Improved measurement of evoked neural response properties.

Abstract

An implantable device for controllably delivering neural stimulation is disclosed. The device includes a plurality of electrodes, including one or more stimulation electrodes and one or more measurement electrodes; a stimulation source configured to provide neural stimulation delivered via the one or more stimulation electrodes to a patient's neural pathway to elicit a neural response from the neural pathway; a measurement circuit configured to capture a signal window sensed on the neural pathway via the one or more measurement electrodes following each neural stimulation; and a control unit configured to control the stimulation source to provide the neural stimulation according to a stimulation intensity parameter and to measure the strength of the evoked neural response in the captured signal window following the provided neural stimulation by correlating the captured signal window with a template. The template is orthogonal to a predetermined artifact basis that models an artifact component in the captured signal window.

Description

[Technical Field]

[0001] This application claims priority to Australian Provisional Patent Application No. 2022902847, filed on 1 October 2022, the contents of which are incorporated herein by reference in their entirety.

[0002] The present invention relates to implantable spinal cord stimulation, and in particular to improvements in measuring the characteristics of neural responses to stimulation. [Background technology]

[0003] There are a wide variety of situations in which it is desirable to apply neural stimulation to alter neural function, a process known as neuromodulation. For example, neuromodulation is used to treat a variety of disorders, including chronic neuropathic pain, Parkinson's disease, and migraines. Neuromodulation devices apply electrical pulses (stimuli) to nerve tissue (fibers or neurons) to produce a therapeutic effect. Generally, the electrical stimulation generated by a neuromodulation device induces a neural response known as an action potential in the nerve fiber, which can have either an inhibitory or excitatory effect. Inhibitory effects can be used to modulate undesirable processes, such as pain transmission, while excitatory effects can be used to induce desired effects, such as muscle contraction.

[0004] When used to relieve neuropathic pain originating in the trunk and extremities, electrical pulses are applied to the dorsal columns (DCs) of the spinal cord, a procedure called spinal cord stimulation (SCS). Such devices typically include an implantable electrical pulse generator and a power source, such as a battery, which may be rechargeable percutaneously by wireless means such as inductive transfer. An electrode array is connected to the pulse generator and implanted adjacent to the target nerve fiber within the spinal cord, typically in the dorsal epicardial space above the posterior columns. When electrical pulses of sufficient intensity are applied to the target nerve fiber by the stimulating electrode, neurons within the fiber depolarize, resulting in the generation of an action potential within the fiber. Action potentials propagate along the fiber in both orthodromic (for afferent fibers, this refers to the cephalad, i.e., rostral direction) and antidromic (for afferent fibers, this refers to the caudal, i.e., caudal direction) directions. Action potentials propagating along stimulated Aβ (Abeta) fibers in this way inhibit pain transmission from the area of ​​the body (dermatome) innervated by the target nerve fiber to the brain. To maintain the pain relief effect, stimulation is applied repeatedly, for example at a frequency in the range of 30 Hz to 100 Hz.

[0005] For effective and comfortable neuromodulation, stimulation intensity must be maintained above the recruitment threshold. Stimulation below the recruitment threshold fails to recruit a sufficient number of neurons to generate therapeutic action potentials. While responses from a single class of fibers are desired in almost all neuromodulation applications, the stimulation waveforms used may induce action potentials in other classes of fibers, causing unwanted side effects. Therefore, for pain relief, it is desirable to apply stimulation at intensities below the discomfort threshold, because exceeding the discomfort threshold can result in unpleasant or painful sensations due to over-recruitment of Aβ fibers. Excessive recruitment of Aβ fibers can cause unpleasant sensations. High-intensity stimulation may also recruit Aδ (A-delta) fibers, which are sensory nerve fibers associated with acute pain, cold sensation, and warm sensation. Therefore, it is desirable to maintain stimulation intensity within the therapeutic range between the recruitment threshold and the discomfort threshold.

[0006] The task of maintaining adequate neural recruitment is further complicated by electrode migration (changes in position over time) and / or postural changes in the implant recipient (patient), both of which can significantly alter the neural recruitment and, therefore, the therapeutic area resulting from a given stimulation. The epidural space provides a space within which the electrode array can move, and movement of such an array due to movement or changes in posture alters the electrode-to-fiber distance, thereby altering the recruitment efficiency of a given stimulation. Furthermore, the spinal cord itself can move relative to the dura mater within the cerebrospinal fluid (CSF). During posture changes, the volume of CSF and / or the distance between the spinal cord and the electrode can change significantly. This effect can be so great that a previously comfortable and effective stimulation regime can become ineffective or painful simply due to postural changes.

[0007] Attempts to address such issues have been made through feedback or closed-loop control, such as using the method described in the applicant's International Patent Publication No. WO 2012 / 155188. Feedback control attempts to compensate for relative nerve / electrode movement by controlling the intensity of the delivered stimulation to maintain substantially constant neural recruitment. The intensity of the neural response evoked by the stimulation can be used as a feedback variable representing the amount of neural recruitment. A signal representing the neural response can be sensed by measurement electrodes in electrical communication with the recruited nerve fibers and processed to obtain the feedback variable. Based on the response intensity, the intensity of the applied stimulation can be adjusted to maintain the response intensity within a therapeutic range.

[0008] Therefore, it is desirable to accurately measure the strength and other characteristics of the neural response evoked by stimulation. The action potentials generated by depolarization of many fibers due to stimulation are summed to form a measurable signal known as the evoked compound action potential (ECAP). Thus, the ECAP is the sum of the responses from many single-fiber action potentials. The ECAP generated from the depolarization of a group of similar fibers can be measured at the measurement electrode as a positive peak potential, followed by a negative peak, and then a second positive peak. This morphology results from the activation region passing through the measurement electrode as the action potential propagates along each individual fiber.

[0009] A proposed approach for obtaining neural response measurements is described in the applicant's International Patent Publication No. WO2012 / 155183, the contents of which are incorporated herein by reference.

[0010] However, neural response measurement can be a challenging task because the neural response component in the sensed signal typically has a maximum amplitude in the microvolt range. In contrast, the stimulus applied to elicit a response is typically a few volts, manifesting itself in the sensed signal as crosstalk of that magnitude. Furthermore, stimulation typically produces stimulus artifacts, which manifest in the sensed signal as a decaying output of a few millivolts after the stimulus has ended. Because neural responses can occur simultaneously with stimulus crosstalk and / or stimulus artifacts, neural response measurement presents a challenging measurement amplifier design challenge. For example, resolving a 10 μV ECAP with 1 μV resolution in the presence of 5 V of stimulus crosstalk requires an amplifier with a dynamic range of 134 dB, which is impractical for implantable devices. In practice, many nonideal aspects of the circuitry can produce artifacts, which primarily result in time-decaying artifact waveforms of positive or negative polarity, making their identification and removal tedious.

[0011] Evoked neural responses are relatively easy to measure when they appear later than the artifact or when the signal-to-noise ratio is sufficiently high. Because artifacts are often limited to a time window of 1-2 ms after stimulation, measurements of neural responses can be more easily obtained if the neural response is measured after this time window. This is especially true when the distance between the stimulating and measuring electrodes is large (e.g., 60 ms). -1 This is the case for surgical monitoring (>12 cm for nerves conducting at 100 Hz), resulting in propagation times from the stimulation site to the measurement electrode exceeding 2 ms, which is longer than the typical duration of stimulation artifacts.

[0012] However, characterizing responses from the dorsal columns requires high stimulation currents. Similarly, the compact size of any implantable neuromodulation device necessitates close proximity of the stimulating and measuring electrodes to monitor the effects of the applied stimulation. In such situations, the measurement process must directly overcome artifacts.

[0013] One approach to measuring the strength of an evoked neural response is to measure the peak-to-peak amplitude of the sensed signal, assuming that its extrema coincide with the peaks of the neural response. Another approach to measuring strength is to measure its root-mean-square (RMS) amplitude. However, both of these measurements are subject to contamination by artifacts that may be present in the sensed signal. In addition, even if the artifacts can be initially removed from the sensed signal, such intensity measurements have other drawbacks. Peak-to-peak amplitude is highly sensitive to spurious noise, resulting in a noisy measurement with frequent outliers. Because RMS amplitude is nonnegative, it has non-Gaussian noise statistics, even if the sensed signal itself is subject to Gaussian noise. This complicates subsequent processing of the measured RMS amplitude.

