EP3174461A1

EP3174461A1 - Measurement of a patient analyte using wavelet transform analysis

Info

Publication number: EP3174461A1
Application number: EP14898849.6A
Authority: EP
Inventors: Feras AL HATIB; John Michael Dobbles
Original assignee: Edwards Lifesciences Corp
Current assignee: Edwards Lifesciences Corp
Priority date: 2014-08-01
Filing date: 2014-08-01
Publication date: 2017-06-07
Also published as: WO2016018432A1; EP3174461A4

Abstract

The measurement of a patient analyte using wavelet transform analysis is disclosed. Embodiments of the present invention can provide precise identification of erroneous analyte sensor signals and validate good signals. A wavelet transform is applied to a sensor signal to produce a time-scale representation of the sensor signal. In some embodiments a Haar wavelet is used for the transform and energy measurements are derived from the time-scale representation. The time-scale representation of the sensor signal can then be compared to a known sensor signal representation. Embodiments like those disclosed herein are useful in blood glucose monitoring systems in which an analyte sensor signal comprises a blood draw portion and a blood clear portion.

Description

MEASUREMENT OF A PATIENT ANALYTE USING

WAVELET TRANSFORM ANALYSIS

BACKGROUND

[0001] Devices for measuring various physiological parameters of a patient have been a standard part of medical care for many years. The vital signs of some patients typically are measured on a substantially continuous basis to enable physicians, nurses, and other healthcare providers to detect sudden changes in a patient's condition. Patient monitors are typically employed to display a variety of physiological patient data to physicians and other healthcare providers. Such patient data facilitates diagnosis of abnormalities or the patient' s current condition.

[0002] In some circumstances, a hospital subject is continuously monitored or

repeatedly tested for changes in some blood analyte levels, such as for diagnosing, monitoring and/or prognosticating a subject's medical status. In some circumstances, blood samples are collected at regular intervals, and sent for laboratory analysis. For example, of importance for health care providers with some patients is accurate information representing the blood glucose levels of the subject, especially in a surgical or intensive care setting. In some circumstances, a bedside analyte monitor can be used to monitor the levels of the analyte and the current numerical value can be displayed. The importance of this value being an accurate reflection of the actual analyte concentration cannot be overstated. Historical levels of an analyte can also be displayed along with various numerical indicators calculated from these historical values. Any inaccuracies in the historical values are reflected in the calculated numerical indicators. SUMMARY

[0003] While modern monitoring systems generally provide accurate measurement of blood analytes, errors can occur. For example, a glucose monitoring system can take readings from a sensor based on periodic draws of small amounts of blood from a patient's vein into a small chamber housing the sensor. Prior to each blood glucose measurement, the system flushes the sensor chamber using a saline solution or a solution with a known glucose concentration, which is used to clear or flush and/or recalibrate the sensor. The saline or calibrant fluid used to flush and/or recalibrate an in-vivo blood analyte sensor may take a long time to move away from the vicinity of the sensor, causing dilution errors, in subsequent readings by the sensor of the patient' s blood.

[0004] In example embodiments of the concepts disclosed herein, a plurality of

analyte sensor signals is received as an electrical waveform from a sensor. Each analyte sensor signal corresponds to a periodic measurement of a patient analyte. A wavelet transform is applied to a sensor signal from among the plurality of sensor signals to produce a time-scale representation of the sensor signal. The time-scale representation of the sensor signal can then be compared to a known sensor signal representation to provide a determination whether the signal is in fact an erroneous analyte sensor signal from among the plurality of analyte sensor signals being received. At least some embodiments of the concepts disclosed can find use in blood glucose monitoring systems in which each analyte sensor signal comprises a blood draw portion and a blood clear portion. [0005] In at least some embodiments, the comparison process and selective identification of an erroneous signal can include measuring at least one energy value corresponding to at least a portion of the time-scale representation of the sensor signal, and comparing the energy value to at least one pre-determined value for the known sensor signal representation. In at least some embodiments, energy values for a plurality of time-scale zones in the time-scale representation of the sensor signal are used for comparison purposes to identify erroneous sensor signals. In some embodiments, the wavelet transform uses a Haar wavelet function, though the process could be adapted to use other types of wavelet functions for the wavelet transform.

[0006] The process described above can enable erroneous analyte sensor signal

readings to be discarded and/or replaced, so that a display device shows numerical values, graphs, and the like, substantially corresponding to only non-erroneous analyte sensor signal readings or subsequent calculated values therefrom. The display device can be part of a computer system, instruction execution platform, or a workstation with appropriate input and output capabilities. Embodiments of the concepts disclosed may also be implemented on a patient monitoring system including a display device and a processor operatively connected to the display device and connected with a memory. The memory may be used to store data and information corresponding to the patient analyte as well as non-transitory computer program code which, when executed, causes the processor to carry out all or a portion of the process of an embodiment of the invention. This hardware along with a sensor interface and any other input and output components form at least some of the means to carry out the various process elements. BRIEF DESCRIPTION OF THE DRAWINGS

[0007] FIG. 1 illustrates a waveform from a glucose sensor that might find use with an embodiment of the concepts disclosed. The waveform includes a plurality of signals corresponding to periodic glucose measurements over time.

[0008] FIG. 2 is a graphical plot of measured glucose levels that can be produced from the waveform illustrated in FIG. 1.

[0009] FIG. 3 is a schematic block diagram of a system that can implement example embodiments of the concepts disclosed.

[0010] FIGs. 4, 5, 6 and 7 are waveforms showing periodic glucose measurement signals over time, where these signals show evidence of being erroneous, and can be detected by example embodiments of the concepts disclosed.

[0011] FIGs. 8 and 9 illustrate time-scale representations of good signals from a

sensor like that used to produce the waveforms illustrated thus far. The time-scale representations show energy level. FIGs. 8 and 9 illustrate measurements at two different glucose levels.

[0012] FIG. 10 illustrates a time-scale representation of an erroneous signal from a sensor during blood the draw phase like that used to produce the waveform illustrated in FIG. 5. The time-scale representations show energy level. The error in the case of

FIG. 10 is most likely caused by a blood obstruction.

