SYSTEM, METHOD, AND NON-TRANSITORY COMPUTER READABLE MEDIUM FOR AUTOMATED ASSESSMENT OF NEURAL RESPONSE RECORDINGS - Patent application

The system automates the evaluation of neural response recordings in neuromodulation systems by using basis functions to decompose CAPs and artifacts, enhancing signal quality detection and optimizing therapeutic settings in implantable devices.

JP7595639B2Active Publication Date: 2024-12-06SALUDA MEDICAL PTY LTD
View PDF 4 Cites 0 Cited by

Patent Information

Application Number
JP2022501298
Authority / Receiving Office
JP · JP
Patent Type
Patents
Current Assignee / Owner
Priority Date
2019-07-12
Filing Date
2020-07-13
Publication Date
2024-12-06
Estimated Expiration
2040-07-13

AI Technical Summary

Technical Problem

The challenge in neuromodulation systems is accurately detecting and recording compound action potentials (CAPs) amidst stimulation artifacts and noise, particularly in implantable devices with limited power budgets and complex electrode configurations, which complicates the identification of optimal therapeutic settings.

Method used

A system and method for automated evaluation of neural response recordings using basis functions to decompose neural recordings, determining parameters that estimate CAPs and artifacts, and outputting quality indicators to identify optimal stimulation and recording configurations.

Benefits of technology

This approach enables efficient, automated identification of optimal neuromodulation settings, reducing manual clinical work and improving therapeutic efficacy by providing high-quality ECAP signal detection, even in power-constrained implantable devices.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure 0007595639000027
    Figure 0007595639000027
  • Figure 0007595639000028
    Figure 0007595639000028
  • Figure 0007595639000029
    Figure 0007595639000029
Patent Text Reader

Abstract

The automated evaluation of neural response recordings includes storing a set of basis functions including at least one compound action potential basis function and at least one artifact basis function. Neural recordings of electrical activity in neural tissue are obtained by applying stimulation using a single stimulation and recording configuration. Each neural recording is decomposed by determining at least one parameter that estimates at least one of the compound action potential and the artifact. The at least one parameter is determined for each of a plurality of neural recordings to yield a plurality of values. A spread of the plurality of values ​​is determined. A small spread results in an indication of high quality of the neural response recording. A large spread results in an indication of low quality of the neural response recording.
Need to check novelty before this filing date? Find Prior Art

Description

[Technical field]

[0001] CROSS-REFERENCE TO RELATED APPLICATIONS This application claims the benefit of Australian Provisional Patent Application No. 2019902485, filed July 12, 2019, which is incorporated herein by reference.

[0002] The present invention relates to electrical recordings of neural activity, such as compound action potentials evoked by neural stimulation, and in particular to systems and methods for improving detection of neural responses in recordings when the recordings are obtained in the presence of stimulation artifacts, noise, and the like. [Background technology]

[0003] Electrical neuromodulation is used or envisioned for use to treat a variety of disorders, including chronic pain, Parkinson's disease, and migraines, as well as to restore functions such as hearing and motor functions. Neuromodulation systems apply electrical pulses to neural tissue to produce a therapeutic effect. Such systems typically include an implanted electrical pulse generator and a power source, such as a battery, that is rechargeable by transcutaneous inductive transfer. An electrode array is connected to the pulse generator and positioned near the neural pathway of interest. The electrical pulses applied to the neural tissue by the electrodes cause depolarization of neurons, which generates antidromic, orthodromic, or both propagating action potentials to achieve a therapeutic effect.

[0004] For example, when used to reduce chronic pain, electrical pulses are applied to the dorsal (DC) portion of the spinal cord and an electrode array is placed in the spinal epidural space. Repeated stimulation of the dorsal fibers in this manner reduces the transmission of pain sensations from that portion of the spinal cord to the brain.

[0005] Generally, the electrical stimulation generated within a neuromodulation system induces neural action potentials that have either an inhibitory or excitatory effect: an inhibitory effect can be used to modulate undesired processes, such as pain transmission, or an excitatory effect can be used to induce desired effects, such as muscle contraction or stimulation of the auditory nerve.

[0006] Action potentials generated among many fibers sum to form a compound action potential (CAP). The CAP is the sum of the responses from many single fiber action potentials. When recording the CAP electrically, the measurement includes the depolarization of many different fibers. Propagation velocity is determined primarily by fiber diameter, with speeds of up to 60 ms for the large myelinated fibers found near the dorsal root entry zone (DREZ) and dorsal columns. -1 CAPs generated from the firing of a group of similar fibers are measured as a positive peak P1, followed by a negative peak N1, followed by a second positive peak P2 in the recorded potential. This is caused by a region of activation passing through the recording electrode as the action potential propagates along the individual fibers, resulting in a typical three-peaked response profile. Depending on the polarity of the stimulus and the configuration of the recording electrode, some CAP measurement profiles may be of reverse polarity, with two negative peaks and one positive peak.

[0007] To better understand the effects of neuromodulation and / or other neurostimulation, and to provide, for example, stimulators controlled by neural response feedback, it is desirable to accurately detect and record CAPs evoked by stimulation. Evoked CAPs (ECAPs) are easier to detect if they appear at a later time point than the artifact or if the signal-to-noise ratio is high enough. The artifact is often limited to a time window 1-2 ms after stimulation, so measurements of the response can be more easily obtained provided that the neural response is detected after this time window. This can be achieved with long distances between the stimulating and recording electrodes (e.g., 60 ms). -1 For nerves that are conducted via the stimulatory system, the propagation time from the stimulation site to the recording electrode may exceed 2 ms (>12 cm for nerves that are conducted via the stimulatory system).

[0008] However, to characterize the response from the spine, high stimulation currents and close proximity between the electrodes are required. Similarly, any implanted neuromodulation device is necessarily compact in size, and therefore the stimulating and recording electrodes are necessarily close together in order for such devices to monitor the effects of the applied stimulation. In such a situation, the measurement process must directly overcome the artifacts. However, overcoming such artifacts can be a difficult task, since the ECAP signal components observed in the neuromeasurement typically have maximum amplitudes in the microvolt range. In contrast, the stimulation applied to induce ECAP is typically a few volts, which generates electrode artifacts evident in the neuromeasurement as a damped output of a few millivolts that is partially or completely coincident with the ECAP signal, presenting a significant obstacle to isolating and even detecting the much smaller ECAP signal of interest.

[0009] The difficulty of this problem is exacerbated when attempting to implement CAP detection within an implanted device. A typical implant has a power budget that allows a limited number of processor instructions per stimulus, e.g., hundreds or thousands, in order to preserve a desired battery life. Thus, if one were to use a CAP detector for an implanted device periodically (e.g., on the order of once per second), one would need to ensure that the detector consumes only a small portion of the power budget.

[0010] Further complexity arises from the increased configurability of the stimulation and recording modes of the neurostimulator. The variables include the selection of stimulation and / or recording electrodes from a potentially large number of available electrodes based on an implanted electrode array, multiple stimulation parameters, and multiple recording parameters. Clinical validation of the proper operation of a neurostimulator should ideally involve identifying the optimal settings of such variables for optimal therapeutic effect, but the number of combinations that must be tested can be very large and currently must be done in large part by the clinician, making the clinical fitting process time-consuming and costly.

[0011] Any discussion of documents, acts, materials, devices, articles or the like which has been included in the present specification is merely for the purpose of providing a context for the present invention, and should not be construed as an admission that any or all of the contents thereof form part of the prior art or were common general knowledge in the field relevant to the present invention by reason of their existence prior to the priority date of each claim of this application.

[0012] It will be understood that throughout this specification the word "comprise" or variations such as "comprises" or "comprising" imply the inclusion of a stated element, integer, or step, or group of elements, integers, or steps, but not the exclusion of any other element, integer, or step, or group of elements, integers, or steps.