[0014] Any discussion of documents, acts, materials, devices, articles or the like which has been included in the present specification is solely for the purpose of providing a context for the present invention and is not to be construed as an admission that any or all of such matters form part of the prior art or were common general knowledge in the art relevant to this invention as they existed prior to the priority date of each claim in this application.

[0015] Throughout this specification the word "comprise" or variations such as "comprises" or "comprising" will be understood to mean 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.

[0016] As used herein, a statement that an element is "at least one of" a list of alternatives is understood to mean that the element may be any one of the listed alternatives, or any combination of two or more of the listed alternatives. [Prior art documents] [Patent documents]

[0017] [Patent Document 1] International Patent Publication No. WO2012 / 155188 [Patent Document 2] International Patent Publication No. WO2012 / 155183 [Patent Document 3] International Patent Publication No. WO2015 / 074121 [Patent Document 4] International Patent Publication No. WO2020 / 124135 Summary of the Invention [Problem to be solved by the invention]

[0018] Disclosed herein are methods and devices configured to measure the strength or other characteristics of evoked neural responses in a manner that is accurate and robust against artifacts. The disclosed techniques use a correlation detector with a template that is the residual of projecting a signal containing a representative ECAP onto an artifact basis. The resulting correlation strength index is orthogonal to any artifact that can be perfectly represented using the artifact basis and is accurate as long as the representative ECAP represents the actual ECAP in the sensed signal. Furthermore, the correlation strength index has noise statistics that reflect the noise statistics of the sensed signal, making it more suitable for subsequent processing required for both programming and operating neuromodulation devices in a closed-loop manner. [Means for solving the problem]

[0019] According to a first aspect of the present technology, an implantable device for controllably delivering neural stimulation is provided. The implantable device includes: a plurality of electrodes, including one or more stimulation electrodes and one or more measurement electrodes; a stimulation source configured to provide neural stimulation delivered via the one or more stimulation electrodes to a patient's neural pathway to elicit a neural response from the neural pathway; a measurement circuit configured to capture a signal window from a signal sensed on the neural pathway via the one or more measurement electrodes following each neural stimulation; and a control unit configured to control the stimulation source to provide the neural stimulation according to a stimulation intensity parameter and to measure the strength of the evoked neural response in the captured signal window following the neural stimulation by correlating the captured signal window with a template. The template is orthogonal to a predetermined artifact basis that models an artifact component in the captured signal window.

[0020] According to a second aspect of the present technology, there is provided an automated method for measuring an evoked response to neural stimulation delivered to a patient's neural pathway. The method includes delivering neural stimulation to the patient's neural pathway to evoke a neural response from the neural pathway, the neural stimulation being delivered according to a stimulation intensity parameter. The method further includes acquiring a signal window sensed on the neural pathway following the delivered neural stimulation, and measuring the strength of the neural response evoked by the delivered neural stimulation within the captured signal window by correlating the captured signal window with a template. The template is orthogonal to a predetermined artifact basis that models artifact components in the captured signal window.

[0021] According to a third aspect of the present technology, a neural stimulation system is provided. The system includes an implantable device for controllably delivering neural stimulation. The device includes a plurality of electrodes, including one or more stimulation electrodes and one or more measurement electrodes; a stimulation source configured to provide neural stimulation delivered to a patient's neural pathway via the one or more stimulation electrodes to elicit a neural response from the neural pathway; a measurement circuit configured to capture a signal window sensed on the neural pathway via the one or more measurement electrodes following each neural stimulation; a control unit configured to control the stimulation source to provide each neural stimulation according to a stimulation intensity parameter; and a processor configured to instruct the control unit to control the stimulation source to provide the neural stimulation according to the stimulation intensity parameter and to measure the strength of the evoked neural response in the captured signal window following the provided neural stimulation by correlating the captured signal window with a template. The template is orthogonal to a predetermined artifact basis that models an artifact component in the captured signal window.

[0022] According to a fourth aspect of the present technology, there is provided a neural stimulation system comprising an implantable device for controllably delivering neural stimulation, the device comprising: a plurality of electrodes including one or more stimulation electrodes and one or more measurement electrodes; a stimulation source configured to provide neural stimulation delivered via the one or more stimulation electrodes to a patient's neural pathway to elicit a neural response from the neural pathway; a measurement circuit configured to capture a signal window from a signal sensed on the neural pathway via the one or more measurement electrodes following each neural stimulation; a control unit configured to control the stimulation source to provide the neural stimulation according to a stimulation intensity parameter and to measure the strength of the evoked neural response in the captured signal window following the neural stimulation by correlating the captured signal window with a template, the template being part of a clinical setting of the implantable device; and a processor configured to derive the template to be orthogonal to a predetermined artifact basis that models an artifact component in the captured signal window and to store the template in a memory of the implantable device as part of the clinical setting of the implantable device.

[0023] According to a fifth aspect of the present technology, there is provided an automated method for programming an implantable neuromodulation device for a patient. The method includes deriving a template orthogonal to a predetermined artifact basis that models artifact components in a signal window captured by the implantable neuromodulation device. The method further includes storing the template in a memory of the implantable neuromodulation device as part of clinical configuration of the implantable neuromodulation device.

[0024] References herein to “estimating,” “determining,” “comparing,” etc., are understood to refer to automated processes performed on data by a processor operating to execute predefined procedures suitable for performing the described estimating, determining, and / or comparing steps. The techniques disclosed herein can be implemented in hardware (e.g., using a digital signal processor, an application-specific integrated circuit (ASIC), or a field-programmable gate array (FPGA)), or software (e.g., using instructions tangibly stored on a non-transitory computer-readable medium to cause a data processing system to perform the steps described herein), or a combination of hardware and software. The disclosed techniques can also be embodied as computer-readable code on a computer-readable medium. A computer-readable medium may include any data storage device capable of storing data which can thereafter be read by a computer system. Examples of computer-readable media include read-only memory (ROM), random-access memory (RAM), magnetic tape, optical data storage devices, flash storage devices, or any other suitable storage device. The computer readable medium can also be distributed over several network-coupled computer systems so that the computer readable code is stored and / or executed in a distributed fashion.

[0025] One or more implementations of the present invention will now be described with reference to the accompanying drawings. [Brief explanation of the drawings]

[0026] [Figure 1] FIG. 1 is a diagram that schematically illustrates an implantable spinal cord stimulator, in accordance with one implementation of the present technology. [Figure 2] FIG. 2 is a block diagram of the stimulator of FIG. 1. [Figure 3] FIG. 2 is a schematic diagram illustrating the interaction of the implantable stimulator device of FIG. 1 with a nerve. [Figure 4a] FIG. 1 illustrates ideal activation plots for a patient undergoing neurostimulation in a given position. [Figure 4b] FIG. 10 shows the variation of activation plots with changes in patient posture. [Figure 5] FIG. 1 is a schematic diagram showing elements and inputs of a closed-loop nerve stimulation (CLNS) system, in accordance with one implementation of the present technology. [Figure 6] FIG. 1 shows typical morphology of electrically evoked compound action potentials (ECAPs) in healthy subjects. [Figure 7] FIG. 2 is a block diagram of a neural stimulation therapy system including the implantable stimulator of FIG. 1 in accordance with one implementation of the present technology. [Figure 8] FIG. 8 is a block diagram illustrating data flow in a neurostimulation therapy system such as the system of FIG. 7. [Figure 9] 6 is a flowchart illustrating a method for deriving a template for a correlation-based ECAP detector, such as the ECAP detector that forms part of the CLNS system of FIG. 5. [Figure 10a] 10 is a graph illustrating the first five basis functions in a CPE artifact basis that may be used in the method of FIG. 9. [Figure 10b] 10 is a graph illustrating the first five basis functions in an SVD artifact basis that may be used in the method of FIG. 9. [Figure 11] 10 is a graph showing the explanatory power of SVD artifact bases obtained from a library of artifact-only sensed signals as a function of the rank of the bases. [Figure 12a] 10b is a graph including a trace representing BURD obtained from the CPE artifact basis of FIG. 10a and a trace representing BURD obtained from the SVD artifact basis of FIG. 10b using the method of FIG. 9; [Figure 12b] 10 is a graph including traces representing BURD obtained using a four-lobe filter in the method of FIG. 9. [Figure 13a]10 is a graph showing a set of activation plots estimated using a four-lobe filter from simulated signal windows captured at different stimulus intensities under different models of the dependence of the artifact on stimulus intensity. [Figure 13b] 13b is a graph containing a set of activation plots estimated from the same simulated signal window as FIG. 13a using a BURD template derived from the SVD artifact basis. [Figure 14] 10 is a flow chart illustrating a method for maintaining a valid BURD template as treatment parameters change. DETAILED DESCRIPTION OF THE INVENTION

[0027] FIG. 1 schematically illustrates a spinal cord stimulator 100 implanted in a patient 108 according to one implementation of the present technology. The stimulator 100 includes an electronic module 110 implanted in a suitable location. In one implementation, the stimulator 100 is implanted in the patient's lower abdomen or posterior superior gluteal region. In other implementations, the electronic module 110 is implanted elsewhere, such as in the flank or subclavian region. The stimulator 100 further includes an electrode array 150 implanted in the epidural space and connected to the module 110 by a suitable lead. The electrode array 150 may include one or more electrodes, such as electrode pads on a paddle lead, a circular (e.g., ring) electrode surrounding a lead body, a conformable electrode, a cuff electrode, a segmented electrode, or any other type of electrode capable of forming unipolar, bipolar, or multipolar electrode configurations for stimulation and measurement. The electrodes may penetrate or be directly fixed to the tissue itself.