[0013] FIGs. 11, 12, 13, 14 and 15 illustrate time-scale representations of erroneous signals from a sensor like that used to produce the waveforms illustrated in FIGs. 1, 4, 5, 6 and 7. The time-scale representations show energy level. The errors in these cases are all dilution errors.

[0014] FIG. 16 is a flowchart illustrating the process of collecting sensor readings, identifying and discarding erroneous readings, and displaying data without the erroneous measurements, as might be carried out by the system of FIG. 3

implementing an example embodiment according to the concepts disclosed herein.

[0015] FIGs. 17, 18, 19 and 20 are graphical plots of measured glucose levels that show erroneous measurements that are or have been discarded using an embodiment according to the concepts disclosed herein.

DETAILED DESCRIPTION OF EXAMPLE EMBODIMENTS

[0016] Exemplary embodiments and/or aspects of a system such as those presently disclosed provide a precise way to identify erroneous measurements of a blood analyte resulting from dilution or other adverse occurrences.

[0017] The following detailed description teaches specific example embodiments of the concepts disclosed. Other embodiments do not depart from the scope of the present invention. The terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. As used herein, the singular forms "a", "an" and "the" are intended to include the plural forms as well, unless the context clearly indicates otherwise. It will be further understood that the terms "includes" and/or "including" when used herein, specify the presence of stated features, integers, steps, operations, elements, and/or components, but do not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and/or groups thereof.

[0018] Unless otherwise defined, all terms (including technical and scientific terms) used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs. It will be further understood that terms used herein should be interpreted as having a meaning that is consistent with their meaning in the context of this specification and the relevant art and will not be interpreted in an idealized or overly formal sense unless expressly so defined herein. Unless otherwise expressly stated, comparative, quantitative terms such as "less" and "greater", are intended to encompass the concept of equality. As an example, "less" can mean not only "less" in the strictest mathematical sense, but also, "less than or equal to."

[0019] As will be appreciated by one of skill in the art, embodiments of the concepts disclosed may include a method, device, article, system, computer program product, or a combination of the foregoing. Any suitable computer usable or computer readable medium may be used for a computer program product including non- transitory computer program code to implement all or part of an embodiment of the concepts disclosed. The computer usable or computer readable medium may be, for example but not limited to, a tangible electronic, magnetic, optical, electromagnetic, or semiconductor system, apparatus or device. More specific examples (a non- exhaustive list) of the computer readable medium would include the following: a portable computer diskette, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or Flash memory), a portable compact disc read-only memory (CD-ROM), or an optical storage device. The computer usable or computer readable medium may be one or more fixed disk drives or flash drives deployed in instruction execution platforms, such as servers or workstations, forming a "cloud" or network.

[0020] Computer program code for carrying out operations of an embodiment of the of the concepts disclosed herein or for assisting in the carrying out of a method according to an example embodiment may be written in an object oriented, scripted or unscripted programming language such as Java, Perl, Smalltalk, Matlab, C/C++ or the like. However, the computer program code for carrying out the concepts disclosed may also be written in conventional procedural programming languages, such as the "C" programming language or similar programming languages.

[0021] Computer program instructions may be provided to a processor of an

instruction execution platform such as a general purpose computer, special purpose computer, server, workstation or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions/acts necessary to carry out an embodiment of the concepts disclosed.

[0022] A processor used to implement an exemplary embodiment according to this disclosure may be a general purpose digital signal processor, such as those commercially available from Texas Instruments, Inc., Analog Devices, Inc., or Freescale Semiconductor, Inc. It may also be a general purpose processor such as those typically provided for either workstation or embedded use by companies such as Advanced Micro Devices, Inc. or Intel Corporation. It could as well be a field programmable gate array (FPGA) as are available from Xilinx, Inc., Altera

Corporation, or other vendors. The processor could also be a fully custom gate array or application specific integrated circuit (ASIC). Any combination of such processing elements may also be referred to as a processor, microprocessor, controller, or central processing unit (CPU). In some embodiments, firmware, software, or microcode can be stored in a non-transitory form on or in a tangible medium that is associated with the processor. Such a medium may be a memory integrated into the processor, or may be a memory chip that is addressed by the processor to perform various functions. Such firmware, software or microcode is executable by the processor and when executed, causes the processor to perform its control and calculation functions. Such firmware or software could also be stored in or on a tangible medium such as an optical disk or traditional removable or fixed magnetic medium such as a disk drive used to load the firmware or software into a monitoring system according to embodiments of the concepts disclosed herein.

3] The term "analyte" as used herein relates to a substance or chemical constituent in a biological sample (e.g., bodily fluids, including, blood, serum, plasma, interstitial fluid, cerebral spinal fluid, lymph fluid, ocular fluid, saliva, oral fluid, urine, excretions, or exudates. Analytes can include naturally occurring substances, artificial substances, metabolites, and/or reaction products. The analyte for measurement by the sensor, devices, and methods is inclusive of glucose. Any other physiological analyte or metabolite can be substituted or combined with the measurement of glucose. The term "subject" as used herein relates to mammals, inclusive of warm-blooded animals (domesticated and non-domesticated animals), and humans.

[0024] The term "calibration" as used herein refers to one or more process of

determining the relationship between sensor data and a corresponding reference data. A continuous analyte sensor can be initially calibrated, calibration can be updated or recalibrated over time (whether or not if changes in the relationship between the sensor data and reference data occur), for example, due to changes in

disconnection/reconnection, sensitivity, baseline, analyte transport, metabolism, and the like. The sensed values produced by a calibrated sensor can be referred to as "calibrated values."

[0025] The phrases "operatively connected" and "operably connected" as used herein relate to one or more components linked to one or more other components, such that a function is enabled. The terms can refer to a mechanical connection, an electrical connection, or any connection that allows transmission of signals between the components. For example, one or more electrodes can be used to detect the amount of analyte in a sample and to convert that information into a signal; the signal can then be transmitted to a circuit. In such an example, the electrode is "operably connected" to the electronic circuitry. The terms include wired and wireless connections, and situations where there is are or may be intervening components.

