methods for source localization. In particular, building a subject-specific forward model using finite element modeling (see reference book 20, which describes how the forward model is built using techniques of forward element modeling), and building a reverse model using multi-scale geodetic Sparse Bayesian Learning (SBL) with Laplacian priors (see reference book 16, which describes how the reverse model is built using SBL with Laplacian priors), is widely considered suitable for locating distributed sources in the domain of non-invasive imaging. These models make the problem of registration between subjects more manageable, given the state of the art methods for mapping structural and functional MRI data to standard brain templates. In particular, a non-linear non-parametric Automatic Registration Toolkit (ART) approach has been shown to outperform other approaches in reducing anatomical variability across subjects (see reference bibliography 1).

A similar problem is to register non-invasive stimulation leads between trials where stimulation is applied and between subjects. To this end, subject-specific forward models for high-resolution stimulation induced currents are built using finite element modeling and segmentation of individual subjects' anatomical MRI (T1) into different tissue classes (e.g., brain, skull, cerebrospinal fluid (CSF), electrodes) based on related state-of-the-art methods (see bibliography 7). The aforementioned ART protocol will then register the evoked currents within the brain volume between subjects. In this way, the solution of the present system can treat any position of the scalp electrodes for non-invasive stimulation.

(3.2.5) dimensionality reduction (element 302)

A unique dimensionality reduction method is used for challenging datamation of EEG signals obtained from different scalp electrode placement locations across a subject. "datamation" refers to the conversion of phenomena or data into a computable format that helps extract knowledge. The EEG-derived current density source image may be noisy considering that the number of voxels in the transformed EEG data is much larger than the number of actual electrode channels. In addition, the number of voxels is large enough to prohibit instantiating a graph for the EEG modality. To address these associated problems, an innovative tool for dimension reduction using sparse Support Vector Machines (SVMs) and Sparse Canonical Correlation Analysis (SCCA) is employed. Fig. 9 shows that the sparse SVM derives current density volumes (element 902) from the EEG based on the results of various behavioral tests (element 904) and reselects a discrimination voxel (element 900), then Sparse Canonical Correlation Analysis (SCCA) determines an additional subset of voxels (element 906) that are most correlated with the known voxel locations (element 908) of invasive LFP recordings.

Thus, the subset of most significant voxels is determined in two steps: firstly, selecting a voxel (using sparse SVM (element 900)) with the maximum discrimination between success and failure memory recall results in a corresponding retrieval test; and then further down-selected based on maximizing (using SCCA (element 906)) the overall correlation with the simultaneously recorded local field potential data (element 908) in the three brain regions.

For the second step, the SCCA projects the current density estimate for the discriminatory voxel set and the spatial localization LFP (element 908) onto the common feature space in order to determine the semantics of the underlying neuron activity pattern. From these semantics, the SCCA will learn the voxel subset (Y) of the EEG source whose time series is most correlated with the time series (X) of the low-dimensional ground truth from the invasive LFP measurements. Adding sparsity constraints to SCCA mitigates the effects of outlier EEG source estimation in neuronal data and is appropriate when many voxels may not provide information for neural decoding (see reference bibliography 2). Sparse SVMs successfully reduce the feature dimension by a factor of 50 to 100 while maintaining a high level of classifier accuracy.

SCCA is a robust, scale-invariant method that can be easily extended to adapt the nonlinear relationship between LFP (element 908) and EEG source estimation (element 902) using the kernel method (see reference bibliography 9), feature learning for multiple subjects for increased statistical power (see reference bibliography 6), and Independent Component Analysis (ICA) if it is determined that the model for multiple subject analysis requires feature independence within the subject but not feature correlation between them (see reference bibliography 18). This is the first example of SCCA (element 906) used to determine the spatiotemporal mapping between an invasive LFP (element 908) and an inferred current density source from a non-invasive EEG signal (element 902).

(3.3) discovery (element 312)

The unique analysis and discovery tools of the system according to embodiments of the present disclosure obtain output from the above-described datamation system in response to querying brain attributes in a particular modality and stimulus-induced currents linked to strong memory-encoded behavioral outcomes in a high-level conceptual space. This section details how these tools can help find optimal stimulation leads to promote behavioral enhancement.