[0028] Many aspects of the operation of the implantable stimulator 100 may be programmable by an external computing device 192, which may be operable by a user, such as a clinician or patient 108. Additionally, the implantable stimulator 100 serves as a data collector, with the collected data being communicated to the external device 192 via a transcutaneous communication channel 190. The communication channel 190 may be active substantially continuously, at regular intervals, at irregular intervals, or upon request from the external device 192. Thus, the external device 192 may provide a clinical interface configured to program the implantable stimulator 100 and recover data stored in the implantable stimulator. This configuration is accomplished by program instructions, collectively referred to as a clinical programming application (CPA), stored in an instruction memory of the clinical interface.

[0029] FIG. 2 is a block diagram of the stimulator 100. The electronics module 110 includes a battery 112 and a telemetry module 114. In implementations of the present technology, the telemetry module 114 can use any suitable type of transcutaneous communication channel 190, such as infrared (IR), radio frequency (RF), capacitive, or inductive transfer, to transfer power and / or data to and from the electronics module 110 via the communication channel 190. The module controller 116 has an associated memory 118 that stores one or more of clinical data 120, clinical settings 121, a control program 122, etc. The controller 116 is configured by the control program 122, sometimes referred to as firmware, to control the pulse generator 124 in accordance with the clinical settings 121 to generate stimuli, for example in the form of electrical pulses. The electrode selection module 126 switches the generated pulses to selected electrodes of the electrode array 150 and delivers the pulses to tissue surrounding the selected electrodes. The measurement circuitry 128 may include an amplifier and / or an analog-to-digital converter (ADC) and is configured to process signals including neural responses sensed at measurement electrodes of the electrode array 150 selected by the electrode selection module 126.

[0030] FIG. 3 is a schematic diagram illustrating the interaction of an implantable stimulator 100 with a nerve 180 in a patient 108. In the implementation shown in FIG. 3, the nerve 180 may be located within the spinal cord; however, in alternative implementations, the stimulator 100 may be positioned adjacent to any desired neural tissue, including peripheral nerves, splanchnic nerves, parasympathetic nerves, or brain structures. The electrode selection module 126 selects a stimulating electrode 2 of the electrode array 150 for delivery of a pulse from the pulse generator 124 to the surrounding tissue, including the nerve 180. The pulse may include one or more phases; for example, a biphasic stimulation pulse 160 includes two phases. The electrode selection module 126 also selects a return electrode 4 of the electrode array 150 for return of the stimulation current in each phase, maintaining zero net charge transfer. An electrode may function as both a stimulating electrode and a return electrode throughout the complete multiphasic stimulation pulse. This use of two electrodes for delivery and return of current in each stimulation phase is referred to as bipolar stimulation. In alternative embodiments, other forms of bipolar stimulation may be applied or more stimulating and / or return electrodes may be used. A set of stimulating and return electrodes and their respective polarities is referred to as a stimulating electrode configuration. The electrode selection module 126 is shown connected to ground 130 of the pulse generator 124 to allow return of the stimulation current through the return electrode 4. However, in other implementations, other connections for returning the current may be used.

[0031] Delivery of appropriate stimulation to the nerve 180 via the stimulation electrodes 2 and 4 elicits a neural response 170 including an evoked compound action potential (ECAP), which propagates along the nerve 180 at a speed known as the conduction velocity, as shown. ECAPs may be induced for therapeutic purposes, and in the case of spinal cord stimulators for chronic pain, may produce paresthesia at a desired location. To this end, the stimulation electrodes 2 and 4 are used to periodically deliver stimulation at any therapeutically appropriate frequency, such as 30 Hz, although other frequencies, including frequencies in the kHz range, may also be used. In alternative implementations, stimulation may be delivered in a non-periodic manner, such as in bursts or sporadically, as appropriate for the patient 108. To program the stimulation device 100 for the patient 108, a clinician can have the stimulation device 100 deliver various configurations of stimulation intended to produce sensations experienced by the user as paresthesia. When a stimulation electrode configuration is found that induces paresthesia in a location, of a size that matches the pain-affected area of ​​the patient's body, and of a quality that is comfortable to the patient, the clinician or patient designates that configuration for continued use. The treatment parameters may be loaded into the memory 118 of the stimulator 100 as clinical settings 121.

[0032] FIG. 6 shows a typical morphology of an ECAP 600 from a healthy subject, recorded with a single measurement electrode referenced to system ground 130. The shape and duration of the single-ended ECAP 600 shown in FIG. 6 are predictable because they are the result of ionic currents generated by groups of fibers depolarizing in response to stimulation and generating action potentials (APs). Evoked action potentials (EAPs) generated synchronously across multiple fibers are summed to form the ECAP 600. The ECAP 600 generated from the synchronous depolarization of similar groups of fibers includes a positive peak P1, followed by a negative peak N1, followed by a second positive peak P2. This shape results from the activation region passing through the measurement electrode as the action potential propagates along each individual fiber.

[0033] ECAP may be recorded differentially using two measurement electrodes, as shown in FIG. 3. Differential ECAP measurements are less susceptible to common-mode noise in surrounding tissue than single-ended ECAP measurements. Depending on the polarity of the recording, the differential ECAP may have the opposite shape to that shown in FIG. 6, i.e., two negative peaks N1 and N2 and one positive peak P1. Alternatively, depending on the distance between the two measurement electrodes, the differential ECAP may resemble the time derivative of ECAP 600, or more generally, the difference between ECAP 600 and its time-delayed copy.

[0034] The ECAP 600 may be characterized by any suitable characteristic, some of which are shown in FIG. 6. The amplitude of the positive peak P1 is A p1 and time T p1 The amplitude of the positive peak P2 is A p2 and time T p2 The amplitude of the negative peak P1 is A n1 and time T n1 The peak-to-peak amplitude is A p1 +A n1 The recorded ECAP typically has a maximum peak-to-peak amplitude in the microvolt range and a duration of 2-3 ms.

[0035] Stimulator 100 is further configured to measure the strength of ECAP 170 propagating along nerve 180, whether such ECAP is induced by stimulation from electrodes 2 and 4 or by other factors. To this end, any electrode in array 150 can be selected to function as recording electrode 6 and reference electrode 8 by electrode selection module 126, which selectively connects the selected electrodes to inputs of measurement circuit 128. Thus, signals sensed by measurement electrodes 6 and 8 after each stimulation are passed to measurement circuit 128, which may include a differential amplifier and an analog-to-digital converter (ADC), as shown in FIG. 3 . The recording and reference electrodes are referred to as a measurement electrode configuration. Measurement circuit 128 can operate, for example, according to the teachings of International Patent Publication No. WO 2012 / 155183, discussed above.

[0036] The signals sensed by measurement electrodes 6 and 8 and processed by measurement circuit 128 are further processed by an ECAP detector implemented within controller 116 configured by control program 122 to obtain information regarding the effect of the applied stimulation on nerve 180. In some implementations, the sensed signals are processed by the ECAP detector in a manner that measures and stores one or more characteristics from each evoked neural response or group of evoked neural responses included in the sensed signals. In one such implementation, the characteristics include peak-to-peak ECAP amplitude in microvolts (μV). For example, the sensed signals may be processed by the ECAP detector to determine the peak-to-peak ECAP amplitude according to the teachings of International Patent Publication No. WO 2015 / 074121, the contents of which are incorporated herein by reference. Alternative implementations of the ECAP detector may measure and store other characteristics from the neural responses, or may measure and store more than one characteristic from the neural responses.