[0026] The term "sensor" as used herein relates to a device, component, or region of a device capable of detecting and/or quantifying and/or qualifying an analyte in the intravascular and/or subcutaneous space of a subject. The phrase "sensor system" as used herein relates to a device, or combination of devices operating at least in part in a cooperative manner, that is inclusive of the sensor. In some aspects, sensor relates to a device, component, or region of a device capable of detecting and/or quantifying and/or qualifying an analyte in the intravascular and/or subcutaneous space in-vivo. The sensor can be of any design that provides, generates, or corresponds to an individual signal and/or a waveform- like signal, such as an electrochemically or optically based sensor.

[0027] The term "signal" and its variants are used herein at times in conjunction with the term "waveform." Generically, these two terms can be used interchangeably. For purposes of this disclosure however, a waveform is meant to refer a continuous sensor reading or its graphical representation. Such a construct would often include quiescent periods where there is essentially no signal from a sensor. A signal, by contrast, will usually be meant to designate the substantially non-zero portion of the waveform that represents a sensor calibration and/or reading. A signal is typically detected and processed to determine an analyte level at a given point in time.

[0028] Reference is also made herein to data or information corresponding to,

representative of, or which is a "representation" of a sensor signal or a measured value. Such terminology is meant to indicate that the data or information could be a direct reproduction or representation of a signal, its corresponding analyte level, or could be a representation of a calculated, transformed, or derived value or collection of values. For example, a time-scale plot generated by a continuous wavelet transform of a sensor signal might be said to be representative of or corresponding to the sensor signal or an analyte reading that the sensor signal in turn indicates. A statistical indication of historical sensor signals or their representation can be said to be a representation of the collection of signals in a waveform or the analyte value over time. It should also be noted that the identification of erroneous signals is necessarily the validation of the non-erroneous signals in a waveform. These two concepts can be considered interchangeable in terms of the function of the system described herein.

[0029] The term "clear" when used in reference to methods disclosed herein is

synonymous with "flushing" or "flush" and these terms may be used interchangeably herein.

[0030] Glucose control in some hospital patients, such as those in the ICU, has been proven difficult with an increased risk for hypoglycemic episodes. Accurate and reliable in-vivo continuous blood glucose measurement is needed, and a variety of methods for continuous in-vivo measurement of blood glucose using electrochemical sensors have been developed. These devices can be relatively complex, and are subject to a number of inaccuracies due to the devices' sensitivities to the complicated biochemical environment in which in-vivo measurements must be made.

[0031] To overcome the technical issues outlined above, a method of measuring

blood glucose concentration with periodic and continuous recalibration of the electrochemical sensor using a saline solution with a known glucose concentration can be used. The system uses periodic (continuous or non-continuous) draws of small amounts of blood from a patient's vein into contact with the glucose sensor. Prior to each blood glucose measurement, the system flushes the sensor using a saline solution with a known glucose concentration, which may also be used to continuously or non- continuously recalibrate the sensor. After recalibration, the system draws a small amount of blood from the patient's vein into contact with the glucose sensor to perform the next blood glucose concentration measurement. The measurement is performed using the latest calibrated sensor value. After taking the measurement, the system clears the blood back into the patient and performs another

flushing/recalibration process. This process of blood clear and blood draw is repeated periodically. The sensor can be integral with a vascular accessing device configured for introduction to the circulatory system of a subject, such as a catheter, or, the sensor can be positioned external to the circulatory system of the subject, e.g., in a housing fluidically coupled to the vascular accessing device. A description of such a system can be found in U.S. Patent Application Publication No. 2011/0319728, the entire disclosure of which is incorporated herein by reference.

[0032] The introduction of a blood sample or calibrant flush solution to the sensor results in a signal from the sensor that is configured to be received by the system in accordance with the methods herein disclosed. The signal from the sensor can be a current or voltage, which can be directly or indirectly related to the analyte sensed by the sensor. For brevity, the use of glucose as the analyte is exemplary recited hereinafter with the understanding that the methods and systems herein disclosed are applicable to other analytes, alone or in combination, without limitation.

[0033] The process above as used with embodiments of the concepts disclosed herein results in a sensed glucose waveform with the pattern illustrated in FIG. 1. FIG. 1 shows a chart 100 of sensor current counts vs. time in hours. Each signal that corresponds to a sample includes a blood draw portion 102 and a blood clear portion 104. This periodic sensor recalibration provides accurate blood glucose measurement, as it compensates for the impact of the complex biochemical environment on the electrochemical properties of the glucose sensor. However, obstructions can occur. Also, since the blood glucose measurement is performed in a vein (often a peripheral vein), when the system flushes saline over the sensor in order to calibrate it, venous blood flow may take a long time to carry away the saline from the vicinity of the sensor. Thus, the saline may cause dilution of the blood in the proximity of the sensor via mixing or diffusion, which can introduce error in a blood glucose measurement.

[0034] An example of blood glucose measurement errors caused by dilution is shown in FIG. 2. Graph 200 includes plotted points of measured blood glucose values over time. The vast majority of sample points such as point 202 represent normal measurements from good sensor signals. However, the dilution effect discussed above results in measurements such as those at points 204 and 206, which are erroneous. The dilution effect could be further intensified in physiological conditions that cause low peripheral flow or in conditions of vein collapse around the catheter carrying the sensor, or where there is a change in position or motion by the subject. Such errors can especially cause problems in the ICU, where patients experience extreme cardiovascular physiology and where glucose measurements are particularly important. Embodiments of the present invention can detect these erroneous measurements. Moreover, embodiments of the concepts disclosed can detect erroneous measurements that are less apparent statistical outliers than those indicated by points 204 and 206, and thus are more difficult to detect with simple statistical analysis.

[0035] Currently, erroneous readings, such as readings 204 and 206 shown in FIG. 2 can be detected with a variety of signal processing methods using the raw sensor signal of FIG. 1. Such methods can include time-domain techniques such as measuring and comparing slopes, time intervals, derivatives, amplitudes, etc. of the blood draw or blood clear portions of the raw sensor signal of FIG. 1. Other methods could include signal interpolation, filtering, or parametric/nonparametric recursive and nonrecursive prediction techniques to measure signal trends in the different phases. Current error detection methods could also include frequency-domain methods such as fast Fourier transform (FFT) based approaches to detect the presence or absence of different frequency harmonics.