Fig. 5 illustrates converting the processed data into a concept lattice 512 and then searching the concept lattice to find the best behavioral result and identify concepts that represent the desired brain state (element 314). Recall that those conceptual features 510 are comprised of an automatically encoded (element 508) portion of the input data graph (element 402) from which the target brain state (element 508) of the individual brain voxel is identified. Once the desired brain state is known (element 314), the key functional inference challenge is to solve the difficult high-dimensional inverse mapping problem of estimating the stimulation pattern, which, whether applied invasively or non-invasively, leads to a specific current density distribution within the brain volume.

(3.3.1) optimal stimulation for memory enhancement (element 310)

As shown in fig. 10, a self-organizing critical (SOC) search is used to solve the inverse problem of estimating invasive and non-invasive neurostimulation leads that give rise to an arbitrary distribution of current density through the brain volume. For additional details regarding SOC searching, reference is made to bibliography 21 and U.S. application No.14/747407, which are incorporated herein by reference as if fully set forth herein.

As shown in fig. 6, the KR system 304 discovers and extracts a signature template that stimulates an evoked current based on a query (element 600) related to a behavioral phenomenon of interest (e.g., strong mnemonics) (element 602). For example, the query (element 600) may be of the form: "which currents lead to enhanced memory? ". This is possible because the KR system 304 acquires experimental data (i.e., a spatial lateral graph annotated with the relationship 604) related to memory retrieval results and associated multi-modal multi-scale brain measurements, as well as calculated data of estimated evoked currents 606 registered between subjects in voxel space. Inputs (element 610) for the datamation (element 300) and forward model 608 include, but are not limited to, EEG, spike, LFP, tCS, and iCS data.

To facilitate personalized optimization of memory-enhanced stimulation leads, signature templates of electrical currents are mapped to specific brain structures. The optimization algorithm/tool (element 310) then operates on the personalized template of desired currents (i.e., the forward model 608) to estimate compatible stimulation leads (i.e., the neurostimulation pattern 612 that will have the desired effect) that specify the scalp location of the electrodes and the stimulation current parameters (e.g., amplitude, frequency) at each electrode. The constraints of the optimization are the maximum number of electrodes, the maximum current per electrode and the maximum cumulative current from all electrodes. In summary, a system according to embodiments of the present disclosure builds a knowledge representation (element 304) from a set of data (element 614) and behavior (element 602), identifies which currents are primarily responsible for a desired behavioral effect (element 608), and then infers which neural stimulation patterns will produce the effect (element 612).

In one embodiment, an iterative optimization algorithm is used. In each iteration step, firstly, the SOC search is used to improve the placement of the electrodes and, secondly, the gradient descent (see bibliography 24 for the description of the gradient descent) is used to improve the parameters for a given placement. For electrode placement positions, a limited set of possible stimulation sites (e.g., 256 predetermined locations on the cranial lid) is used, and the goal is to find the best electrode pattern from these locations. SOC searches use a self-organizing threshold (SOC) process (see reference bibliography 3) to generate search patterns. As shown in fig. 10, the SOC process may be defined on any graph 1000. In each iterative step of the optimization, i.e., the sample sequence 1002 of avalanche, each avalanche (e.g., element 1004) is mapped one-to-one in the graph of the SOC process onto the graph defining the optimization problem (element 1006).

Fig. 10 illustrates an operation of SOC search to find a ground state of an octyl (Ising) spin glass. The Esin spin glass is an array of spins 1000 (e.g., with electromagnetic torque, electrons up or down) and coupling between spins such that the energy of the spins depends on the orientation (up or down) of the surrounding spins. The ground state is the arrangement of spin orientations that minimizes the total energy of all spins. To find the minimum, the SOC search iterates over the patterns obtained from the SOC process (e.g., avalanche sequence 1002 in the Bak-Tang-Wiesenfeld model (ref 3)). Here, individual avalanches 1004 are mapped onto the spin array 1006 and the spins to be flipped are marked to test whether a lower energy state can be obtained. If the change results in a lower energy, the resulting spin arrangement forms a new baseline for optimization, otherwise the new spin arrangement is discarded. This process is repeated until the desired optimum value is reached. For further details, reference is made to reference 21.

Due to the high dimensionality of the electrode placement location problem (N-105), the optimization is solved with an SOC search that iteratively finds and refines the location of the electrode that should be active. The idea is to initially approximate the solution with a coarser resolution and then iteratively use the estimated solution to locally increase the resolution (adding columns in the a matrix). This approach significantly improves the accuracy of the solution (reduces residual) and goes beyond the ability of the state-of-the-art inverse model for non-invasive stimulation to be able to handle spatially distributed current fields in the brain.