[0037] The stimulator 100 may apply stimulation for potentially extended periods of time, such as days, weeks, or months, and during this time may store neural response characteristics, clinical settings, target levels of paresthesia, and other operating parameters in memory 118. To implement appropriate SCS therapy, the stimulator 100 may deliver tens, hundreds, or thousands of stimuli per second for several hours each day. Each neural response or group of responses generates one or more characteristics, such as a measure of the neural response's strength. Thus, the stimulator 100 may generate such data at rates on the order of tens of Hz, hundreds of Hz, or even kHz; over the course of several hours or days, this process may result in large amounts of clinical data 120 that are stored in memory 118. However, the capacity of the memory 118 is necessarily limited, and therefore care must be taken to select a compact data format for storage in memory 118 so that the memory 118 does not become exhausted before the data is expected to be acquired wirelessly by the external device 192, typically no more than once or twice a day.

[0038] An activation plot or growth curve is an approximation of the relationship between stimulation intensity (e.g., the amplitude of the current pulse 160) and the strength of the stimulation-evoked neural response 170 (e.g., ECAP amplitude). FIG. 4a shows an ideal activation plot 402 for a given posture of the patient 108. The activation plot 402 shows a linear increase in ECAP amplitude for stimulation intensity values ​​above a threshold 404, referred to as the ECAP threshold. The ECAP threshold arises because fiber recruitment is binary; that is, if the field strength is too low, no fibers are recruited. However, once the field strength exceeds the threshold, fibers begin to be recruited, and individual evoked action potentials are independent of field strength. Thus, the ECAP threshold 404 reflects the field strength at which a significant number of fibers begin to be recruited, and an increase in response strength at stimulation intensities above the ECAP threshold reflects an increase in the number of recruited fibers. Below the ECAP threshold 404, the ECAP amplitude can be considered zero. Above the ECAP threshold 404, the activation plot 402 has a positive, approximately constant slope, indicating a linear relationship between stimulation intensity and ECAP amplitude. Such a relationship can be modeled as follows:

[0039]

number

[0040] where s is the stimulus intensity, y is the ECAP amplitude, T is the ECAP threshold, and S is the slope of the activation plot (referred to herein as patient sensitivity). The slope S and the ECAP threshold T are the key parameters of the activation plot 402.

[0041] FIG. 4a also shows an discomfort threshold 408, which indicates the stimulation intensity above which the patient 108 experiences an unpleasant or painful stimulus. FIG. 4a also shows a perception threshold 410. The perception threshold 410 corresponds to the ECAP amplitude that is barely perceptible to the patient. Several factors, including the patient's posture, can affect the location of the perception threshold 410. As shown in FIG. 4a, the perception threshold 410 may correspond to a stimulation intensity greater than the ECAP threshold 404 if the patient 108 does not perceive low levels of neural activation. Conversely, the perception threshold 410 may correspond to a stimulation intensity less than the ECAP threshold 404 if the patient has a high perceptual sensitivity to lower levels of neural activation than can be detected by ECAP or if the ECAP signal-to-noise ratio is low.

[0042] For an implantable neuromodulation device such as stimulator 100 to operate effectively and comfortably, it is desirable to maintain the stimulation intensity within a therapeutic range. A stimulation intensity within the therapeutic range 412 is above the ECAP threshold 404 and below the discomfort threshold 408. In principle, it is easy to measure these limits and ensure that a precisely controllable stimulation intensity always falls within the therapeutic range 412. However, the activation plot, and therefore the therapeutic range 412, changes depending on the position of the patient 108.

[0043] Figure 4b shows the changes in activation plots with changes in patient posture. Changes in patient posture can change the impedance of the electrode-tissue interface or the distance between the electrode and the neuron. While Figure 4b only shows activation plots 502, 504, and 506 for three postures, the activation plot for a given posture may lie between or outside the activation plots shown and may vary continuously depending on the posture. As a result, changes in patient posture result in changes in the ECAP threshold, as indicated by the ECAP thresholds 508, 510, and 512 of the activation plots 502, 504, and 506, respectively. Additionally, changes in the slope of the activation plots also result in changes in the patient posture, as indicated by the change in slope of the activation plots 502, 504, and 506. Generally, as the distance between the stimulating electrode and the spinal cord increases, the ECAP threshold increases and the slope of the activation plot decreases. Therefore, activation plots 502, 504, and 506 correspond to an increase in the distance between the stimulating electrode and the spinal cord and a decrease in patient sensitivity.

[0044] To maintain the applied stimulation intensity within a therapeutic range as a patient's posture changes, in some implementations, an implantable neuromodulation device such as stimulator 100 may adjust the applied stimulation intensity based on a feedback variable determined from one or more measured ECAP characteristics. In one implementation, the device can adjust the stimulation intensity to maintain the measured ECAP amplitude at a target response intensity. For example, the device can calculate the error between the target ECAP amplitude and the measured ECAP amplitude and adjust the applied stimulation intensity to reduce the error as much as possible, e.g., by adding the scaled error to the current stimulation intensity. Neuromodulation devices that operate by adjusting the applied stimulation intensity based on measured ECAP characteristics are said to operate in a closed-loop mode and are also referred to as closed-loop neurostimulation (CLNS) devices. By adjusting the applied stimulation intensity to maintain the measured ECAP amplitude at an appropriate target response intensity, such as the target ECAP amplitude 520 shown in FIG. 4b, CLNS devices generally maintain the stimulation intensity within a therapeutic range even as a patient's posture changes.

[0045] The CLNS device includes a stimulator that receives a stimulation intensity value and converts it into a neural stimulus consisting of a series of electrical pulses that follow a predefined stimulation pattern. The stimulation pattern is parameterized by multiple parameters, including stimulation amplitude, pulse width, number of phases, phase sequence, number of stimulation electrode poles (e.g., 2 poles for bipolar, 3 poles for tripolar, etc.), and stimulation rate or frequency. At least one of the stimulation parameters, e.g., stimulation amplitude, is controlled by a feedback loop.

[0046] In one example of a CLNS system, a user (e.g., a patient or clinician) sets a target response strength, and the CLNS device performs proportional-integral-derivative (PID) control. In some implementations, the derivative contribution is ignored, and the CLNS device uses a first-order integral feedback loop. The stimulator generates stimulation according to stimulation strength parameters, which elicit a neural response in the patient. The strength of the evoked neural response (e.g., ECAP) is measured by the CLNS device and compared to the target response strength.

[0047] The measured neural response strength and its deviation from the target response strength are used by a feedback loop to determine possible adjustments to stimulation intensity parameters to maintain the neural response at the target strength. If the target intensity is appropriately selected, the patient will receive consistently comfortable and therapeutic stimulation through postural changes and other perturbations to stimulus / response behavior.

[0048] 5 is a schematic diagram illustrating elements and inputs of a closed-loop nerve stimulation (CLNS) system 300, according to one implementation of the present technology. The system 300 includes a stimulator 312 that converts a stimulation intensity parameter (e.g., stimulation current amplitude) s into a nerve stimulus comprising a series of electrical pulses on stimulation electrodes (not shown in FIG. 5) in conjunction with a set of predefined stimulation parameters. According to one implementation, the predefined stimulation parameters include the number and order of phases, the number of stimulation electrode poles, the pulse width, and the stimulation rate or frequency.

[0049] The generated stimulus traverses from the electrode to the spinal cord, represented in FIG. 5 by dotted box 308. Box 309 represents the stimulation eliciting a neural response y, as previously described. Box 311 represents the induction of an artifact signal a, which depends on the stimulus intensity and other stimulus parameters, as well as the electrical environment of the measurement electrodes. In addition to the artifact a, various sources of measurement noise n may be added to the evoked response y in summing element 313 to form a sensed signal r, including electrical disturbances generated by the body, such as electrical noise from external sources, such as 50 Hz mains power, neural responses evoked by other sources rather than the device, e.g., peripheral sensory input, and electrical noise from EEG, EMG, and measurement circuitry 318.