[0013] As used herein, a statement that an element may be "at least one" of a list of options should be understood to mean that the element may be any one of the listed options, or any combination of two or more of the listed options. Summary of the Invention [Means for solving the problem]

[0014] According to a first aspect, the present invention provides a system for automated evaluation of neural response recordings, the system comprising: a memory that stores a set of basis functions including at least one of (a) compound action potential basis functions and (b) artifact basis functions; an input for receiving a plurality of neural recordings of electrical activity in neural tissue, the neural recordings being obtained by repeatedly applying stimulation using a single configuration of stimulation and recording; a processor configured to decompose each neural recording by determining at least one parameter that estimates at least one of a compound action potential and an artifact from a set of basis functions, the processor being further configured to iteratively determine a plurality of values ​​of the at least one parameter for each of the plurality of neural recordings, the processor being further configured to determine a spread of the plurality of values, the processor being further configured to output an indication of high quality of the neural response recording if the spread is small, and to output an indication of low quality of the neural response recording if the spread is large; Includes.

[0015] According to a second aspect, the present invention provides a method for automated evaluation of neural response recordings, the method comprising: storing a set of basis functions including at least one compound action potential basis function and at least one artifact basis function; receiving a plurality of neural recordings of electrical activity in neural tissue, the neural recordings being obtained by repeatedly applying stimulation using a single configuration of stimulation and recording; decomposing each neural recording by determining at least one parameter that estimates at least one of a compound action potential and an artifact from a set of basis functions, and iteratively determining multiple values ​​of the at least one parameter for each of the multiple neural recordings; determining a spread of a plurality of values; outputting an indication that the quality of the neural response recording is high if the diffusion is low, and outputting an indication that the quality of the neural response recording is low if the diffusion is high. Includes.

[0016] According to a further aspect, the present invention provides a non-transitory computer readable medium for automated assessment of neural response recordings comprising instructions which, when executed by one or more processors, cause the method of the second aspect to be performed.

[0017] The indication of quality of the neuroresponse recording output by the processor may be a binary indication of high or low quality, e.g., the spread is compared to a threshold. Alternatively, the indication of quality of the neuroresponse recording may be defined on a scale of three or more quality indication levels, or may be based on a substantial continuum from high to low quality. For example, a quality score may be output, and may be normalized to fall somewhere within a desired range, such as [0:1]. The determination of the quality score may be calibrated by reference to a clinician's scoring of a test set of neurorecordings. Similarly, the normalization of the quality score may be calibrated by reference to a clinician's scoring of a test set of neurorecordings, e.g., the clinician may use the test set to define the midpoint, spread, or growth rate of a normalizing function, such as a sigmoid.

[0018] The ECAP quality score can be used to evaluate a selected configuration of stimulation and recording. Separate ECAP quality scores can be additionally obtained for one or more other configurations of stimulation and recording, for example by changing the selection of stimulation electrodes and / or the selection of recording electrodes and generating a new ECAP quality score for the new configuration. By comparing the quality scores of each configuration, a selection of stimulation and recording configurations for ongoing treatment can be made. A preferred embodiment can include an implant and / or associated clinical software configured to test all possible configurations of stimulation and recording in an automated manner, where all implanted electrodes are used sequentially for stimulation to yield a matrix or set of ECAP quality scores for the entire implanted electrode array, and all possible recording electrodes for each such stimulation configuration are used sequentially to obtain an ECAP quality score for each stimulation and recording configuration. Such an embodiment thus provides an automated means by which an optimal configuration of stimulation and recording can be quickly identified by reference to a set of ECAP quality scores. Such an embodiment can thus eliminate difficult manual clinical tasks, improve the time and cost of optimally fitting a neurostimulator, and / or improve the therapeutic outcomes of the implanted individual. Additionally or alternatively, some embodiments may allow a matrix or set of ECAP quality scores to be continually created or updated for some or all possible electrode configurations during operation of the implanted device. For example, a processor of the implanted device may be configured to create or update a matrix or set of ECAP quality scores at predefined time intervals, or after a certain number of stimuli have been delivered, and / or at other times as needed. Based on such ECAP quality scores created during ongoing operation of the device, the device may be configured to employ updated stimulation configurations, such as selection of which electrodes to use as stimulation electrodes for ongoing treatment, and to utilize optimal or preferred ECAP quality scores associated with the updated stimulation configurations.Additionally or alternatively, based on such ECAP quality scores generated during ongoing operation of the device, the device can be configured to adopt an updated recording configuration, such as selecting which electrodes to use as recording electrodes during ongoing treatment, and to take advantage of an optimal or preferred ECAP quality score associated with the updated recording configuration.

[0019] The spread may be calculated as the standard deviation of a parameter, the variance of a parameter, the interquartile range or decile range of a parameter, or may include any other suitable statistical measure of the spread of the data.

[0020] In some embodiments, at least one parameter may include a correlation between the observed ECAP and a predefined basis function, including an analytically defined compound action potential basis function, and such parameters are referred to herein as correlation parameters. Such embodiments recognize that when determining the quality of a recording, it is advantageous to consider how well the observed ECAP correlates with a predefined analytical or "ideal" ECAP.

[0021] Additionally or alternatively, the at least one parameter can include a frequency of the observed ECAP, measured, for example, from the duration of one or more lobes of the observed ECAP and / or from the time offset of the ECAP peaks in the recording and / or from spectral analysis of the recording, such parameters being referred to herein as a frequency parameter. Such embodiments recognize that frequency is a particularly useful parameter to monitor, as it has been found that large variations in ECAP frequency from stimulus to stimulus correlate with poor ECAP signal quality and suboptimal treatment.

[0022] Additionally or alternatively, the at least one parameter can include a time offset of the observed ECAP relative to the time of stimulation, referred to herein as the offset parameter. Such embodiments recognize that offset is a particularly useful parameter to monitor, as it has been found that large variability in ECAP offset from stimulation to stimulation correlates with poor ECAP signal quality and suboptimal treatment.

[0023] In some embodiments, the basis functions are analytically defined compound action potential basis functions. In such embodiments, the rate at which ECAP is detected among multiple recordings can further be used to define the quality of the neural response recordings. This rate is referred to herein as the detection rate.

[0024] In some embodiments, more than one nerve recording can be obtained for each ECAP, such that a comparison parameter derived by comparing the two or more recordings can additionally or alternatively be used to assess the quality of the ECAP. For example, the conduction velocity and / or variance of each ECAP can be determined from two or more nerve recordings of that ECAP, and the spread of conduction velocity and / or the spread of variance can be used to derive ECAP signal quality.

[0025] In embodiments where multiple parameters are available, the multiple parameters may be processed by any suitable predefined function to generate a single quality score. For example, in one embodiment, the quality score may be determined as follows: Score = (detection rate * correlation) / (frequency spread + offset spread).

[0026] In such embodiments, each element of the function may be scaled or adjusted by any suitable tuning constant, factor, etc., to better calibrate the output to the clinician's opinion. For example, if the offset spread is measured in ms, then this parameter may be multiplied by 100 in the above function.

[0027] Note that a higher detection rate and higher correlation correspond to higher ECAP signal quality, and a preferred function is proportional to these parameters and / or places these parameters in the numerator of the function. Note that conversely, a larger frequency spread and a larger offset spread correspond to lower ECAP signal quality, and a preferred function is inversely proportional to these parameters and / or places these parameters in the denominator of the function. Thus, other embodiments may utilize any other suitable function consistent with these observations.

[0028] In embodiments utilizing differential ECAP recording by using two sense electrode inputs to a differential measurement amplifier, some or all of the above parameters can be obtained for both the positive ECAP component of the differential ECAP recording and the negative ECAP component of the differential ECAP recording.

[0029] The ECAP signal quality score may be normalized, for example to the range [0:1], by any suitable function, such as a sigmoid function. The normalized score may be determined, for example, by the following formula: Normalized score = 1-1 / (1+α * Score).