6] There are potential shortcomings with such existing time or frequency domain techniques. Firstly, the sensor signal during dilution or occlusion conditions could exhibit a very wide variety of signal features. Thus, in order to detect erroneous conditions, separate signal processing techniques would need to be put in place to detect each particular signal feature. Since each erroneous signal feature needs to be detected with its own signal processing technique, it is not practically possible to cover all the different variations of the signal exhibited during erroneous conditions with existing time and frequency domain techniques. Secondly, with many erroneous conditions, the sensor signal looks very similar to the sensor signal during normal conditions, with very similar time and frequency domain characteristics. Thus, existing time or frequency domain methods may be unreliable for detection of erroneous conditions. These issues can lead to very poor sensitivity and specificity of identification of erroneous signals with standard time or frequency domain methods, leading to unacceptable numbers of both false positives and false negatives. [0037] FIG. 3 depicts an operating environment and a system that incorporates aspects of embodiments of the concepts disclosed in this document. In this example, an analyte information display system 10 is configured to monitor blood glucose levels in a subject. The subject in this case is a hospitalized medical patient. The system is configured around a monitor and control unit 12, which comprises a programmed processor to control the functioning of the system, and a display device including a visible panel configured to communicate information regarding the system and its functions, as well as the condition of the patient, to a user of the system. The display panel may also serve as a touch screen interface through which a user can enter information and commands for controlling the system's operations. Various other configurations of functionally equivalent components can be employed.

[0038] In FIG. 3, an access point 14 provides access for the system to the patient's body. The access point may provide an entry site for a catheter placed in the vein or another vessel in the vasculature of the patient, access to interstitial fluid located under the patient's skin, or to any other site through which access can be provided to fluids or materials bearing glucose or otherwise providing a reliable indicator of blood glucose levels or corresponding information relevant to the subject. FIG. 3 illustrates a configuration in which an intravascular catheter is disposed inside a vein of the patient, and through which blood can be drawn over an electronic sensor to measure glucose levels in the patient's blood. The sensor in this configuration can be a glucose oxidase sensor configured to produce a current or voltage proportional to the patient's blood glucose level. It cannot be overemphasized that this arrangement is an example. Other mechanisms can be used to monitor blood glucose, and a system according to embodiments of the concepts disclosed can be devised to measure other analytes. Sensor systems that can be used with at least some embodiments of the concepts disclosed are described in U.S. Patent Publication No. 2007/0027385, the entire disclosure of which is incorporated herein by reference.

[0039] Still referring to FIG. 3, the system further includes a supply of infusion fluid 16, housed inside an infusion vessel 32, and in fluid communication with the sensor and the access point 14 through a fluid line 22 between the infusion bag and the access point. A flow control device 20 controls the flow of fluids back and forth through the fluid line between the access point and the fluid bag. In this particular example, the flow control device is a pump, but other devices can be used. Under control of the flow control device, blood may be drawn from the patient's vein over at least a portion of the sensor. At other times, infusion fluid may be used in the sensor system by being directed from the bag to flow over and rinse the sensor. The infusion fluid may also at times be infused into the patient. The infusion fluid may comprise normal medical saline solution or other physiologically acceptable solutions. The infusion fluid may further comprise glucose at a known concentration, which may be directed over the sensor from time-to-time in order to calibrate the sensor as previously described by reading the resultant current or voltage from the sensor at times when the known-concentration glucose solution in the infusion bag is being directed over and in contact with the sensor.

[0040] The system 10 of FIG. 3 further includes an insulin supply controller 18

configured to supply insulin in a controlled fashion through an insulin supply line 19 to the fluid line 22 and from there into the body of the patient. The elements of system 10 can be mounted on and supported by a stand 34, so that the system can be placed with or moved with the patient as needed. It should be noted that the term "system" may be used herein to refer to the entire arrangement of electronic elements and connection cables described above. However, the term system may also be used to refer only to the monitor and control unit including installed software and/or firmware that together direct and execute the functions described herein.

[0041] Continuing to refer to FIG. 3, a signal line 36 provides electrical

communication between the sensor near the access point 14 and the monitor and control unit 12. Electrical communication is similarly provided between the monitor and control unit and the fluid pump 20 through a data and control cable (not shown), and similarly between the monitor and control unit and optionally an insulin supply controller 18 through another data and control cable. The system 10 pictured in FIG. 3 can be configured as a "closed- loop" system capable of both monitoring the patient's blood glucose level and control of that level by administering one or more doses of glucose or drugs such as insulin, all as determined and controlled by software and circuitry of the monitor and control unit 12. The visual display and touch-screen control of the monitor and control unit communicate information to a user of the system. The user may further input commands to the system through that same visual display and touch-screen control, for example, to control the amount and/or rate of administration of the drug, to terminate administration of the drug and/or to provide a bolus of the drug.

[0042] It should be noted that data analysis features and functions described herein can also be used in a so-called "open-loop" system to great advantage. An open loop system is one where an appropriate medication or other physiological component, in this exemplary case insulin or glucose, is administered manually by medical personnel in response to indications presented by the display panel of a monitor and control unit like the monitor and control unit 12. Hybrid systems also exist and can incorporate embodiments of the concepts disclosed. A hybrid system combines various aspects of open and closed- loop systems. For example, a hybrid system may determine and recommend a dose of a medication, but rely on medical personnel to administer the medication, possibly through confirmation of the dosage via user input.

3] Still referring to FIG. 3, the system includes I/O interface 302, which may in turn include an appropriate connector, and circuitry to monitor signals from the sensor system. This circuitry may include analog-to-digital converters, encoders, decoders, and the like. I/O interface 302 is coupled to a central processing unit (CPU) 304, which controls the operation of the entire system and may be referred to herein merely as a "processor." The I/O interface receives sensor signals and may also send signals to control the supply of insulin to the subject with some embodiments of the concepts disclosed. CPU 304 is further operatively connected to memory 306. Memory 306 stores all of the information needed for the system to operate. Such information may be stored in a temporary fashion, or may be stored more permanently. This memory may include a single, or multiple types of memory. For example, a portion of the memory connected with CPU 304 may be "flash" memory which stores information semi-permanently for use by the system. In either event memory 306 of FIG. 3 in this example embodiment includes computer program code 308 which, when executed by CPU 304, causes the system to carry out the various processes according to example embodiments of the concepts disclosed. Memory 306 also stores data 310, which in example embodiments includes historical numerical values for the analyte being measured.