[0050] Stimulation-induced neural recruitment is affected by mechanical changes, including changes in posture, gait, breathing, heart rate, etc. Mechanical changes can cause changes in impedance or changes in the position and orientation of nerve fibers relative to the electrode array. As discussed above, the strength of the evoked response is a measure of the recruitment of the fibers being stimulated. Generally, the stronger the stimulation, the greater the recruitment and the stronger the evoked response. Evoked responses typically have maximum amplitudes in the microvolt range, while the voltage resulting from the stimulation applied to elicit a response is typically several volts.

[0051] Measurement circuit 318, which may be identical to measurement circuit 128, amplifies sensed signal r (including evoked neural response, artifacts, and measurement noise) and samples the amplified sensed signal r to capture a "signal window" 319 containing a predetermined number of samples of the amplified sensed signal r. ECAP detector 320 processes the signal window 319 and outputs a measured neural response strength d. In one implementation, the neural response strength includes a peak-to-peak ECAP amplitude. The measured response strength d is input to feedback controller 310. Feedback controller 310 includes a comparator 324 that compares the measured response strength d (an example of a feedback variable) with a target ECAP amplitude set by target ECAP controller 304 and provides an indication of the difference between the measured response strength d and the target ECAP amplitude. This difference is an error value e. The error value e is input to feedback controller 310.

[0052] The feedback controller 310 calculates an adjusted stimulation intensity parameter s with the goal of maintaining the measured response strength d equal to the target ECAP amplitude. Thus, the feedback controller 310 adjusts the stimulation intensity parameter s to minimize the error value e. In one implementation, the controller 310 makes appropriate adjustments to the stimulation intensity parameter s using a first-order integral function with a gain element 336 and an integrator 338. In such an implementation, the current stimulation intensity parameter s may be determined by the feedback controller 310 as follows: s=∫Kedt (2)

[0053] where K is the gain of gain element 336 (controller gain). This relationship can also be expressed as: δs=Ke (3)

[0054] where δs is the adjustment to the current stimulus intensity parameter s.

[0055] The target ECAP amplitude is input to feedback controller 310 via target ECAP controller 304. In one embodiment, target ECAP controller 304 provides an indication of a particular target ECAP amplitude. In another embodiment, target ECAP controller 304 provides an indication to increase or decrease the current target ECAP amplitude. Target ECAP controller 304 provides an input to CLNS system 300 through which a patient or clinician can input the target ECAP amplitude or an indication thereof. Target ECAP controller 304 includes a memory in which the target ECAP amplitude is stored, from which the target ECAP amplitude is provided to feedback controller 310.

[0056] The clinical setting controller 302 provides the system 300 with clinical settings, including stimulation parameters for the feedback controller 310 and the stimulator 312 that are not under the control of the feedback controller 310. In one example, the clinical setting controller 302 may be configured to adjust the controller gain K of the feedback controller 310 to adapt the feedback loop to patient sensitivity. The clinical setting controller 302 provides an input to the CLNS system 300, through which the patient or clinician can adjust the clinical settings. The clinical setting controller 302 may include a memory in which the clinical settings are stored and provided to the components of the system 300.

[0057] In some implementations, two clocks (not shown) are used: a stimulus clock running at the stimulus frequency (e.g., 60 Hz) and a sample clock (e.g., running at a 10 kHz sampling frequency) for sampling the sense signal r. Because the ECAP detector 320 is linear, only the stimulus clock affects the dynamics of the CLNS system 300. At the next stimulus clock cycle, the stimulator 312 outputs stimulation according to the adjusted stimulus intensity s. Thus, there is a delay of one stimulus clock cycle before the stimulus intensity is updated according to the error value e.

[0058] 7 is a block diagram of a neurostimulation system 700. The neurostimulation system 700 is centered around a neuromodulation device 710. In one example, the neuromodulation device 710 is implemented as the stimulator 100 of FIG. 1 and may be implanted within a patient (not shown). The neuromodulation device 710 is wirelessly connected to a remote controller (RC) 720. The remote controller 720 is a portable computing device that allows the patient to control stimulation in a home environment by enabling control of the functions of the neuromodulation device 710, including one or more of the following functions: enabling or disabling stimulation, adjusting stimulation intensity or target neural response strength, and selecting a stimulation control program from control programs stored in the neuromodulation device 710.

[0059] The charger 750 is configured to recharge the rechargeable power source of the neuromodulation device 710. While charging is shown as wireless in Figure 7, it may be wired in alternative implementations.

[0060] The neuromodulation device 710 is wirelessly connected to a clinical system transceiver (CST) 730. The wireless connection may be implemented as the transcutaneous communication channel 190 of FIG. 1. The CST 730 may act as an intermediary between the neuromodulation device 710 and a clinical interface (CI) 740 to which the CST 730 is connected. Although a wired connection is shown in FIG. 7, in other implementations, the connection between the CST 730 and the CI 740 is wireless.

[0061] The CI 740 may be implemented as the external computing device 192 of Figure 1. The CI 740 is configured to program the neuromodulation device 710 and recover data stored in the neuromodulation device 710. This configuration is accomplished by program instructions, collectively referred to as a clinical programming application (CPA), stored in the instruction memory of the CI 740.

[0062] FIG. 8 is a block diagram illustrating a data flow 800 for a neural stimulation therapy system, such as the system 700 of FIG. 7, according to one implementation of the present technology. Once implanted within a patient, the neuromodulation device 804 delivers stimulation over a potentially extended period of time, such as weeks or months, and records neural responses, clinical settings, target levels of paresthesia, and other operating parameters, as described below. The neuromodulation device 804 may comprise a closed-loop neural stimulation (CLNS) device, in that the recorded neural responses are used in a feedback configuration to continuously or on-goingly control the clinical settings. To implement appropriate SCS therapy, the neuromodulation device 804 may deliver tens, hundreds, or thousands of stimuli per second for hours each day. A feedback loop can operate over most or all of this time by acquiring a sensed signal after each stimulus, or at least periodically acquiring such a sensed signal. Each sensed signal generates a feedback variable, such as a measure of the amplitude of the evoked neural response, which in turn causes the feedback loop to vary at least one stimulation parameter for the next stimulus. Thus, the neuromodulation device 804 generates such data at rates of tens of Hz, hundreds of Hz, or kHz, and over the course of hours or days, this process generates a large amount of clinical data, unlike traditional neuromodulation devices, such as open-loop SCS devices, which lack the ability to record neural responses.

[0063] Once within range of the receiver, the neuromodulation device 804 transmits data, for example via the telemetry module 114, to a clinical programming application (CPA) 810 installed on a clinical interface. In one implementation, the clinical interface is the CI 740 of FIG. 7. Data is categorized into two main sources: (1) data collected in real time during a programming session, and (2) data downloaded from the stimulator after a period of non-clinical use by the patient. The CPA 810 collects and compiles the data into a clinical data log file 812.

[0064] All clinical data transmitted by the neuromodulation device 804 may be compressed using appropriate data compression techniques before transmission by the telemetry module 114 and / or storage in memory 118 to enable high-resolution storage by the neuromodulation device 804. This high resolution allows the neuromodulation device 804 to provide more data for post-analysis and more detailed data mining regarding events in use. Alternatively, compression allows standard-resolution clinical data to be transmitted more quickly.

[0065] The clinical data log file 812 is manipulated, analyzed, and efficiently presented for on-site diagnosis by a clinical data viewer (CDV) 814, such as a clinician, field clinical engineer (FCE), or the like. The CDV 814 is a software application installed on the clinical interface (CI). In one implementation, the CDV 814 opens one clinical data log file 812 at a time. The CDV 814 is intended to be used to diagnose patient problems and optimize therapy for the patient in the field. The CDV 814 may be configured to provide the user or clinician with a simple, single-view page summary of neuromodulation device usage, therapy output, and errors as soon as the log file is compiled after device connection.

[0066] The clinical data uploader 816 is an application that runs in the background on the CI and uploads files generated by the CPA 810, such as the clinical data log file 812, to the data server. The database loader 822 is a service that runs on the data server and monitors the patient data folder for new files. Once the clinical data log file is uploaded by the clinical data uploader 816, the database loader 822 extracts the data from the file and loads the extracted data into the database 824.

[0067] The data server further includes a data analytics web API 826 that provides data for third-party analysis, such as by an analytics module 832 located remotely from the data server. This technology enables the capture, storage, download, and analysis of large amounts of neuromodulation data to improve patient outcomes in challenging conditions and enables the collection of statistical information across patient populations for future analysis for the purposes of fast, cost-effective, and more accurate troubleshooting and patient status assessment, diagnosing disease etiology, and predicting patient outcomes.