[0030] In such an embodiment, a tuning constant a can be selected to calibrate the normalized score output to the clinician's opinion, e.g., a=800 in one embodiment. In alternative embodiments, a can be replaced by any suitable tuning constant or scaling factor, etc. For example, if the scores designated by a human clinician are selected from "unsatisfactory," "marginal," and "satisfactory," a or other constant can be selected as needed such that the normalized score generated for at least 90% of the signal sets labeled "unsatisfactory" by the skilled clinician is less than 0.4. This form represents an implementation-independent threshold that the on-site clinical engineer may refer to when deciding which stimulator configuration to use, where a normalized score below 0.4 indicates that additional programming is required, while a normalized score above 0.6 predicts that the existing stimulation and recording configuration program will result in a clinically usable growth curve. In such an embodiment, if a normalized score between 0.4 and 0.6 is output, the stimulator configuration is considered marginal, meaning that it is unknown whether the stimulator configuration will result in a clinically usable growth curve.

[0031] Importantly, embodiments of the present invention recognize that a system for automatically assessing neural response recordings should preferably provide an output that is insensitive to the stimulation current used. Because ECAP amplitude depends on the stimulation current, this requirement ensures that the system does not erroneously equate a larger ECAP amplitude with a better quality of stimulation and recording configuration. The parameters selected in preferred embodiments of the present invention advantageously do not depend solely on the ECAP amplitude, and therefore such embodiments do not erroneously equate ECAP amplitude with the quality of the stimulation and recording configuration. It should further be noted that the magnitude of ECAP is posture dependent, due to both the varying distance from the stimulation electrode to the nerve and the varying distance from the nerve to the recording electrode, providing another reason why it is advantageous to select parameters that do not only represent the amplitude of the ECAP recorded.

[0032] References to estimation or determination herein should be understood to refer to automated processes performed on data by a processor operating to execute a predefined estimation or determination procedure. The techniques presented herein can be implemented by hardware (e.g., using an application specific integrated circuit (ASIC)), or by software (e.g., using instructions tangibly stored on a computer readable medium to cause a data processing system to perform the steps described above), or by a combination of hardware and software. The present invention can also be embodied as computer readable code on a computer readable medium. A computer readable medium may include any data storage device that can store data that can thereafter be read by a computer system. Examples of computer readable media include read only memory ("ROM"), random access memory ("RAM"), CD-ROM, DVD, magnetic tape, optical data storage device, flash storage device, or any other suitable storage device. The computer readable medium may be distributed across computer systems coupled to a network such that the computer readable code is stored and executed in a distributed fashion.

[0033] Thus, embodiments of the present invention may provide a partially or fully automated process for clinically validating proper operation of a neurostimulator device by reference to ECAP signal quality using an automated process for testing multiple or all combinations of stimulation variables in a computationally efficient manner that reduces the time and cost of clinical adaptation. In particular, the described embodiments provide a process that leverages data parameters that can be rapidly acquired by a largely automated process, and by specifically leveraging such parameters and avoiding or minimizing steps that require the involvement of a human clinical expert, those embodiments of the present invention advantageously avoid the significant time and expense of conventional approaches that involve clinically observing ECAP recordings and / or clinically deriving ECAP growth curves in each relevant position to identify optimal treatment settings for the device. Some embodiments may generate a signal quality score in a fraction of a second, such as within 250 ms, and may be updated iteratively as quickly as once within 62.5 ms.

[0034] Further embodiments of the present invention may utilize the signal quality score to continually control the operation of the feedback loop of the implanted neuromodulation device. For example, such embodiments may cause the feedback loop to stop operating or to react more slowly when the ECAP signal quality score is low. Additionally or alternatively, such embodiments may cause the feedback loop to start operating or to react more quickly when the ECAP signal quality score is high.

[0035] Examples of the invention will now be described with reference to the accompanying drawings, in which: [Brief description of the drawings]

[0036] [Figure 1] 1 illustrates a schematic diagram of an implanted spinal cord stimulator. [Diagram 2] FIG. 1 is a block diagram of an implanted neurostimulator. [Diagram 3] 1 is a schematic diagram showing the interaction of an implanted stimulator with a nerve. [Figure 4] 1 illustrates a scrubbing process. [Diagram 5] The signal flow diagram is shown. [Figure 6] ECAP and artifact basis functions and their results are presented. [Figure 7] 1 shows a system for ECAP and artifact estimation. [Figure 8] 1 illustrates an architecture for a signal quality indicator according to one embodiment of the present invention. [Figure 9] 1 illustrates a clinical system according to one embodiment of the present invention. [Figure 10] FIG. 2 is a state machine diagram illustrating one implementation of measurement electrode scanning (MES), in accordance with one embodiment of the present invention. [Figure 11] 1 is a flow diagram of the MES procedure performed by the implant. [Figure 12] An example of how to configure the MES position when the stim electrode is E2 is shown. [Figure 13] An example of MES output is shown below. [Figure 14] An example of MES output is shown below. [Figure 15] An example of MES output is shown below. [Figure 16] An example of MES output is shown below. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0037] 1 shows a schematic representation of an implanted spinal cord stimulator 100. The stimulator 100 comprises an electronic module 110 that is implanted at a suitable location in the patient's lower abdomen or upper posterior gluteal region, and an electrode assembly 150 that is implanted in the epidural space and connected to the module 110 by suitable leads. Many aspects of the operation of the implanted neuro device 100 are reconfigurable by an external control device 192. Additionally, the implanted neuro device 100 serves to collect data that is communicated to the external device 192.

[0038] 2 is a block diagram of the implanted neurostimulator 100. The module 110 includes a battery 112 and a telemetry module 114. In embodiments of the invention, any suitable type of transcutaneous communication 190, such as infrared (IR), electromagnetic, capacitive, and inductive transfer, may be used by the telemetry module 114 to transfer power and / or data between an external device 192 and the electronics module 110.

[0039] The module controller 116 has an associated memory 118 that stores patient settings 120, control programs 122, etc. The memory 118 also stores a set of basis functions including at least one of (a) compound action potential basis functions and (b) artifact basis functions to aid in adapting or refining device operation based on the ECAP quality score. The external device 192 also stores a set of basis functions including at least one of (a) compound action potential basis functions and (b) artifact basis functions to enable clinical adaptation based on the ECAP quality score. The controller 116 controls the pulse generator 124 to generate stimulation in the form of current pulses according to the patient settings 120 and the control program 122. The electrode selection module 126 switches the generated pulses to appropriate electrodes of the electrode array 150 to deliver the current pulses to tissue surrounding the selected electrodes. The measurement circuitry 128 is configured to capture measurements of neural responses sensed at sense electrodes of the electrode array selected by the electrode selection module 126.

[0040] 3 is a schematic diagram illustrating the interaction of the implanted stimulator 100 with a nerve 180, which in this case is the spinal cord, although alternative embodiments may be placed adjacent to any desired neural tissue, including peripheral nerves, visceral nerves, the parasympathetic nervous system, or brain structures. The electrode selection module 126 selects the stimulation electrode 2 of the electrode array 150 to deliver triphasic current pulses to the surrounding tissue including the nerve 180, although other embodiments may additionally or alternatively deliver biphasic tripolar stimulation. The electrode selection module 126 also selects the return electrode 4 of the array 150 to restore the stimulation current and maintain zero net charge transfer.

[0041] Delivering appropriate stimulation to the nerve 180 induces a neural response including a compound action potential that propagates along the nerve 180 as shown, with the therapeutic objective being to create a sensory illusion at a desired location, which in the case of a spinal cord stimulator for chronic pain, may be. Stimulating electrodes are used to deliver stimulation at 30 Hz to do so. To fit the device, the clinician applies a stimulus that results in a sensation that the user experiences as a sensory illusion. If the sensory illusion is in a location and size consistent with the user's body area affected by pain, the clinician recommends the configuration for continued use. While this clinical fitting process has traditionally been laborious, the embodiments described herein provide a means for automatically assessing the fit of the device based on an ECAP quality score, including stimulation configuration and recording configuration, to improve the efficiency of this fitting process.