[0044] Continuing with FIG. 3, monitoring and control unit 12 may also include a network interface 313. This network interface can allow the system to be connected to a wired or wireless network to allow monitoring on a remote display (not shown). For example, the remote display could duplicate, or be used in place of the local display panel. Network interface 313 could also be simply used to trigger an alarm at a nurse's station or on a mobile device. In the embodiment of FIG. 3, a local display device, 317, is connected with CPU 304 via a graphics engine 324. The local display device may be an LCD panel, plasma panel, or any other type of display component and accompanying circuitry to interface the display device to graphics engine 324. Graphics engine 324 may be on its own chip, or in some embodiments it may be on the same chip as CPU 304. Note that display device 317 may include user input functionality, for example an optical or capacitive touchscreen over the display screen. In such a case, monitoring control unit 12 may include additional circuitry to process such input. Alternatively, such circuitry may be included in the display device itself, the graphics engine, or the CPU 304.

[0045] As previously discussed, when blood glucose concentration is diluted by saline in the proximity of the analyte sensor, the sensor signal during a blood draw becomes noisy and distorted, as it represents the saline glucose in addition to the blood glucose. Because of discontinuities introduced by the mechanical draw and clear, and because of physiological complexities of the dilution, this noise is non- stationary and varying. FIGs. 4, 5, 6 and 7 show some illustrative examples of a glucose sensor signal that shows the effect of dilution. FIG. 4 shows waveform graph 400, with signals 402 and 404 evidencing dilution errors. FIG. 5 shows waveform graph 500, with signals 502 and 504 evidencing dilution errors. FIG. 6 shows waveform graph 600, with a dilution error evidenced in signal 602 and FIG. 7 illustrates waveform graph 700 with dilution error evidenced in signal 702. As can be readily observed in this figures, the effects of dilution can be very diverse and in some cases, can produce a signal that is very similar to the sensor signal in normal conditions, meaning an erroneous measurement in such a circumstance is difficult to detect with standard statistical techniques and/or techniques that are carried out only in the time domain or the frequency domain.

6] Embodiments of the concepts disclosed herein detect erroneous analyte measurement signals using a time-frequency approach, based on a wavelet transform. This transform may be continuous or discrete and take various forms, however for the purposes of the detailed examples herein it will be referred to as a continuous wavelet transform (CWT). This wavelet-based process enables detailed local analysis of an analyte sensor signal to detect any unusual signal features. In the case of blood analyte monitoring as described above, the CWT approach also detects the time when those unusual signal features occur among the blood draw, blood clear and calibration time-segments of the signal. Thus, this wavelet-based process is capable of revealing aspects of an analyte sensor signal that other traditional signal analysis techniques may miss such as trends, breakdown points, discontinuities, and self-similarity.

Furthermore, because it presents a different view of the signal than that presented by classical techniques, the wavelet-based process can be further used to de-noise the signal without any appreciable degradation.

[0047] Example embodiments of the concepts disclosed apply a continuous wavelet transform (CWT) to a raw analyte sensor signal. For purposes of the examples in this disclosure, the sensor is a glucose sensor as previously discussed. The CWT uses inner products to measure the similarity between the sensor signal and an analyzing function ^(wavelet). The CWT in effect compares the glucose signal to time-shifted and frequency altered (compressed or stretched) versions of the wavelet function Ψ. The stretching or compressing the wavelet function corresponds to the physical notion of frequency or scale. Scale can be thought of as the amount of space taken up by a given section of the wavelet.

[0048] One can appreciate the relationship with frequency if one imagines a wave being stretched or compressed. With such a change, the peak-to-peak distance or wavelength of the wave is changed. Assuming such a wave always moves at the same speed, the number of peaks passing a given point, or the frequency, changes. Shifting the wavelet function corresponds to changing its position in time. By comparing the signal to the wavelet at various scales and positions, one can obtain a function of three variables representing these quantities over various shifted positions of the wavelet. Thus, the CWT generates coefficients that can be used as a third dimension in a representation of the signal. The signal typically consists of electric current (but could also be voltage) and the coefficients represent the energy in the signal for a given time and level. For scale parameter a > 0, and position b, the CWT of a glucose sensor signal in an exemplary embodiment can be represented as: C(a, b; g(t) t)) = C_∞g(t j= * ±) dt, where

C is the CWT coefficients;

g(t) is the glucose sensor signal;

Ψ(ι) is the wavelet function;

a is the scale parameter; and

b is the position parameter (time 9] The equation above mathematically demonstrates that not only do the values of scale and position affect the CWT coefficients, but the choice of wavelet also affects the values of the coefficients. Since, a blood analyte sensor signal of the type illustrated thus far typically contains marked pulses introduced by the discontinuities of the mechanical blood draw and clear, a Haar wavelet, similar to a step function or square wave, can be used. Other wavelet functions could also be used and may produce better results with different types of sensor signals. For example, some Mathieu wavelets might be used in some circumstances. The Haar wavelet is defined mathematically as follows:

By applying the Haar wavelet to the signal and by continuously varying the values of the scale parameter, a, and the position parameter, b, one can obtain the CWT coefficients C(a,b). The plot of the CWT coefficients versus the scale and position gives a time-scale representation of the glucose sensor signal that includes a "z" dimension. Such a representation includes scale plotted against time, but also includes energy level for a given scale and time as represented by the coefficient or z value. In example embodiments of the concepts disclosed, the CWT is

programmatically applied to the portion of the sensor signal between blood draw and blood clear to detect any distortion, such as distortion that may be caused by the dilution effect.

[0050] Two examples of acceptable blood glucose measurements and their time-scale representations are shown in FIGs. 8 and 9. FIG. 8 illustrates a blood glucose sensor signal plot 800 and its time-scale representation 802. The z-dimension in the time- scale representation can be visualized as running in and out of, or perpendicular to the page. Portion 804 of the plot is the blood draw portion and portion 806 is the phase just before initiating blood clear. Scale 808 is a key for the coefficient or energy value indicated by the z dimension across the time-scale representation 802. This scale represents energy level. The energy profile in the blood draw portion of the transform is shown at 810 and energy spike 812 is an edge effect, which needs to be discarded from the analysis.