[0068] Assisted Programming System As mentioned above, obtaining patient sensory feedback during programming of closed-loop neurostimulation therapy is important, but intervention by a trained clinical engineer is costly and time-consuming. Therefore, it would be advantageous if patients could program their implanted devices themselves, or with some assistance from a clinician. However, current programming system interfaces are not intuitive and technical in nature, and are generally not suitable for direct patient use. Therefore, there is a need for a CPA that is as intuitive as possible for non-technical users while avoiding patient discomfort. Assisted programming systems (APS) implementations of the present technology are generally configured to meet this need.

[0069] In some implementations, the APS includes two elements: an assist programming module (APM), which forms part of the CPA, and an assist programming firmware (APF), which forms part of the control program 122 executed by the controller 116 of the electronic module 110. Data acquired from the patient is analyzed by the APM to determine the clinical settings of the neural stimulation therapy to be delivered by the stimulator 100. The APF is configured to complement the operation of the APM by delivering specific stimuli to the patient in response to commands issued by the APM to the stimulator 100 via the CST 730, and by transmitting measurements of the neural responses to the delivered stimuli back via the CST 730.

[0070] In other implementations, all processing of the APS according to the present technology is performed by the APF, i.e., data acquired from the patient is not passed to the APM but is analyzed by the controller 116 of the device 710 configured by the APF to determine the clinical settings of the neurostimulation therapy delivered by the stimulator 100.

[0071] In an implementation of the APS in which the APM analyzes data from the patient, the APS instructs the device 710 to capture and transmit a signal window back to the CI 740 via the CST 730. In such an implementation, the device 710 captures the signal window using the measurement circuitry 128, bypasses the ECAP detector 320, and temporarily stores the data representing the raw signal window in memory 118 before transmitting the data representing the captured signal window to the APS for analysis.

[0072] After programming, the APS can load the determined program into the device 710 to control subsequent neurostimulation therapy. In one implementation, the program includes clinical settings 121, also referred to as therapy parameters, which are entered into the neuromodulation device 710 by or stored in the clinical settings controller 302. The patient can then use the remote controller 720, as described above, to control the device 710 to deliver therapy according to the determined program. The determined program can simultaneously, or alternatively, be loaded into the CPA for verification and modification.

[0073] Measurement of neural response properties As described above, according to one implementation of an ECAP detector, the sensed signal can be processed to determine the peak-to-peak ECAP amplitude according to the teachings of International Patent Publication No. WO 2015 / 074121. International Patent Publication No. WO 2015 / 074121 discloses processing the sensed signal using a correlation detector, in which the sensed signal is correlated with a predetermined template called a "four-lobe filter." The correlation value, which can be positive, negative, or zero, is a measure of both the RMS amplitude of the neural response and the degree to which the neural response is in phase with the template. The sign of the correlation represents the phase alignment of the neural response with the template, with a positive sign indicating the two signals are in phase and a negative sign indicating the two signals are out of phase. A correlation value of zero indicates that the two signals are in quadrature, i.e., 90 degrees out of phase. Because phase matching depends on the offset at which the correlation is calculated, International Patent Publication No. WO 2015 / 074121 also discloses a method for selecting an offset that optimally aligns the template and the sensed signal so that the correlation value reflects the RMS amplitude of any neural responses in the sensed signal. This process is effective when the template is morphologically similar to the neural response. The correlation value thus calculated can be scaled by a predetermined scalar to convert the RMS amplitude into a measure of the peak-to-peak amplitude of the neural response.

[0074] The templates are chosen not only to resemble neural responses but also to be partially orthogonal to the artifact component in the sensed signal, which is modeled as the sum of two decaying exponential functions with different time constants, and the correlation measure is therefore somewhat insensitive to artifacts in the sensed signal that satisfy such a model.

[0075] However, due to the lack of perfect orthogonality, some artifact components will "bleed through" into the correlation value. Additionally, regardless of the choice of time constant, this artifact model may be imperfect in that a particular artifact is not adequately modeled as the sum of two decaying exponential functions with different time constants. The template will be even less orthogonal to such artifacts, resulting in additional artifact components "bleeding through" into the correlation value. Additionally, the template will never be perfectly in phase with the neural response components of the sensed signal, no matter how carefully the delays are selected. Both of these effects reduce the accuracy of the correlation value as a measure of the magnitude of the neural response components. The present disclosure derives a template for a correlation-based ECAP detector that is less susceptible to at least one of these effects. Such a template improves the accuracy of neural response measurements under a wider range of artifact conditions, or at least provides an alternative.

[0076] In the following, the term "artifact" may be interpreted to mean "unwanted non-stochastic signals within the captured signal window." An example is the stimulation artifact described above. However, the term also includes other neural responses not related to ECAP, such as EMG and late responses.

[0077] The sensed signal may be represented as a vector y of samples in the captured signal window. The correlation detector is implemented as the dot product of the sensed signal y and the template vector d, returning a correlation value C. C=<y,d> (4)

[0078] The template d is normalized to have unit norm. ||d||=1 (5)

[0079] where the norm of a vector is its RMS amplitude.

[0080] The sensed signal y is expressed as a neural response component y E , artifact component y A , and a noise signal e. y=y E +y A +e (6)

[0081] Since the correlation is linear, the correlation in equation (4) can be expanded as follows: C= <y E ,d>+ <y A ,d>+<e,d> (7)

[0082] The purpose of the detector is to measure the neural response component y E However, the artifact component y A The existence of E Even if the normalized replica (matched filter) of E We can see that || means not equal.

[0083] If the noise vector is an independent and identically distributed (IID) Gaussian noise with mean 0, the inner product of the template d and the noise e will also be an IID Gaussian distribution with mean 0 and the same variance as the noise vector e. That is, the variance Var[e] of the noise vector e is σ 2 If so, the variance of the inner product is also σ 2 which is shown by the following set of identities: Var(Σd[i]e[i])=Σd[i] 2 Var(e[i])=σ 2 Σd[i] 2 =σ 2 (8)

[0084] Artifact component y A is the r basis functions φ i can be reliably modeled as any linear combination of the weights a i basis function φ with i Any artifact component y is weighted as a sum of A For the set of coefficients {a i :i=1,...,r} can be obtained.

[0085]

number

[0086] Artifact basis A={φ i ,i=1,…,r} may be assumed to be orthonormal (if they are not, they may be orthonormalized by applying orthonormality such as Gram-Schmidt to the basis). Therefore, the artifact component y A Coefficient a that models i is the artifact component y A and the basis function φ i It can be obtained by the dot product with a i = <y A ,φ i > (10)

[0087] The optimal template d is<e,d> against <y E ,d> (the signal-to-noise ratio of the correlation detector) and any artifact component y A against <y A , d> is zero, i.e., the template is orthogonal to the artifact basis A. Using the expansion of equation (9), it can be seen that d is the sum of all artifact basis functions φ i If , then orthogonality to the artifact basis is guaranteed. <d,φ i >= 0=1,…,r for all i (11)

[0088] The template d that simply maximizes the signal-to-noise ratio of the correlation detector is

[0089]

number

[0090] In other words, it is a normalized matched filter. If such a template d satisfies Equation (11), the correlation value C is simply the neural response component y E which is the desired output of the correlation detector.

[0091] however, <y E ,φ i ≥ 0, i.e., the pure neural response is rarely orthogonal to the artifact. Therefore, templates d other than matched filters are advantageous.

[0092] Since the artifact basis A is orthonormal, the least squares approximation of the sensed signal y by the artifact basis A is given by the sensed signal y and the basis function φ i It can be found by calculating the dot product with

[0093]

number

[0094] Ordinary least squares (OLS) coefficients

[0095]

number

[0096] represents the projection of the sensed signal y onto the artifact basis A. The unnormalized template

[0097]

number

[0098] Define y as the residual of the projection of the sensed signal y onto the artifact basis A, so that

[0099]

number

[0100] Denormalized Templates

[0101]

number

[0102] is orthogonal to the artifact basis A, i.e.

[0103]

number

[0104] It can be shown that satisfies equation (11).

[0105] So the denormalized template

[0106]

number

[0107] The correlation between y and the sensed signal y' containing any artifacts is simply the neural response component y of the sensed signal y'. E and

[0108]

number

[0109] (ignoring noise). In other words, the unnormalized template

[0110]

number

[0111] is completely insensitive to the artifacts represented by the artifact basis A.