[0042] The device 100 is further configured to detect the presence and electrical profile of a compound action potential (CAP) propagating along the nerve 180, whether the CAP is induced by stimulation from electrodes 2 and 4 or otherwise. To that end, any electrode of the array 150 can be selected by the electrode selection module 126 to serve as the measurement electrode 6 and the measurement reference electrode 8. The stimulator case can also be used as the measurement or reference electrode or the stimulation electrode. The signals detected by the measurement electrodes 6 and 8 are conveyed to the measurement circuit 128, which can operate, for example, according to the teachings of the applicant of the present invention, WO2012155183, the contents of which are incorporated herein by reference. The present invention recognizes that in situations such as that shown in FIG. 3, where the recording electrode is close to the stimulation site, stimulation artifacts represent a significant impediment to obtaining an accurate recording of the compound action potential, but reliable and accurate CAP recording is a key enabler for a range of neuromodulation techniques.

[0043] Specifically, ECAP recordings allow the device to enter a closed-loop feedback mode, where a target ECAP level is continually determined by the device, and the device responds by adjusting future stimulation pulses to perturbations in the feedback loop, such as postural changes. However, the feedback operation is critically dependent on the quality of the response recordings obtained by the device. Although the quality can be reliably assessed by suitably experienced human clinicians, such assessment is laborious. The quality can also be assessed by obtaining a complete growth curve for each configuration, which represents the growth of ECAP amplitude as a function of increasing stimulation current. Such an assessment allows one to ascertain whether the configuration results in a growth curve with a well-defined threshold (stimulation current below which ECAP does not occur) and also whether the growth curve increases monotonically above the threshold (which is important for the stability of the feedback loop). However, obtaining and evaluating the growth curves is also laborious.

[0044] The present invention thus provides a system and method for automatically assessing the quality of neural response recordings.

[0045] More specifically, an embodiment of the present invention uses a set of basis functions in memory to decompose each neuronal recording by determining at least one parameter that estimates at least one of the compound action potential and the artifact. This is therefore a method for separating composite signals when the signal components belong to a closed space of signals that can be represented by distinct basis sets. In neuromodulation, this is used to separate the "ECAP part" and the "artifact part" of the recording signal.

[0046] A composite signal is a signal that is constructed by the sum of other signals, referred to herein as basis signals. The basis element signal separation technique of the present invention estimates the basis signals of the composite signal given only the composite signal and without knowledge of the exact basis signals. An embodiment of the present invention provides a blind signal separation algorithm that can estimate some knowledge about the basis signals. That is, an embodiment of the present invention recognizes that it can be assumed that each basis signal can be represented by a set of basis functions. Unlike blind signal separation algorithms with multiple inputs and one output, an embodiment of the present invention exploits this assumption to provide a deterministic estimate of the basis signals.

[0047] In the field of neurostimulation, the mixed signal may be a combination of ECAP and stimulation artifacts. In some cases, it becomes necessary to decompose the signal and analyze the components. Analyzing the individual components can reveal characteristics of the signal components that can be used in many advantageous ways. In some cases, analyzing the components of the mixed signal can reveal errors in the system. Furthermore, there may be situations where the mixed or composite signal has dominant but unwanted components that mask the essential components. In such cases, it is necessary to decompose the mixed signal into its components, eliminate the unwanted components, and analyze the essential components and their properties.

[0048] An embodiment of the invention decomposes a mixed signal by determining at least one of the signals that make up the mixed signal from a set of basis functions. This embodiment separates the mixed signal into its fundamental components by modeling each fundamental component with a basis. This embodiment can be applied in neuromodulation in separating the ECAP waveform from the artifact waveform (as well as noise) given a signal recording that is a mixture of those signals. Such separation results in more robust feature extraction from the ECAP, including the ECAP magnitude, which is a feature used by the closed loop control system of Figures 1-3. Additional features such as ECAP peak location can also be measured more robustly, which is of scientific interest. An embodiment of the invention estimates both the artifact and ECAP simultaneously, balancing the ECAP and artifact signal contributions to provide the "best" representation of the recorded signal. An embodiment of the invention provides a noise-free ECAP estimate and can impose certain signal characteristics (e.g., a baseline of 0V) subject to the definition of the ECAP basis set. Furthermore, embodiments of the present invention are efficient (O(n)) and run in deterministic time (unlike non-deterministic methods), which means that embodiments of the present invention can potentially be built into firmware to provide improved real-time ECAP magnitude estimation without the need for human-tuned filters.

[0049] FIG. 4 illustrates the scrubber process 400. A scrubber is an algorithm that estimates the ECAP and artifact components of any composite signal, shown at 410. A composite signal is defined as a signal that is composed of the sum of multiple separate components. In the context of ECAP measurements, the components of the composite measurement are artifacts, neurophysiological responses to stimuli (ECAP), and everything else. The primary goal of the scrubber 420 is to isolate the ECAP. However, artifact estimation is usually a by-product of this task, and is useful in itself, as insight into artifact mechanisms can help minimize artifacts in future designs. What is left behind consists of electronic noise and neurophysiological noise unrelated to the stimulus.

[0050] An embodiment of the present invention employs the following process: Each basis signal is represented as a linear combination of basis functions. Consider a composite signal having two basis signals.

[0051]

number

[0052] The basis functions are derived empirically based on experience and alternative models of the underlying signals. For the purposes of illustration, they are considered constants. A set of linear equations can be written that can be solved by matrix inversion to calculate pairwise dot products of the basis functions and between each basis function and the composite signal to obtain a set of coefficients alpha and beta. Given the alpha coefficient, a basis representation of f(x) can be written and f(x) can thus be estimated. Similarly, given the beta coefficient, g(x) can be estimated. This method is not limited to composite signals containing two components, although the problems to which this method is applied in the described field of neuromodulation have only two components.

[0053] The basis element signal separation approach of the present invention is a mathematical tool for decomposing a composite signal. Consider a signal that includes an ECAP component f(t) and an artifact component g(t). The signal σ(t) measured in the patient is then

number

[0054] The embodiments of the present invention work on the assumption that the ECAP and artifact signal components belong to separate function families, i.e., ECAP is always a short oscillatory event, whereas artifact is an exponential looking signal. For each separate function family, a basis for representing it can be predefined. The basis coefficients can be calculated for the appropriate basis functions, and the basis expansions of ECAP and artifact can be separated, respectively. The basis expansion of ECAP gives an estimate of the artifact-free ECAP component.

[0055] The calculation of the basis coefficients balances the contributions of each of the basis functions in such a way that the overall signal is approximated as best as possible. In other words, the contributions of the estimated ECAP and artifacts are balanced to best model the recorded signal. To achieve better performance, embodiments of the present invention assume that all ECAPs belong to a certain family of functions and that no forms of ECAP outside that family exist. At the time of writing, ECAPs with slow responses, such as those described in WO2015070281, are outside the ECAP family of functions used by embodiments of the present invention and therefore cannot be properly estimated. Therefore, it may be more appropriate to use other scrubbers when dealing with signals that are not adequately modeled by the ECAP basis used at the time.

[0056] The above methods form blocks in the signal flow diagram of FIG. 5. In some embodiments, pre-processing and post-processing are used to improve the signal estimate. For example, pre-processing can be used to reduce high frequency noise in the signal. However, a feedback mechanism is used to improve the construction of the basis set. A rough "first guess" basis can be used to approximate the signal, and the resulting estimate can be used to refine the basis set on subsequent passes. For example, a first pass can guess an ECAP basis to obtain a good estimate of the artifact. Subtracting artifacts from the signal and using signal correction methods can be used to refine the selection of the ECAP basis. Re-running the algorithm with the improved basis results in a better estimate of both the ECAP and the artifact.