[0051] FIG. 9 illustrates a blood glucose sensor signal plot 900 and its time-scale representation 902. Portion 904 of the plot is the blood draw portion and portion 906 is the phase just before initiating blood clear. Scale 908 is the key for the added dimension of the time-scale representation 902. This scale represents energy level. The energy profile in the blood draw portion of the transform is again shown at 910 and an edge effect is shown at 912. FIG. 8 is for a glucose reading of 200 mg/dl and FIG. 9 is for a glucose reading of 100 mg/dl. There is more noise evident in the latter signal because of its lower signal level. However, use of the CWT results in time- scale representations for acceptable signal readings that are nearly identical, despite a significant difference between the levels of the readings.

[0052] It should be emphasized that higher scales in the time-scale representations correspond to the most "stretched" wavelets. The more stretched the wavelet, the longer the portion of the signal with which it is being compared, and therefore the coarser the signal features measured by the wavelet coefficients. Thus, in the time- scale representation of an analyte sensor signal, higher scales correspond to lower frequencies and lower scales correspond to higher frequencies.

[0053] Compared to the time-scale representations shown in FIG. 8 and FIG. 9, distortion caused by dilution as well as other anomalies such as obstructions will generate a very different time-scale representation of the glucose sensor signals. FIG. 10 illustrates the time-scale representation of a signal that is likely at least in part erroneous because of an obstruction, which can in turn cause dilution. The signal graph 1000 is shown above its time-scale representation 1002 as before. Note that there is almost no energy anywhere in the time-scale domain other than that caused by the edge effect.

[0054] FIGs. 11-15 show examples of different glucose sensor waveforms measured during dilution episodes and their corresponding CWT-generated time-scale representations. The transformation of a specific signal from each waveform is indicated in each case by an arrow. In FIG. 11, time-scale representation 1102 results from signal 1104, which is part of the waveform shown in plot 1106. In FIG. 12, time-scale representation 1202 results from signal 1204, which is shown in plot 1206. In FIG. 13, time-scale representation 1302 results from signal 1304, which is illustrated as part of waveform plot 1306.

[0055] FIG. 14 illustrates that very little energy is present in the bulk of the time-scale representation of the signal where there is significant dilution, e.g. erroneous data. In FIG. 14, time-scale representation 1402, showing little energy outside of the edge effect, results from the transformation of signal 1404, which is part of the waveform illustrated in plot 1406. In FIG. 15, the time-scale representation 1502 is very similar to that of a normal signal because signal 1504 from waveform plot 1506 only shows evidence of slight dilution. This signal might be treated as normal using traditional statistical techniques, but the CWT-based technique, example of which are disclosed herein, can identity this signal as erroneous.

[0056] FIG. 16 shows a flowchart illustrating the process carried out by a system like that shown in FIG. 3, which is implementing an exemplary embodiment of concepts disclosed herein. Like most flowcharts, FIG. 16 illustrates process 1600 as a series of sub-process blocks. Process 1600 begins at block 1602. At block 1603, the wavelet function to be used, for example, the HAAR wavelet as previously discussed, is initialized or reconstructed. At block 1604, the processor in the system obtains the next sensor reading. This sensor reading consists essentially of an analyte sensor signal from among a waveform with a plurality of analyte sensor signals. At block 1606, the continuous wavelet transform (CWT) is applied to compare the sensor signal to a plurality of shifted versions of the wavelet. The CWT is applied by using inner products to measure the similarity between a sensor signal and the Haar wavelet, using the following equation: 1 (t - b\

g(t)— V — - V a / where:

C is the CWT coefficients;

g(t) is the glucose sensor signal;

a is the scale parameter; and

b is the position parameter (time shift).

Ψ(ι) is the Haar wavelet, defined as:

7] The theoretical analog equations of the CWT illustrated above can be practically implemented in the digital domain using the following algorithm:

1. Determine the coefficients of the scaling filter corresponding to the wavelet. The scaling filter coefficients for the Haar wavelet are:

W = [0.5, 0.5]

2. Compute the coefficients of the orthogonal reconstruction wavelet filters based on the scaling filter coefficients. The following equations are used:

Low -pass reconstruction wavelet filter:

Lo_R = V2 x W

Lo_R = [0.7071, 0.7071]

High-pass reconstruction wavelet filter:

The coefficients of the high-pass reconstruction wavelet filter are computed by reversing the order of the coefficients of the low-pass reconstruction filter and then reverse the sign of the even coefficients. For this case, the high-pass filter coefficients are simply:

Hi_R = [0.7071, -0.7071]

3. Reconstruct the wavelet function using the low-pass and high-pass reconstruction filters. In this example embodiment of the invention, dyadic interpolation (upsampling and convolution) of a vector of incremental numbers X=[l,2,3...] and the reconstruction filter coefficients LOR and HIR are used. The length of the interpolation in this example is 1024 samples. A standard library function can be used to perform the dyadic interpolation. The reconstruction generates a 1024- sample approximation function of the Haar wavelet.

4. Integrate the reconstructed Haar wavelet function over the 1024 samples. This is done using cumulative sum of all the 1024 samples of the reconstructed Haar wavelet function. Note that the integrated function of the Haar wavelet will have a triangular shape.

5. Compute the wavelet coefficients. For this example implementation, 128 scales are used: from scale 1 through scale 128 with a step of 1.

To compute the wavelet coefficients a one-dimensional convolution of the sensor signal and the integrated function of the reconstructed Haar wavelet for each scale are used. For this example implementation, the blood draw portion of the signal is used: from the start of blood draw to just before the start of the blood clear phase. This results in the entire sensor signal used in the convolution being 714 samples. The resulting coefficient matrix is a 128x714 matrix representing the coefficients for each scale and for each sensor signal's samples. If the wavelet coefficients are plotted on a two-dimensional plot: scale vs. time-samples, one obtains the scale-time plots shown in FIG. 8 through FIG. 15.

[0058] At block 1610, the system determines coefficients for scale and position to produce a time-scale representation of the sensor signal that was discussed above.