Fig. 7 illustrates how the knowledge representation (ultimately the conceptual lattice derived by steps 1-4 (described below)) relates neuroscience data conceptual features (elements 510) to behavior, and can be applied to find neural correlations behind memory enhancement. If the experiment involves different neurostimulation protocols, the system can be used to optimize the protocol to produce the best behavioral result. The brain signals are processed by a datamation module (step 1, element 700) which establishes a link to the relevant brain signals (step 2, element 702). Sub-graphs are identified by means of linked clustering (step 3, element 704) and sub-graphs are mapped to conceptual features (element 510) by means of an auto-encoder (step 4, element 706). The concept lattice correlates features with behavior (step 5, element 708) and the system maps back to find neural correlations behind memory enhancement (step 6, element 710). Fig. 8 illustrates a signal graph 800. The adjacency matrix of the graphical representation may capture intra-model and cross-model links. For example, the intra-model links are correlations between EEG channels (element 506) and the cross-model links are correlations between EEG and LFP (element 610). The entire adjacency matrix is an example of a statistical layer 404 in knowledge graph 402.

In summary, as illustrated in fig. 11, the invention described herein is a method and system for influencing a desired behavioral outcome by sensing and intelligently stimulating the brain of a subject 1100 (or a group of subjects) that includes a brain monitoring subsystem 1102 that incorporates a set of monitoring electrodes 1104 that can sense brain state (element 314) either invasively (i.e., chronically implanted electrodes providing neuron or LFP data) or non-invasively (i.e., via EEG). A large number of monitor electrodes 1104 will provide high resolution results. The system additionally includes a brain stimulation subsystem 1104 (shown in conjunction with a brain monitoring subsystem) that incorporates a set of stimulation electrodes 1106 that can apply current stimulation invasively (i.e., chronically implanted electrodes) or non-invasively (i.e., transcranial current stimulation). As will be appreciated by those skilled in the art, the brain monitoring and brain stimulation subsystem 1104 may also be a component of a separate device. In addition, although multiple monitoring and stimulation electrodes are used in the invention described herein, for illustrative purposes, only one of each type is shown in FIG. 11. In one embodiment, the monitoring electrode and the stimulation electrode are co-located.

Further, the forward model predicts the behavioral results of different stimuli (element 904), and the search method (FIG. 10) is used to identify ideal stimulus patterns associated with the desired behavioral results. The forward model consists of the following parts: the graph representation 402, its nodes (e.g., elements 410 and 412) describe data, and edges describe relationships between data items. These relationships are topological (element 408), statistical (element 406), and causal (element 404) (see FIG. 4). The data of graphical representation 402 may include stimulation leads 316 (if any) applied to the brain non-invasively or via chronically implanted electrodes. Additionally, the data may include EEG data recorded from the monitoring electrodes as a result of the stimulation leads 316 applied via the stimulation electrodes.

The data diagram/graphical representation 402 is processed in a series of steps including one or more of the following. Statistical relationships between graph nodes are computed using wavelet transforms (element 406), and causal relationships are quantized with Transfer Entropy (TE) and Associated Transfer Entropy (ATE) (element 404), which transform continuous time values into a symbol time series. The pseudo-indirect linking is removed using the correlation matrix technique described in reference bibliography 4. The measurements and resulting data of the non-invasive modality are transformed into resulting currents (element 606) in voxels of the brain volume by means of a forward model (element 608), and those currents (element 606) are integrated into the graphical representation 402. Data from multiple trials and/or multiple subjects are registered by deformation (warp).

Dimension reduction was performed on noisy EEG data across subjects using a Support Vector Machine (SVM) (element 900) and Sparse Canonical Correlation Analysis (SCCA) (element 302 and fig. 9) to select the most significant voxels. The local regions of the graphical representation 402 are compressed with an auto-encoder 508 and become concepts 510 in a concept lattice 512. The concept lattice 512 is formed by a table whose rows are different experiments (i.e., different electrical stimulation leads) and whose columns are a set of concepts (patterns produced by the respective local area automatic encoders) and behavioral outcomes.