[0057] According to an embodiment of the present invention, the artifacts are modeled using three basis functions:

number

[0058] The unitary basis function φ1 captures the DC content of the measured signal. The linear basis function φ2 captures the artifact component due to amplifier drift. The exponential basis function φ3 captures the chemical charge relaxation component of the artifact. The decay constants of the exponential components can be any suitable variable, and the above values ​​were empirically determined based on the performance of the model against a library of human artifact recordings. Different devices may exhibit different artifacts and / or ECAP results and therefore require different constants, which can also be obtained empirically.

[0059] When the algorithm of the present invention is applied, the artifact component of the signal is represented by the following equation: A(t)=αφ1(t)+βφ2(t)+γφ3(t)

[0060] Although this model is simple, it has been applied to thousands of representative neural recordings from human patients and has been shown to work well. In combination with the ECAP basis functions, the composite model accurately estimates the recorded signal.

[0061] Abnormal neurological artifacts such as background neuronal activity or slow responses are not modeled in embodiments of the present invention, but may be incorporated according to alternative embodiments of the present invention. The estimates obtained from the approach of embodiments of the present invention remove such features, and therefore, at least in this embodiment, the results cannot be relied upon to measure non-ECAP neurological features.

[0062] The ECAP basis functions are defined using a product of gamma probability density functions with parameters k=1.7 and θ=0.60. φ(t)=(ft) k-1 e -ft / θ

[0063] This is a piecewise function consisting of one period of a sine wave followed by an exponential function, so that the derivative is continuous at its boundaries:

number

[0064] Scale the time axis by ν, and ν(t- t0 ), the ECAP basis functions can be stretched and shifted in time. Such stretched and scaled ECAP basis functions are called parametric ECAP basis functions: φ ν,t0 (t).

[0065] There are two separate ECAP models, one for single-ended measurements and one for differential measurements. The single-ended ECAP basis consists of one parametric ECAP basis function, where ECAP E is

[0066]

number

[0067] The differential ECAP basis is formed by the difference of two parametric ECAP basis functions giving the ECAP model:

[0068]

number

[0069] In either model, the time stretch and time offset (corresponding to the ECAP vibration frequency) are selected such that κ or κ+ is positive and κ- is negative. A set of ECAP frequencies and offsets are checked by embodiments of the present invention to ensure that this condition holds. The frequencies and offsets selected to model the ECAP component of the recorded signal are selected to provide the best possible fit to the recordings using both the ECAP model and the artifact model.

[0070] It should be noted that the single-ended ECAP model assumes a fixed ratio between peak height and peak time: neurophysiological parameters such as width at half height or ratio of n1:p2 are entirely determined by the temporal stretch v applied to the parametric basis functions.

[0071] As with the artifacts, this assumption is verified by fitting parametric basis functions to real-world single-ended measurements.

[0072] In the case of the differential model, such neurophysiological parameters can vary independently of v+ and v- and additionally depend on the composition of the ECAP estimates; that is, κ+ and κ- provide additional degrees of freedom. Although the relative neurophysiological parameters are allowed to vary, they have limited degrees of freedom compared to the more free-form ECAP model. As with the assumptions of single-ended ECAP, the constraints of this model have been verified by fitting the differential ECAP basis to real-world differential measurements.

[0073] The range of parametric ECAP frequencies is limited to a set of linearly spaced frequencies from 500 Hz to 2 kHz. The upper limit of 2 kHz was chosen to minimize interference of broadband (up to 8 kHz) noise on the parameter selection procedure. The lower limit of 500 Hz was chosen to limit interference of artifacts on the parameter selection procedure. Parametric ECAP that is slow enough closely resembles artifacts within a limited time window. The range of offsets examined was chosen to be significantly broader to model real-world ECAP, yet reasonably constrained to maintain computational performance.

[0074] So far, we have assumed that each recorded signal contains an ECAP. However, in practice, this is never the case for subthreshold signals, i.e., when the applied stimulus is insufficient to recruit any neural response, so that in such situations the recorded signal necessarily does not contain any ECAP. Including ECAP basis functions in the model for subthreshold signals would be problematic, since the ECAP would be fitted to the noise in the signal, making the estimate meaningless. In addition, the artifactual components of the signal would be misrepresented as ECAPs, and artifactual features would be balanced in the composite model.

[0075] It is therefore desirable to include a mechanism to detect the presence of ECAP in a signal so that the ECAP basis can be included in the overall model only if the underlying ECAP is real. An embodiment of the present invention incorporates such a mechanism. A signal is modeled using an artifact-only basis and a combined ECAP and artifact basis. A set of signal features is derived from the estimates provided by both models and combined with signal features from the recorded signal. A set of signals known to contain both ECAP and artifacts or only artifacts is analyzed by an embodiment of the present invention and the derived set of features is saved. Machine learning is used to train a classifier with categories of "ECAP" or "no ECAP". After sufficient training, the resulting classifier can automatically determine the presence of ECAP in a signal. An embodiment of the present invention is directed to detecting ECAP in signals containing ECAP with 85% accuracy and rejecting ECAP in signals containing only artifacts with 95% accuracy.

[0076] These concepts are combined to arrive at the complete algorithm of an embodiment of the present invention, shown in FIG.

[0077] The recorded signal is first modeled using an artifact-only basis, under the assumption that it does not include ECAP. Such modeling results in an estimate of the artifact in terms of basis coefficients, regardless of the presence of ECAP. If ECAP is present, this estimate can be refined by also including an ECAP basis. To help better determine the parametric ECAP basis, the initial artifact estimate is subtracted from the recorded signal. The estimated artifact and derived features are passed to the "ECAP presence classification" (or ECAP detector) block for further use.

[0078] Once the parameters for the parametric ECAP basis are determined, the coefficients of the combined ECAP and artifact basis are determined. The resulting estimates and feature set are passed to the ECAP detector.

[0079] The ECAP detector now has everything it needs to classify the presence of ECAP in the recorded signal and based on that decision, return an estimate of ECAP and artifacts, or just artifacts.

[0080] The steps of this method are as follows: a. Capturing / recording a composite signal, the composite signal having two or more additive components. Selecting a first basis set from the pool of basis sets corresponding to the first signal component, and selecting a second basis set from a separate pool of basis sets corresponding to the second signal component. c) determining a first component and a second component of the composite function based on the basis functions, where an estimate of the first component is determined as a linear expansion of the first basis set and an estimate of the second component is determined as a linear expansion of the second basis set. d. Iteratively improve the basis set using estimated components from previous iterations.

[0081] The following discussion delves into the mathematics behind embodiments of the present invention. The coefficients are determined as follows: Let the recorded signal be σ(t) and the underlying ECAP and artifact components be f(t) and g(t), respectively. The problem we are trying to solve is to find estimates of the unknown f(t) and g(t) using the known recorded signal σ(t). For simplicity we will assume there is no noise in the signal. Thus,

[0082]

number

[0083] Next, a set of finite basis functions

number

number

[0084]

number

[0085] Then by simple substitution we get:

number

[0086] At this stage of the problem, we know the basis set but not the coefficients for a particular signal σ(t). We can use the coefficients to recover estimates of f(t) and g(t). We then recover them:

[0087] fφ i Consider the dot product of the following function with respect to any basis function of (t), and by linearity of the dot product we obtain:

number

[0088] Similarly, g:φ l Consider the inner product of a function with respect to any basis function of (t).

number

[0089] Equations (4) and (5) give a system of n+m linear equations with n+m unknowns (coefficients a k and b j ). Determining the coefficients is therefore a problem of solving the following linear equations: Hv=b However, the following equation holds:

[0090]

number

[0091] Therefore, H -1 The coefficients can be solved by b. The matrix H is invertible if and only if none of the basis functions from the ECAP basis belong to the span of the artifact basis (and vice versa) and the basis functions with the ECAP and artifact basis are distinct. The basis functions should be scaled to unit power so that relatively large or small inner products do not cause calculation errors during the inversion of H.

[0092] In reality there is noise in the signal that is not modeled by either basis, but the error introduced is minor because the dot product of an independent noise source with any signal is zero for a dot product taken over an infinite time interval. Limiting the dot product to a finite number of samples when computing b propagates some error, but this error is not significant.