[0059] Still referring to FIG. 16, at block 1612, a plurality of timescale zones in the timescale representation of the sensor signal are identified. In the example embodiments, two timescale zones are identified. Zone 1 that is referred to herein as

"early time-high scales" is defined as:

Time: samples 10 to 220

Scale: 108 to 128

[0060] Zone 1 corresponds to the initial time during the blood draw and the low- frequency components of the sensor signal (high scales). Since during the blood draw, the sensor signal is characterized by either a sharp increase or decrease, the CWT coefficients using the Haar wavelet must exhibit very high values in that particular timescale zone. This can be observed easily in FIG. 8 and FIG. 9. Any catheter obstruction or blood dilution will cause smoothing of the sensor signal during blood draw. This smoothing causes a significant decrease in the wavelet coefficients values. The decrease is easily observable in FIG. 10 and FIG. 14, where the timescale energy, as defined by the wavelet coefficients, in zone 1 is practically missing.

[0061] Zone 2 that is referred to herein as the "late time-all scales" zone is defined as:

Time: samples 250 to 600

Scale: 1 to 128 [0062] Zone 2 corresponds to the time after the initial blood draw and all the frequency components of the signal (all scales). Since during this time, the sensor signal reaches its current blood glucose value and therefore remains constant, the CWT coefficients using the Haar wavelet must exhibit very low values. This situation is also observable in FIG. 8 and FIG. 9. Any signal disturbances, such as those caused by noise, motion artifacts, dilution, etc. will generate variations in the sensor signal during that time. The wavelet coefficients will therefore exhibit high values during that time, when the sensor signal is noisy. This situation is observable in FIGs. 12, 13 and 14, where there is significant energy in zone 2 as defined by the wavelet coefficients.

[0063] At block 1614, the energy and each time-scale zone is measured to determine early and late time energies. As previously discussed, energy can be determined by calculating the volume under the surface of the time-scale representation, as follows. The energy for early time-high scale (zone 1) is: where ai represent the scales and b_j represent the time samples

N=128

n=108

L=220

1=10 The energy for early time-high scale (zone where ai represent the scales and b_j represent the time samples

N=128

n=l

L=600

1=250

4] At block 1616 of FIG. 16, the processor in the system compares the time-scale representation of the sensor signal to a known sensor signal representation. In example embodiments, this comparison is based on energy measurements of the two signals. In at least some example embodiments, the comparison is performed as follows. In zone 1, the energy should be high if the sensor signal is good and should be low if the sensor signal is bad. In order to meet the "good signal" requirement, the Energy in zone 1 must exceed the following threshold:

E_zonA > 25000

In zone 2 the energy should be low if the sensor signal is good and should be high if the sensor signal is bad. In order to meet the "good signal" requirement, the energy in zone 2 must not exceed the following threshold:

E Zone ₉2 < 9 10⁴ The thresholds for the Energy in zone 1 and zone 2 where determined experimentally, using a large amount of clinical measurements on ICU patients.

[0065] At block 1618, if the comparison determines that the sample is an erroneous measurement, the sample can be discarded at block 1620. Otherwise the signal can be used to provide a measured numerical value that can be stored and/or graphically displayed on a display device block 1622 in accordance with the system's settings and configuration. At block 1624, a determination is made as to whether the system is still active, and there are more samples in a waveform to be analyzed in accordance with embodiments of the concepts disclosed. If so, processing branches back to block 1604. Otherwise, process 1600 and at block 1626.

[0066] Note that it is possible with an embodiment of the concepts disclosed herein to use the CWT process to analyze at least some erroneous measurements and determine what contribution to the measurement the actual analyte level has made. In such a case, the discarded measurement can be replaced with a good measurement corresponding to the same signal. The good measurement can be determined using curve fitting and extrapolation techniques. It could be also discerned using different filtering techniques, such as recursive and nonrecursive filters, Kalman filters, or different statistical filters, such as probability filters, Bayesian filters, adaptive filters, etc. In FIG. 16, if the erroneous signal were replaced with a technique like one of those mentioned above in block 1620, processing would branch to block 1622 so that the corrected sample could be stored and/or displayed. This alternative is illustrated by use of the dashed arrow in FIG. 16. [0067] FIGs. 17, 18, 19 and 20 illustrate historical plots, 1700, 1800, 1900 and 2000 of blood glucose readings over time for patients in a hospital setting. Measurements that have been identified as erroneous by use of an embodiment like those disclosed herein are indicated as bull's-eyes such as that shown at point 1702 of FIG. 17 and good measurements are shown as plain dots such as point 1704 of FIG. 17. It should be noted that some erroneous data points, such point 1706 of FIG. 17, may be close enough to other good measurements that at least some traditional statistical techniques would not identify the measurement as an erroneous measurement.

[0068] It should be noted that the CWT technique as shown in the examples describe herein can be combined with more traditional statistical techniques to analyze analyte measurement data. In the examples of FIGs. 17-20, once the erroneous points are removed, the remaining data points can be further analyzed to project trends or for any other purpose desired. In such a case, the results of such statistical analysis will have higher accuracy since the starting data will consist essentially of only good data points. For example, a smooth curve-fitting operation with an extrapolation against data like that shown in FIGs. 17-20, but with only the good data points, can produce a prediction of future analyte levels, again with greater accuracy than may be possible without employing an embodiment of the concepts disclosed because the starting data is more accurate. Various statistical techniques can be used to calculate extrapolated numerical values of the analyte. Examples include least square estimation, linear regression, and geometric regression.

[0069] With an embodiment of the concepts disclosed, deviations from acceptable limits can be frequently, accurately recorded and reported as a number of events, with corresponding blood glucose values greater than or less than some acceptable range. More accurate blood glucose monitoring allows calculations of numerical values representing blood glucose values above or below acceptable limits. By combining the severity of out-of-permissible-range events with durations of occurrence of such deviations, physicians and other caregivers can assess the severity of patients' conditions, and devise optimal therapies for controlling and counteracting those conditions. The enhanced measurement accuracy brought about through use of an embodiment of the concepts disclosed can also improve dosing in closed-loop and hybrid systems.

[0070] References cited herein, including but not limited to published and

unpublished applications, patents, and literature references, are incorporated herein by reference in their entirety and are hereby made a part of this specification. To the extent publications and patents or patent applications incorporated by reference contradict the disclosure contained in the specification; the present specification supersedes and/or takes precedence over any such contradictory material of the incorporated reference.