The inverse mapping technique is used to find a set of stimulation electrode placement locations and parameters that can reconstruct the desired brain state (element 314). In an embodiment, the inverse mapping technique is a self-organizing critical form of search (SOC search) (fig. 10). The discovered stimulation electrode placement locations and parameters are applied to the brain via a brain stimulation subsystem, and behavioral results are monitored. If the results are not satisfactory, the chart is updated using the brain monitoring subsystem and a new row is created in the conceptual table and the process is repeated. If the results do not achieve the desired behavioral effect, the results are determined to be satisfactory. Depending on the desired effect, the determination of success may be subjective or there may be a task performance metric. For example, if the goal is to weaken the selected memory that leads to post-traumatic stress disorder, then the treatment is successful when the subject is in remission. If the goal is to improve certain memories, the ability to recall those memories after a certain time may be a measure of success.

In embodiments, monitoring and stimulation is accomplished by co-located electrodes. In another embodiment, brain stimulation is accomplished by transcranial magnetic stimulation rather than electrical current. In yet another embodiment of the invention, brain stimulation and/or monitoring is accomplished by chronically implanted electrodes. In another embodiment, behavioral outcomes are promoted for a particular subject as opposed to a population of subjects.

The invention described herein will have a significant scientific impact by facilitating findings related to the neural mechanisms underlying complex behavioral and psychological phenomena and their external control by means of invasive or non-invasive electrical current stimulation. Knowledge representation frameworks, datamation, and discovery tools will enhance the understanding of the mapping of non-invasive and/or invasive stimuli to neural mechanisms, which allows for more finely tuned stimulation interventions that achieve behavioral outcomes.

A system according to embodiments of the present disclosure will not only facilitate understanding of brain function, but will also provide optimal intervention for behavioral deficits (e.g., memory impairment due to traumatic brain injury) and psychological disorders (e.g., depression). The present invention will significantly enhance long-term retention of memory in both normal and elderly populations and potentially restore memory function in subjects with neurodegenerative diseases and brain damage.

Knowledge representation according to embodiments of the present disclosure may generate more subtle assumptions of stimulus evoked currents based on queries for desired behavior. In addition, it is found that the tool can then acquire these outputs by virtue of the unique concepts and techniques described herein to better optimize the location and parameters of the stimulation electrodes.

A system according to embodiments of the present disclosure is a process and apparatus for assisting a human subject suffering from behavioral deficits and psychological disorders. It follows that it may be part of a service that helps injured soldiers and civilians. Moreover, the present invention may be part of a product for memory recovery and enhancement that will have a large potential commercial market.

Finally, while the invention has been described in terms of several embodiments, those of ordinary skill in the art will readily recognize that the invention can have other applications in other environments. It should be noted that many embodiments and implementations are possible. Further, the following claims are in no way intended to limit the scope of the invention to the specific embodiments described above. Additionally, any recitation of "means for … …" is intended to evoke a device plus function reading of the elements and claims, and no element specifically using the recitation of "means for … …" is intended to be read as a device plus function element, even if the claims otherwise include the word "means". Further, although specific method steps have been recited in a particular order, the method steps may occur in any desired order and are within the scope of the invention.

Claims

1. A system for determining brain stimulation that elicits a desired behavior, the system comprising:

a brain monitoring subsystem comprising a set of monitoring electrodes for sensing brain activity;

a brain stimulation subsystem comprising a set of stimulation electrodes for applying electrical current stimulation; and

one or more processors and a non-transitory computer-readable medium having executable instructions encoded thereon such that, when executed, the one or more processors perform operations comprising:

registering a set of multi-scale distributed data into a graphical representation, wherein at least a subset of the set of multi-scale distributed data is sensed brain activity;

identifying a sub-chart in the graphical representation;

mapping the sub-graph to a set of conceptual features, generating a conceptual lattice that relates the set of conceptual features to behavioral effects;

determining a current stimulus to be applied to produce the behavioral effect; and

such that the current stimulus is applied via the set of stimulation electrodes.

2. The system of claim 1, wherein the graphical representation comprises a plurality of nodes, each node representing a data item in the set of multi-scale distributed data, and edges between the plurality of nodes representing relationships between data items, wherein the relationships are at least one of topological relationships, statistical relationships, and causal relationships.

3. The system of claim 1, wherein the set of multi-scale distributed data includes electroencephalographic data recorded from the set of monitoring electrodes as a result of stimulation leads applied via the set of stimulation electrodes.