[0093] ECAP parameter determination is as follows: A parametric ECAP basis is determined using the recorded signal with initial artifacts removed and any residual baseline subtracted. This signal will be called the "refined record." A correlation mesh is determined by sweeping a series of basis ECAP frequencies and offsets and taking the dot product between the refined record and each parametric basis function.

[0094] In single-ended and differential modes, embodiments of the invention sample 16 linearly spaced frequencies between 800 Hz and 2 kHz and offsets between -7 samples and -1 samples. This frequency and offset range has been found to work well for test signals observed in human subjects, but these ranges may be extended. Extensions should be done with care, as over-extending these ranges may allow parametric ECAP to introduce noise or artifacts. The most positive stationary point of the correlation mesh determines the parameters of the first ECAP basis element. If the measurement is single-ended, this is the only ECAP basis element.

[0095] For differential ECAP measurements, a new correlation mesh is calculated by sampling 16 linearly spaced frequencies between 500Hz and the frequency of the previously determined basis element. The reference is always considered to be farther away from the stimulus than the recording electrode. This allows exploiting human neurophysiology, since ECAP frequency decreases monotonically with recording distance. In a similar situation, the offset is examined between the previous ECAP basis offset and 12 samples. Again, these ranges were empirically chosen to work well with good signals from humans. Rather than using the most positive fixed point of the correlation mesh, the most negative fixed point determines the parameters of the quadratic basis function instead. If there are no negative fixed points, only the linear basis function is utilized.

[0096] Most blind signal separation algorithms assume that the fundamental signals are statistically independent and use statistical signal processing techniques to estimate the fundamental signals. This approach cannot solve the problems of ECAP and artifact estimation, since the fundamental signals are inherently interdependent. Instead, embodiments of the present invention assume that each fundamental signal can be represented as a linear combination of basis functions (a stronger assumption), and limit their application to processes where some knowledge of the fundamental signals is already present before they are recorded in the form of a composite signal.

[0097] Artifact models enumerate the basis functions used to model artifacts present in the hardware / recording. FPAP models are the singular basis functions used within the total ECAP basis set. In practice single-ended measurements use one FPAP, while differential measurements use two FPAPs to address reference electrode effects that occur with differential measurements taken between two recording electrodes.

[0098] An alternative embodiment is also presented, in which the process of FIG.

[0099] An artifact estimation scrubber is a scrubber that attempts to estimate only the artifactual component of the signal, g(t), and uses σ(t)-g(t) to derive the ECAP estimate. An exponential scrubber models the artifacts as a sum of exponential functions. Three such models are considered here:

[0100] [Table 1]

[0101] A nonlinear optimization is performed using a simplex hill-climbing Nelder Mead algorithm, where the parameters a, b, c, d, e, f, g and h are all tuned to minimize the value of a cost function. The nonlinear optimization minimizes the sum of squares error between the estimated artifact samples and the samples of the recorded signal. Mathematically the cost function is defined as:

number

[0102] Nonlinear optimization is a non-deterministic algorithm, meaning that it does not finish in a predictable or predeterminable time. This means that it is possible to feed such a scrubber signals that it cannot scrub within a reasonable time frame. Furthermore, nonlinear optimization may get stuck in a local minimum and never find a true optimal solution. In practice this scrubber works well, but it does have limitations that should be known before using it in general. Nevertheless, such an embodiment has use in certain applications.

[0103] A further embodiment is a fractional pole scrubber that works based on the same principles as the exponential scrubber, where non-linear optimization is used to define the parameters a, k, α and h of the following artifact model: g(t) = αexp(-kt) t 1.0-α +h.

[0104] Yet another embodiment is a complex pole scrubber. If we assume that the artifact is a second-order response (double exponentials are a subset of this type of response), we can estimate the parameters of the second-order response that fits the raw signal. For a discrete signal, the artifact g follows the following model: g[n] = b g[n-1] + c g[n-2].

[0105] Given a set of samples, the matrix equation:

number

number

[0106] Therefore,

number

[0107] The above analysis then feeds into an algorithm called the Signal Quality Indicator (SQI), which assigns a quality score to a set of ECAPs recorded under the same stimulator program. Such an algorithm can be used for the signal quality indicator in clinical data analysis software and clinical user interface software.

[0108] To build a system for automated assessment of signal quality, it is necessary to define the signal characteristics that make a signal "good" versus "bad." Test cases for the "good" versus "bad" range can be used to assess the performance of the SQI. However, no such definition of signal quality exists because it is unclear which characteristics of individual signals lead to poor clinical outcomes in closed-loop spinal cord stimulation. In contrast, it is relatively easy to assess the quality of growth curves, a known indicator of clinical success for closed-loop patients.

[0109] Therefore, the quality of the signal constellation recorded under the same stimulator configuration is defined as a prediction of the quality of the growth curve measured using the same stimulator configuration. However, collecting growth curves takes time. Therefore, an objective guideline that a program will produce satisfactory growth curves before collecting growth curves is required by the clinical engineers in the field.

[0110] A stimulator program is defined as a combination of stimulation waveform parameters, stimulation frequency, and electrode placement. If these quantities are held constant, the signal is measured using the same stimulator program. Since embodiments of the present invention operate under the assumption that the ECAP morphology does not change with stimulation current, and only the peak-to-peak magnitude of the ECAP changes with current, the stimulation current may vary with the signal.

[0111] The Signal Quality Indicator (SQI) of an embodiment of the present invention evaluates the quality of multiple signals recorded with the same stimulator program in open-loop mode (i.e., feedback is not enabled) and outputs a measure of the predicted growth curve quality as a single number between 0 and 1. A higher score indicates that the signal recorded with said program is of higher quality and more suitable for use in growth curve measurements. Since the quality estimate should be robust to individual signals of unique quality, multiple recordings (or signals) are required to make the evaluation. Instead, it is desirable for the SQI to provide an indication of the overall signal quality of the stimulator program.

[0112] FIG. 8 illustrates the architecture of an SQI system according to one embodiment of the present invention.

[0113] It should be noted that alternative embodiments can derive the ECAP signal quality score by reference to a reference ECAP derived by other means. For example, a residual signal can be obtained by subtracting an artifact estimate from the recorded signal and simply compared to a clinically validated template ECAP stored in the device from the time of fitting. The clinically validated template ECAP can include, for example, an ECAP recording obtained significantly above a threshold for improving SNR and verified by a clinician as suitable for storage in the device to serve as such a template.

[0114] Growth curves of various qualities were scored by experienced clinicians for their usefulness in closed-loop SCS treatment. A subset of signals recorded with the same programs used to produce those growth curves was used as the signal quality test library. The performance of the SQI is evaluated by its ability to produce quality scores that give a consistent ranking relative to the quality scores assigned to programs in the signal quality test library. Tuning / learning of the algorithm is not performed on the signal quality test library, but rather on the signal quality training library.

[0115] SQI accepts each input signal as a list of samples. SQI also accepts a stimulation current along with each input signal. SQI uses multiple signals and provides a quality estimate when four or more signals are accepted as inputs. SQI is used to assess program quality, so that high-quality programs can be more easily selected for clinical use.

[0116] Additionally, quality can be assessed by measuring the consistency of the estimated ECAP components of the signal. Inconsistent estimates indicate poor signal quality and therefore poor ECAP estimates, or poor modeling of the signal components as may occur given a degraded signal. The SQI outputs a quality estimate in the form of a decimal number between 0 and 1.

[0117] The intent of the quality indicator is to allow the FCE to more quickly find a good program for the patient without having to rely on experience and developed intuition regarding signal quality. Although presenting multiple outputs may reduce the mental / experiential burden on the FCE, training or developed intuition is still required in aggregating the meaning of multiple indicators. Therefore, it is desirable to provide a single indicator as provided by an embodiment of the present invention.