[0112]

number

[0113] Also, denormalized templates

[0114]

number

[0115] A normalized version of o It can also be shown that the correlation detector maximizes the signal-to-noise ratio while still being completely free from artifacts. o is called the "Bespoke Unitary Residual Detector (BURD)."

[0116] Advantageously, in SCS, the morphology of ECAP is fairly constant with stimulation intensity. y E =α(s)E (15) where s is the stimulus intensity parameter, E is the normalized ECAP template, and α(s) is the ECAP component y E Therefore, for all sensed signals y obtained after stimulation with any stimulus intensity parameter s, BURD d o It can be shown that the correlation value C returned by a correlation detector using is given by: C=α(s)C E (16) In the formula, C E teeth, <d o,E> is a constant correlation coefficient equal to d. Thus, the variation of the output of the correlation detector with stimulus intensity is a scaled version of the intensity variation of the evoked neural response with stimulus intensity. This is a useful property of an effective feedback variable. The signal-to-noise ratio of such a correlation detector is d o increases to a degree similar to the ECAP template E, so that the BURD d o It makes sense to derive

[0117] Therefore, the correlation detector template according to the present technique is a non-zero ECAP component y E A representative sensed signal y includes R Applying equation (13) to the obtained template

[0118]

number

[0119] Normalize and BURD d o In one implementation, the representative signal y R may be derived by averaging or accumulating multiple sensed signals known to contain ECAP.

[0120] An ECAP discriminator may be applied to the sensed signal y to determine whether the sensed signal contains ECAP. In one implementation, such an ECAP discriminator is a noise deviation detector (NDD).

[0121] Noise Deviance Detector (NDD) NDD is a statistical detector of the presence of ECAP in a signal window. Prior to operating NDD on the signal window, an "artifact scrubber" may be used to remove artifacts from the signal window. One such artifact scrubber is disclosed in International Patent Publication No. WO2020 / 124135, the entire contents of which are incorporated herein by reference. NDD works by detecting statistically anomalous differences from expected noise in the signal window, which differences are attributable to the presence of ECAP in the signal window.

[0122] Calibration of the NDD may be performed on one or more captured signal windows known to contain no evoked neural responses. In one implementation, such a signal window is a "zero current" signal window captured when no stimulation is applied, with artifacts removed, and can therefore be treated as containing only noise. Calibration involves forming estimates of parameters of a predetermined "noise model" (statistical distribution) from samples within one or more "zero current" signal windows. In one implementation, the noise model is a Gaussian distribution, and the parameters are the mean of the samples.

[0123]

number

[0124] and standard deviation

[0125]

number

[0126] is.

[0127] Once calibrated, the NDD is the number of outliers in a signal window.

[0128]

number

[0129] That is, it can be applied to a signal window by counting the number of samples within the signal window that deviate significantly from the noise model. For a Gaussian noise model, the NDD is calculated by the mean estimate

[0130]

number

[0131] Standard deviation estimate from

[0132]

number

[0133] Number of samples that differ by more than n-fold

[0134]

number

[0135] Count the number of outliers, where n is a small integer.

[0136]

number

[0137] is the signal window average

[0138]

number

[0139] and standard deviation

[0140]

number

[0141] The number of samples expected to occur if the noise consists of only

[0142]

number

[0143] To obtain a metric r that quantifies the proportion of outliers present in the signal window relative to the expected proportion of outliers in the signal window according to the noise model,

[0144]

number

[0145] and

[0146]

number

[0147] The difference between is divided by the number of samples in the signal window, N.

[0148] It can be shown that in a Gaussian noise model, the NDD can estimate the metric r as follows:

[0149]

number

[0150] where Φ is the standard normal cumulative distribution function.

[0151] A negative or zero value of the metric r indicates that the signal window is consistent with the noise model, while a positive value of r indicates a deviation from the noise model, which is attributed to the presence of ECAP within the signal window.

[0152] In one implementation of NDD, n is set to 3. Smaller values ​​of n make the NDD more sensitive, more easily indicating deviations from noise and increasing the rate of Type I errors (false positives), while larger values ​​of n require larger outliers before r indicates deviations from noise and increasing the rate of Type II errors (false negatives).

[0153] In one implementation of NDD, a sigmoid function is applied to the raw metric r to convert the metric r into a quality indicator Q in the interval [0,1]. NDD can be mapped to

[0154]

number

[0155] where γ is a parameter to balance type I and type II errors. NDD has a natural interpretation. NDD < 0.5 corresponds to r ≤ 0, indicating that the signal window is likely to be noise. Conversely, Q NDD >0.5 indicates a deviation from the noise model that is considered representative of ECAP. In one implementation, γ is set to 50.

[0156] FIG. 9 illustrates a template d for a correlation-based ECAP detector, such as ECAP detector 320 that forms part of CLNS system 300. o 9 is a flowchart illustrating a method 900 for deriving (BURD). Method 900 may be performed by an APM-configured CI 740, a controller of an APF-configured device 710, or some combination of the two working in conjunction as described above. The output of method 900 is a correlation-based ECAP detector template d o may be stored as part of the treatment parameters entered into the neuromodulation device 710 by the assisted programming system.

[0157] Method 900 begins at step 910, delivering a stimulus at a fixed stimulus intensity s. Then, at step 915, an ECAP discriminator, such as an NDD, is applied to sensed signal y within the captured signal window to detect whether ECAP is present in sensed signal y. Then, at step 920, the output of the ECAP discriminator is tested. If ECAP was not present in sensed signal y ("N"), method 900 returns to step 910. If ECAP was present in sensed signal y ("Y"), at step 930, sensed signal y is compared to a representative signal y R 9. Optionally, a consistency check may be performed on sensed signal y before performing step 930. If sensed signal y fails the consistency check, step 930 is bypassed. One example of a consistency check is checking whether any samples in sensed signal y are clipped, i.e., whether they have values ​​at both ends of the output range of measurement circuit 318.

[0158] Then, in step 940, sufficient sensed signals including ECAP are generated to generate a representative signal y R If not ("N"), method 900 returns to step 910. If yes ("Y"), in step 950, Equation (13) is applied to the representative signal y using the artifact basis A. R Apply it to the denormalized template

[0159]

number

[0160] Optionally, step 950 first obtains the representative signal y R Divide by the number of accumulated sensing signals to obtain the representative signal y R Finally, in step 960, the unnormalized template

[0161]

number

[0162] to normalize the BURD d used by the correlation-based ECAP detector as part of the treatment program. o Get.

[0163] Artifact Base As mentioned above, BURD d o The artifact basis A is required for the derivation of . BURD d in generating the feedback variables of the ECAP-driven CLNS system o The effectiveness of increases with the effectiveness of the artifact basis A in representing the artifact components in the sensed signal in all their various complexities.

[0164] In one implementation, if the artifact is a stimulation artifact, the artifact basis A is derived from a constant phase element (CPE) model of the electrode-tissue interface. This CPE basis is described in the aforementioned International Patent Publication No. WO2020 / 124135. CPE basis A CPE are the five basis functions φ defined for the biphasic stimulus waveform. i (i=1,…,5). φ1(t)=s(t)-s(t-pw)-s(t-pw-ipg)+s(t-2.pw-ipg) (19) φ2(t)=i(t)-i(t-pw)-i(t-pw-ipg)+i(t-2.pw-ipg) (20) φ3(t)=s(t-pw-ipg)-s(t-2.pw-ipg) (21) φ4(t)=i(t-pw-ipg)-i(t-2.pw-ipg) (22) φ5(t)=1 (23) where pw is the pulse width of each phase of the biphasic stimulation, ipg is the interphase gap between the two phases, and i(t) is the impulse response of the CPE, defined as a decaying exponential function: i(t)=t -α ,t≧0 (24) s(t) is the step response of the CPE and is defined as the integral of the impulse response i(t) over all positive time periods t.

[0165]

number

[0166] The parameter α is a constant that depends on the geometry of the electrode-tissue interface, but in one implementation may be set to 0.364.