[0071] All numbers expressing quantities of ingredients, reaction conditions, and so forth used in the specification are to be understood as being modified in all instances by the term "about." Accordingly, unless indicated to the contrary, the numerical parameters set forth herein are approximations that may vary depending upon the desired properties sought to be obtained. At the very least, and not as an attempt to limit the application of the doctrine of equivalents to the scope of any claims in any application claiming priority to the present application, each numerical parameter should be construed in light of the number of significant digits and ordinary rounding approaches.

2] Although specific embodiments have been illustrated and described herein, those of ordinary skill in the art appreciate that any arrangement which is calculated to achieve the same purpose may be substituted for the specific embodiments shown and that the inventive concept has other applications in other environments. This application is intended to cover any adaptations or variations of the present invention. The following claims are in no way intended to limit the scope of the invention to the specific embodiments described herein.

Claims

WHAT IS CLAIMED IS:

1. A method of processing signals from a patient analyte sensor, the method comprising:

obtaining a plurality of analyte sensor signals, each analyte sensor signal corresponding to a periodic measurement of a patient analyte;

applying by a processor, a wavelet transform to a sensor signal to produce a time-scale representation of the sensor signal; and

comparing by the processor of the time-scale representation of the sensor signal to a known sensor signal representation.

2. The method of claim 1 wherein the comparing further comprises:

measuring by the processor at least one energy value corresponding to at least a portion of the time-scale representation of the sensor signal;

comparing by the processor of the at least one energy value to at least one predetermined value corresponding to the known sensor signal representation; and

selectively identifying the sensor signal as an erroneous analyte sensor signal from among the plurality of analyte sensor signals.

3. The method of claim 2 wherein the at least one energy value comprises a plurality of energy values, one each for a plurality of time-scale zones in the time-scale representation of the sensor signal.

4. The method of claim 3 wherein the wavelet transform uses a Haar wavelet.

5. The method of claim 4 wherein the patient analyte comprises glucose.

6. The method of claim 5 wherein each analyte sensor signal comprises a blood draw portion and a blood clear portion.

7. The method of claim 6 further comprising:

discarding by the processor of the erroneous analyte sensor signal; and graphically displaying on a display device at least one numerical value corresponding to only a non-erroneous analyte sensor signal.

8. The method of claim 6 further comprising:

replacing by the processor of the erroneous analyte sensor signal; and graphically displaying on a display device at least one numerical value corresponding to only a non-erroneous analyte sensor signal.

9. A system for displaying patient analyte information, the system comprising: a display device to display graphical indications of numerical values corresponding to a plurality of analyte sensor signals;

a processor operably connected to the display device; and

a memory operably connected with the processor to store the data corresponding to numerical values for a patient analyte, the memory also operable to store computer program code which, when executed, causes the processor to apply a wavelet transform to a sensor signal to produce a time-scale representation and compare the time-scale representation to a known sensor signal representation.

10. The system of claim 9 wherein the patient analyte comprises blood glucose.

11. The system of claim 10 wherein each analyte sensor signal comprises a blood draw portion and a blood clear portion.

12. The system of claim 11 further wherein the processor is further caused to selectively identify an erroneous analyte sensor signal from among the plurality of analyte sensor signals, and at least one of replace and discard the erroneous analyte sensor signal.

13. The system of claim 12 wherein the processor compares the time-scale representation to a known sensor signal using a method comprising:

measuring at least one energy value corresponding to at least a portion of the time- scale representation; and

comparing the at least one energy value to at least one pre-determined value corresponding to the known sensor signal representation.

14. The system of claim 13 wherein the at least one energy value comprises a plurality of energy values, one each for a plurality of time-scale zones in the time-scale representation of the sensor signal.

15. A computer program product for processing a waveform from a patient analyte sensor, the computer program product including non-transitory computer program code in a tangible medium, the computer program code comprising:

instructions for obtaining a plurality of analyte sensor signals from the waveform, each analyte sensor signal corresponding to a periodic measurement of a patient analyte;

instructions for applying a wavelet transform to a sensor signal to produce a time-scale representation of the sensor signal;

instructions for comparing the time-scale representation of the sensor signal to a known sensor signal representation.

16. The computer program product of claim 15 wherein the computer program code further comprises:

instructions for measuring at least one energy value corresponding to at least a portion of the time-scale representation of the sensor signal;

instructions for comparing the at least one energy value to at least one predetermined value corresponding to the known sensor signal representation; and

instructions for selectively identifying the sensor signal as an erroneous analyte sensor signal from among the plurality of analyte sensor signals in the waveform.

17. The computer program product of claim 16 wherein the at least one energy value comprises a plurality of energy values, one each for a plurality of time-scale zones in the time-scale representation of the sensor signal.

18. The computer program product of claim 17 wherein the wavelet transform uses a Haar wavelet.

19. The computer program product of claim 18 wherein the patient analyte comprises glucose.

20. The computer program product of claim 19 where the computer program code further comprises:

instructions for discarding the erroneous analyte sensor signal; and instructions for graphically displaying on a display device at least one numerical value corresponding to only a non-erroneous analyte sensor signal.

21. The computer program product of claim 20 further comprising instructions for replacing the erroneous analyte sensor signal.

22. Apparatus for processing signals from a patient analyte sensor, the apparatus comprising:

means for obtaining a plurality of analyte sensor signals, each analyte sensor signal corresponding to a periodic measurement of a patient analyte; means for applying a wavelet transform to each sensor signal to produce a time-scale representation of the sensor signal;

means for comparing the time-scale representation of each sensor signal to a known sensor signal representation; and

means for selectively identifying a sensor signal as an erroneous analyte sensor signal from among the plurality of analyte sensor signals based on the comparing.

23. The apparatus of claim 22 further comprising:

means for measuring energy values corresponding to the time-scale representation of each sensor signal; and

means for comparing the energy values to a pre-determined value corresponding to the known sensor signal representation.

24. The apparatus of claim 23 further comprising:

means for discarding erroneous analyte sensor signals; and

means for graphically displaying numerical values corresponding to only non- erroneous analyte sensor signals.

25. The apparatus of claim 24 further comprising means for replacing erroneous analyte sensor signals.