[0118] The approach of an embodiment of the present invention can be represented by the following pseudocode: score=(scoreParams.DetectionRate * scoreParams.MeanPositiveCorrelation) / ((scoreParams.StdPosFreq+100 * scoreParams.StdPosOff)+30 / (scoreParams.DetectionRate+1e-3)); return 1.0-1.0 / (1.0+alpha * score); / / Convert a score from [0,inf] to a score from [0,1].

[0119] The Signal Quality Indicator (SQI) is a tool used to guide the FCE in selecting programming parameters. The SQI is a number between 0 and 1 that, in combination with the SQI measured across various patient programs, gives insight into which of those programs will perform best. For example, if program A has an SQI of 0.9 and program B has an SQI of 0.5, the clinical engineer would choose program A. In this sense, the SQI can be thought of as a predictor of patient outcome.

[0120] Signal quality can be interpreted in one of two ways: objective and subjective. Objective signal quality is represented by objective signal characteristics such as signal to noise, and no amount of the signal can be removed by processing. Subjective signal quality is a measure of how much information can be extracted from the signal given the capabilities of the implant in use. This subjective signal quality category covers signal features such as signal to artifact ratio. An ideal artifact removal technique that is not limited by processing time and capacity can improve the subjective signal quality, but given the limited filter capabilities available in practical implants and in practical clinical programming sessions, embodiments of the present invention instead predict patient outcomes within the constraints of such applications. Signal quality indicators used in various embodiments of the present invention may include a combination of objective and subjective signal quality. Under the assumption that neurophysiological responses change only in amplitude over time and not in morphology, the subjective SQI takes into account the variability of certain signal features and therefore requires a time sequence of the signal. However, the objective SQI can provide a score based on individual signals.

[0121] The SQI in the described embodiment is derived from a time sequence of signal features. The features utilized are: Detection of ECAP determined by the previously described basal element signal separation mechanism Model parameters estimated by the mechanism described in the previous fitting method Model correlations calculated using the previously described mechanisms ·Stimulation current.

[0122] Given a time sequence of features, a time sequence of derived SQI is determined. Embodiments of the invention provide a signal quality indicator that varies over different time scales. Estimation of the variability of certain signal features requires some sample size before an estimate can be made. Using a small sample size gives a rapidly updating SQI compared to a large sample size. The rapidly updating SQI used by embodiments of the invention is:

number

number

number

[0123] Signal statistics are calculated over 32 samples, and at least 32 signals are required before an initial SQI score, s, can be calculated. After this step, the score, s, is not limited to the specified range of [0;1], but can instead extend to ∞ if the variance of both parameters is 0. Therefore, in the next step, we apply a normalization to s to yield a normalized score as follows:

number

[0124] Now s'∈:[0,1]. The constants γ and τ have been tuned using clinical experience to give maximum differentiation between quality scores in a clinical setting. If these parameters are chosen incorrectly, the scores tend to be inappropriately close to 1 or close to 0 most of the time.

[0125] A slowly updated SQI is also utilized. The advantage of a slowly varying SQI is that the scores are assigned over a long history of the signal and are not overly sensitive to local signal changes. Thus, the clinical engineer has stable scores and is better equipped to select a program compared to an SQI that constantly changes the "best" choice of program based on local signal characteristics. A slowly varying SQI can be obtained by increasing the sample size as above, except that in this embodiment a weighted ensemble average is employed. For every n (=32) samples, s' is calculated. The slowly varying SQI is then derived from the weighted average:

number

number

[0126] In one embodiment, the system is configured such that signal quality is presented for four different patient program alternatives in a clinical setting and each quality score is configured to evolve as new signals are observed. The numeric representation of each quality score is scaled to a percentage between 0 and 100, allowing a clinical engineer to perform a closed-loop control programming procedure and use the SQI predictions to refine patient programs before evaluating clinical effectiveness.

[0127] Alternative embodiments of the present invention may similarly implement an SQI derived from any time series of signal features, including signal-to-artifact ratio (SAR), signal-to-noise ratio (SNR), or frequency domain features such as spectral peak locations, etc. In some embodiments, time series of other device program parameters may also be included in the signal quality estimate.

[0128] Thus, embodiments of the present invention may be particularly helpful in automating as much as possible the programming of the device for individual patients.

[0129] An embodiment of the present invention may be particularly beneficial with respect to neuromodulation that utilizes closed-loop feedback based on observed outcomes such as ECAP amplitude. In such a feedback system, a possible behavior of the loop is that if the ECAP signal is somehow lost (e.g., by significant lead migration or additional noise sources) or the signal-to-noise ratio becomes too low, with the effect of reducing the measured ECAP amplitude (but not necessarily by an actual reduction in recruitment), the system will increase the stimulation current to bring the measured ECAP amplitude back up to the specified target. This may result in over-recruitment. Furthermore, if there is a total loss of the ECAP measure, the feedback loop operates to increase the stimulation current until a maximum current limit or compliance voltage limit is hit. Either of these endpoints may result in some discomfort to the patient and more dorsal column activation than intended. In the opposite case, if the measured ECAP amplitude is somehow higher than the actual recruitment, the current is driven to 0 mA and the patient may not receive any therapy and / or may experience intermittent stimulation, which is often frustrating and uncomfortable. Integrating the ECAP signal quality determination according to the present invention into such a feedback loop can improve the operation of the feedback loop by modifying the loop to reduce or prevent such undesired loop excursions from occurring when the ECAP signal quality is poor. For example, a simple step would be to stop operation of the feedback loop entirely when the ECAP signal quality is below a threshold, and resume operation of the feedback loop when the ECAP signal quality rises above that or another threshold. The patient can be notified of such occurrences.

[0130] FIG. 9 illustrates a clinical system according to one embodiment of the present invention, in which a programming application associated with a clinician user arranges for the neurostimulator to perform an automated scan of all possible recording electrode configurations, thereby obtaining a matrix or set of ECAP quality scores for all possible electrode configurations.

[0131] The automated scanning is also referred to herein as measurement electrode scanning (MES). In particular, in this embodiment, the MES is performed by an implanted device, which allows for faster execution of the automated MES, thereby facilitating clinical matching, and further minimizing the possibility that changes in patient posture may affect comparison results.

[0132] The MES results are visually presented by the programming application to allow the clinician user to view the signal quality indication (SQI) of multiple electrode locations in real time. Specifically, the programming application is configured to visually present the estimated neural response to stimulation as measured in the currently selected stimulation and recording configuration, but also simultaneously present the SQI for multiple alternatives that the clinician may wish to consider.

[0133] Measurement electrode scanning allows ECAP from multiple electrode configurations to be displayed simultaneously. It is intended to help optimize the selection of measurement electrodes and settings. By default, measurement electrode scanning is automatically initiated at the start of stimulation. Measurement electrode scanning is configured for up to four measurement electrode configurations: the configuration selected by the user and three other configurations. The electrodes used for scanning are based on the positions of the stimulation, measurement, and reference electrodes selected in the electrode display window, see FIG. 12.

[0134] The settings for selecting electrodes and performing MES are as follows:

[0135] [Table 2]

[0136] 10 is a state machine diagram illustrating one implementation of MES according to one embodiment of the present invention. Define Primary Location as the location used by the neural stimulator to calculate the neural response to stimulation. Define Necap as the number of measurements used for ECAP averaging. If averaging is disabled, Necap equals 1. Define Nmeasurement as the number of ECAP averages required at a defined location. Define N+ as the location of the ECAP measurement electrode. Define N- as the location of the ECAP reference electrode.

[0137] FIG. 11 is a flow diagram of an MES procedure 1100 performed by the implant. In a first step 1102, an electrode configuration is assigned to the MES program. This may be pre-determined or determined by the user. In 1104, the implant firmware captures ECAP measurements for all electrode configurations associated with the user-selected setup. In an exemplary implementation, considerations while selecting electrodes may be that the selected stimulation, measurement, and reference electrodes must be on the same lead. Or that only one stimulation electrode is selected. In this embodiment, the measurement electrode must not be adjacent to the stimulation electrode and the reference electrode must not be on the case of the implantable pulse generator, although such conditions may be allowed in other embodiments. The measurement electrode must be between the stimulation and reference electrodes. In some cases, the primary location set by the user may not be the best location to capture a good quality ECAP recording. The MES program then suggests the best electrode configuration to obtain a robust ECAP.