4. The system of claim 1, wherein the one or more processors further perform the following:

transforming the set of multi-scale distributed data into currents in voxels of the brain volume by means of forward simulation; and

integrating the current into the graphical representation.

5. The system according to claim 1, wherein the one or more processors further perform the operation of identifying a set of stimulation electrode placement locations and parameters that can reproduce the behavioral effect.

6. The system of claim 5, wherein the identified set of stimulation electrode placement locations and parameters are applied via the brain stimulation subsystem, and wherein the brain monitoring subsystem monitors the behavioral effects, wherein if the behavioral effects are not satisfactory, the brain monitoring subsystem updates the graphical representation and the one or more processors perform operations to determine a new current stimulation to apply to produce a satisfactory behavioral effect.

7. The system of claim 1, wherein the one or more processors further perform an operation of searching for electrode locations or settings for brain stimulation using a self-organizing threshold SOC.

8. The system of claim 1, wherein the one or more processors further perform operations to predict behavioral outcomes for various brain stimuli using forward simulation.

9. A computer-implemented method of determining brain stimulation that causes a desired behavior, the method comprising the acts of:

causing one or more processors to execute instructions encoded on a non-transitory computer-readable medium such that, when executed, the one or more processors perform the following:

registering a set of multi-scale distributed data into a graphical representation, wherein at least a subset of the set of multi-scale distributed data is sensed brain activity obtained from a brain monitoring subsystem comprising a set of monitoring electrodes;

identifying a sub-chart in the graphical representation;

determining a current stimulation to be applied via a brain stimulation subsystem comprising a set of stimulation electrodes to produce the behavioral effect; and

10. The method of claim 9, wherein the graphical representation comprises a plurality of nodes, each node representing a data item in the set of multi-scale distributed data, and edges between the plurality of nodes representing relationships between data items, wherein the relationships are topological, statistical, and/or causal relationships.

11. The method of claim 9, wherein the one or more processors further perform the following:

integrating the current into the graphical representation.

12. The method of claim 9, wherein the one or more processors further perform the operation of identifying a set of stimulation electrode placement locations and parameters that can reproduce the behavioral effect.

13. The method of claim 12, wherein the identified set of stimulation electrode placement locations and parameters are applied via the brain stimulation subsystem, and wherein the brain monitoring subsystem monitors the behavioral effects, wherein if the behavioral effects are not satisfactory, the brain monitoring subsystem updates the graphical representation and the one or more processors perform operations to determine a new current stimulation to be applied to produce a satisfactory behavioral effect.

14. The method of claim 9, wherein the one or more processors further perform the operation of using the self-organizing critical SOC to search for electrode locations or settings for brain stimulation.

15. The method of claim 9, wherein the one or more processors further perform operations to predict behavioral outcomes for various brain stimuli using forward simulation.

16. A non-transitory computer readable medium determining brain stimulation causing a desired behavior, the non-transitory computer readable medium having stored thereon computer readable instructions executable by a computer having one or more processors for causing the processors to:

identifying a sub-chart in the graphical representation;

17. The non-transitory computer-readable medium of claim 16, wherein the graphical representation includes a plurality of nodes, each node representing a data item in the set of multi-scale distributed data, and edges between the plurality of nodes representing relationships between data items, wherein the relationships are topological, statistical, and/or causal relationships.

18. The non-transitory computer-readable medium of claim 16, further comprising instructions for causing the one or more processors to:

integrating the current into the graphical representation.

19. The non-transitory computer-readable medium of claim 16, further comprising instructions for causing the one or more processors to further perform operations of identifying a set of stimulation electrode placement locations and parameters that can reproduce the behavioral effect.

20. The non-transitory computer-readable medium of claim 19, wherein the identified set of stimulation electrode placement locations and parameters are applied via the brain stimulation subsystem, and wherein the brain monitoring subsystem monitors the behavioral effects, wherein if the behavioral effects are not satisfactory, the brain monitoring subsystem updates the graphical representation and the one or more processors perform operations to determine a new current stimulation to be applied to produce a satisfactory behavioral effect.

21. The non-transitory computer-readable medium of claim 16, further comprising instructions for causing the one or more processors to further perform operations of searching for electrode locations or settings for brain stimulation using a self-organizing critical SOC.

22. The non-transitory computer-readable medium of claim 16, further comprising instructions for causing the one or more processors to further perform operations for predicting behavioral outcomes for various brain stimuli using forward simulation.