[0138] FIG. 12 shows an example of the MES location configuration method when the stim electrode is E2. The MES program is configured to measure ECAP at each of the selected electrode locations until a set number of ECAPs are accumulated. An SQI score is then calculated at each electrode location using the SQI algorithm. The SQI scores are calculated for various electrode locations by the programming software based on strategies such as fixed distance and fixed reference as shown in FIG. 12. The MES program stops upon user intervention or after calculating the scores for all selected electrodes. The ECAP quality scores at multiple electrodes are provided to the user, which allows the user to select the best possible electrode combination that captures the best quality ECAP.

[0139] Examples of the output are shown in Figures 13-16. Figure 13 shows the Measurement Electrode Scan GUI window showing the SQI for a four fixed distance recording electrode configuration. Simple observation can reveal that E3, referenced against E7, is the best recording electrode in this example. Figure 14 shows the output when the MES scan is stopped for any reason.

[0140] Figure 15 shows the output of SQI=0%, which results when no ECAP is detected. Since no ECAP is observed regardless of the choice of reference electrode, these results suggest that E7 is a poor choice of recording electrode. Figure 16 shows the MES output when investigating which reference electrode is optimal when using E4 as the recording electrode.

[0141] It will be appreciated by those skilled in the art that numerous variations and / or modifications may be made to the invention as shown in the specific embodiments without departing from the spirit or scope of the invention as broadly described, and the present embodiments are therefore to be considered in all respects as illustrative and not restrictive or restrictive. [Explanation of symbols]

[0142] 112 Power supply 114 Telemetry Module 116 Controller 120 Patient Settings 122 Control Program 124 Pulse Generator 126 Electrode Selection 128 Amplifiers & ADCs 150 electrodes

Claims

1. 1. A system for automated evaluation of neural response recordings, comprising: a memory storing a set of basis functions including at least one of: (a) a compound action potential basis function; and (b) an artifact basis function; an input for receiving a plurality of neural recordings of electrical activity in neural tissue, the plurality of neural recordings being obtained by repeatedly applying stimulation using a single configuration of stimulation and recording; 1. A processor comprising: decomposing each neural recording using the set of basis functions by determining at least one parameter that characterizes at least one of an evoked compound action potential (ECAP) and an artifact within the neural recording; iteratively determining a plurality of values ​​of the at least one parameter for each of the plurality of neural recordings; determining a spread of the plurality of values; outputting an indication that the quality of the neural response recording is high if the spread is less than a predetermined threshold, and outputting an indication that the quality of the neural response recording is low if the spread is greater than the predetermined threshold. A processor and Including, the system.

2. The system of claim 1 , wherein the indication of the quality of the neural response recording is a binary indication of high or low quality based on a comparison of the diffusion to the predetermined threshold.

3. The system of claim 1 , wherein the indication of the quality of the neural response recording is based on a continuum from high quality to low quality based on a comparison of the diffusion to the predetermined threshold.

4. 4. The system of claim 1, wherein the indication of quality of the neuronal response recording is calibrated by reference to a clinician scoring of a test set of neuronal recordings.

5. 5. The system of claim 1, wherein the processor is further configured to output a separate indication of the quality of neural response recordings obtained for one or more other configurations of stimulation and recording.

6. The system of claim 5 , wherein the processor is further configured to select a stimulation and recording configuration for ongoing treatment by comparing a quality score of each configuration.

7. The system of claim 1 , wherein the spread is calculated as being the standard deviation of the parameter.

8. The system of claim 1 , wherein the spread is calculated as the variance of the parameter.

9. The system of claim 1 , wherein the spread is calculated as the interquartile range of the parameter.

10. The system of claim 1 , wherein the spread is calculated as being a decile range of the parameter.

11. The system of claim 1 , wherein the at least one parameter comprises a correlation between an observed ECAP and a predefined basis function, the predefined basis function comprising an analytically defined compound action potential basis function.

12. The system of claim 1 , wherein the at least one parameter comprises a frequency of an observation ECAP.

13. The system of claim 1 , wherein the at least one parameter comprises a time offset of an observed ECAP relative to a time of the stimulation.

14. 14. The system of claim 1, wherein the basis functions include analytically defined compound action potential basis functions, and the processor is further configured to use a proportion of ECAPs detected within the plurality of neural recordings to determine a quality of the neural response recordings.

15. The system of any one of claims 1 to 14, wherein the processor is further configured to obtain two or more neural recordings of each ECAP and to evaluate the quality of the ECAP using one or more comparison parameters derived by comparing the two or more recordings.

16. The system of claim 15, wherein the comparison parameters include a conduction velocity of each ECAP determined from two or more neural recordings of the ECAP, and a spread of the conduction velocities is used to derive ECAP signal quality.

17. 17. The system of claim 1, wherein a number of parameters are obtained and processed by a predefined function to generate a single quality score.

18. The quality score is as follows: Score = (detection rate * correlation) / (frequency spread + offset spread) It is determined as 18. The system of claim 17, wherein the detection rate is the rate at which ECAPs are detected in the multiple neural recordings, the correlation is a correlation between the observed ECAPs and predefined basis functions including an analytically defined compound action potential basis function, the frequency spread is the frequency spread of the observed ECAPs, and the offset spread is the spread of time offsets of the observed ECAPs relative to the time of stimulation.

19. The system of claim 18, wherein the processor is further configured to normalize the quality score to the range of [0:1].

20. 20. The system of claim 18, wherein the processor is configured to provide the quality score within 250 ms.

21. 1. A method for automated evaluation of neural response recordings, the method being executed by a processor of a system, the method comprising: storing a set of basis functions including at least one compound action potential basis function and at least one artifact basis function; receiving a plurality of neural recordings of electrical activity in neural tissue, the plurality of neural recordings being obtained by repeatedly applying stimulation using a single configuration of stimulation and recording; decomposing each neural recording using the set of basis functions by determining at least one parameter that characterizes at least one of an evoked compound action potential and an artifact within the neural recording; iteratively determining a plurality of values ​​of the at least one parameter for each of the plurality of neural recordings; determining a spread of the plurality of values; outputting an indication that the quality of the neural response recording is high if the spread is less than a predetermined threshold, and outputting an indication that the quality of the neural response recording is low if the spread is greater than the predetermined threshold. A method comprising:

22. 1. A non-transitory computer readable medium for automated evaluation of neural response recordings, comprising: When executed by one or more processors, storing a set of basis functions including at least one compound action potential basis function and at least one artifact basis function; receiving a plurality of neural recordings of electrical activity in neural tissue, the plurality of neural recordings being obtained by repeatedly applying stimulation using a single configuration of stimulation and recording; decomposing each neural recording using the set of basis functions by determining at least one parameter that characterizes at least one of an evoked compound action potential and an artifact within the neural recording; iteratively determining a plurality of values ​​of the at least one parameter for each of the plurality of neural recordings; determining a spread of the plurality of values; outputting an indication that the quality of the neural response recording is high if the spread is less than a predetermined threshold, and outputting an indication that the quality of the neural response recording is low if the spread is greater than the predetermined threshold. A non-transitory computer readable medium comprising instructions for performing a method comprising:

Citation Information

Patent Citations

  • Blind signal separating system, method and program, and its recording medium

    JP2005084860A

  • Methods and systems for automatically identifying whether a neural recording signal includes a neural response signal

    US20060287609A1

  • Stimulus Artifact Removal for Neuronal Recordings

    US20100069996A1

  • Method and Device for Detecting a Neural Response in Neural Measurements

    US20180228391A1