[0167] Figure 10a shows the CPE artifact base A CPE Five basis functions φ in i A graph showing the

[0168] As in step 950 of method 900, the representative signal y R To apply Equation (13) to the representative signal y R OLS coefficients of the artifact component of

[0169]

number

[0170] CPE base A CPE is not orthonormal, so the representative signal y R Simply applying equation (12) to

[0171]

number

[0172] However, the coefficient

[0173]

number

[0174] is CPE base ACPE and orthonormalize it to obtain the orthonormalized version of the CPE basis A CPE (12) to the original CPE basis A CPE can be obtained by converting

[0175] In another implementation, an artifact base A suitable for any type of artifact is SVD is the sensed signal y , which does not contain neural response components and contains only artifact components and noise. NR The library X can be obtained from singular value decomposition (SVD) of the library X. SVD is a machine learning technique for obtaining a basis that best represents a given training data set. The library X is a set of unresponsive sensed signals y NR In one implementation, such a library X may be obtained from a database of sensed signals from different patients, such as database 824. The library X may be obtained by extracting sensed signals y from the database that are captured after stimulation with the same pulse width and phase and a stimulus intensity s less than half the ECAP threshold. SVD decomposes the matrix X into a product of three matrices: U, Σ, and V*. The SVD artifact basis A SVD r basis functions φ in i can be obtained as the first r columns of the SVD matrix U corresponding to the largest r singular values ​​of the matrix X. In this way, the sensed signal y NR SVD artifact basis A obtained from library X of SVD The first five (i.e., r=5) basis functions φ in i The graph showing these first five basis functions φ i provides a good approximation of the artifact-only sensed signal in library X, explaining 96% of the power of matrix X. Figure 11 shows the SVD artifact basis A obtained from matrix X. SVD The explanatory power of A SVD as a function of rank r.

[0176] SVD artifact basis A SVD is the patient-specific unresponsive sensed signal y NR This results in an artifact basis of lower rank r with the same explanatory power obtained from a library X of non-responsive signals from a multi-patient population. Since the SNR of the correlation detector derived using Equation (13) tends to decrease as the rank r of the artifact basis increases, the patient-specific SVD artifact basis A SVD may be preferable to ensemble-based SVD artifact bases.

[0177] Furthermore, patient-specific SVD artifact basis A SVD can be adapted over time by iteratively recalculating X from a new library of unresponsive sensed signals captured during the patient's treatment.

[0178] Figure 12a shows the CPE artifact base A CPE Retrieved from BURD d o The graph also includes a trace 1210 showing the SVD artifact basis A SVD Retrieved from BURD d o Also included is trace 1220, which represents the two BURDs.

[0179] In some implementations, as in step 950 of method 900, the representative signal y R Applying equation (13) to BURD d o Instead of obtaining a four-lobe filter d FL By applying equation (13) to o may be derived. The resulting BURD d o is a four lobe filter d FL Such an implementation is a four-lobe filter dFL is known to be effective as a correlation template for measuring ECAP characteristics in the sensed signal, and therefore, a representative signal y containing the ECAP component is R This is advantageous because it is valid as FL Retrieved from BURD d o Included in Figure 12b is a graph containing a trace 1230 representing a BURD component. It can be seen that the BURD trace 1230 is not significantly different in shape from the two BURD traces 1210 and 1220 graphed in Figure 12a. Figure 12b also includes a trace 1240 representing a typical artifact component for comparison.

[0180] Figure 13a includes a graph 1300 plotting response strengths estimated using a four-lobe filter against stimulus strength. These response strengths were obtained by applying the four-lobe filter to simulated signal windows obtained using different models of artifact dependence on stimulus strength. Each artifact model is linear and parameterized by a (slope, intercept) parameter pair, shown in legend 1320. To obtain the simulated signal window for each artifact model, a simulated artifact component, obtained by scaling a representative artifact according to the model and corresponding stimulus strength, was added to the set of raw signal windows captured at each stimulus strength. At stimulus intensities below the threshold, artifact leakage into the estimated response strength is clearly visible in the deviation of the response strength to the left of graph 1300 from the ideal horizontal shape of the activation plot in the subthreshold region, as shown in Figure 4a. A completely artifact-insensitive template would return the same response strength at each stimulus strength, regardless of the magnitude of the added artifact. The response magnitudes in graph 1300 do not have this characteristic and vary widely at a given stimulus intensity. Each dashed line, such as line 1310, represents an estimate of the contribution of an artifact component to the response magnitude, modeled by a linear dependence on stimulus intensity in the subthreshold region. The deviations in the dashed lines indicate significant artifact components leaking through the four-lobe filter.

[0181] Figure 13b shows the BURD template d derived from the SVD artifact basis. o Included in the figure is a graph 1350 plotting response strengths estimated using the BURD template d against stimulus strength in the same simulated signal window used to obtain Figure 13a. oThe SVD artifact basis was itself derived from the simulated artifact components used to generate the simulated signal window, as described above. The much higher similarity between response intensities at each stimulus intensity in graph 1350 compared to graph 1300 is due to the use of the BURD template d compared to the 4-lobe filter. o Similar to Figure 13a, each dashed line, e.g., line 1360, represents an estimate of the contribution of the artifact component to the response strength, modeled by a linear dependence on stimulus intensity in the subthreshold region. The similarity between these dashed lines indicates that the BURD template d o We further demonstrate the robustness of

[0000] to artifacts.

[0182] Quantizing the BURD template BURD template d in ECAP detector 320 o In some implementations of the BURD template o To provide for such an implementation, a BURD template d derived according to method 900 may be used. o The coefficients of may be quantized to a fixed number of bits (eg, 14 bits) before being stored as part of the treatment parameters.

[0183] As mentioned above in equation (23), the CPE basis A CPE The fifth basis function φ5(t) is a constant function (φ5(t)=1). This is the CPE basis A CPE If derived using the BURD template d o means that it must be orthogonal to any constant value. In other words, the BURD template d oThe sum of the coefficients of must be zero. To maintain this property after quantization to a fixed-point representation, a "walk-back" algorithm may be used. The purpose of the walk-back algorithm is to distribute any necessary adjustments to the quantized coefficients among the coefficients, starting with the terminal coefficients and working backwards, until the accumulated adjustments are sufficient to make the sum of the quantized coefficients zero.

[0184] The walkback algorithm works as follows. 1. The delta value is calculated as the sum of the template coefficients. If the delta value is zero, the template has already rejected the constant and no further processing is required. 2. The walkback index is set to the position of the last coefficient in the template. 3. The delta value is subtracted from the coefficient pointed to by the walkback index, unless the result exceeds the fixed-point range of the coefficient, i.e., the subtraction causes the coefficient to fall within [-2 n-1 ,2 n-1 -1], where n is the bit depth of the fixed-point representation. 4. The walkback index decreases by 1. 5. The delta value is updated to the recalculated sum of the template coefficients. 6. The algorithm repeats from step 3 until the delta value is zero.

[0185] Maintaining an effective BURD template BURD template by method 900 o The derivation of depends on the measured ECAP and its geometry. The geometry of the ECAP in turn depends on several treatment parameters. These BURD-related treatment parameters include: Stimulating electrode configuration Measuring electrode configuration Ratio of current value of each stimulation electrode to other stimulation electrodes Pulse Width Phase gap Pulse shape (triphasic, biphasic) and polarity (positive-lead, negative-lead) Sampling Delay Sampling Period Stimulus interval Short-circuit interval Number of stimulus sets in a multiple stimulus set program Position of the applied stimulus set in the multiple stimulus set program (the stimulus set that is the target of ECAP measurement)

[0186] Changing any of these treatment parameters can change the shape of the ECAP, thereby "detuning" the BURD template and potentially causing the ECAP characteristic measurements it returns to be inaccurate.

[0187] New therapy parameters may be received when a new therapy program is sent to the stimulator 100 (before therapy begins) or via a program update (after therapy begins).

[0188] According to one aspect of the present technology, numerical values ​​representing relevant current treatment parameters are calculated and can be compared to numerical values ​​calculated in a similar manner from relevant treatment parameters of any new or modified program. Based on this comparison, the BURD template is re-derived, for example, according to method 900 of Figure 9. This allows an effective BURD template to be maintained as treatment parameters change.

[0189] 14 is a flow chart illustrating a method 1400 for maintaining a valid BURD template as treatment parameters change. Method 1400 may be performed by an APM-configured CI 740, an APF-configured controller of device 710, or some combination of the two working in conjunction as described above. Method 1400 begins when the treatment parameters of a treatment program are determined at the end of a programming session.

[0190] Method 1400 begins at step 1410, where a BURD template is derived from the treatment parameters, e.g., using method 900 described above. Step 1410 is optional and may be omitted if the BURD template already forms part of the treatment parameters, as described above. Step 1420 then involves calculating and storing, e.g., in memory 118 of stimulator 100, hash values ​​of the treatment parameters that depend on the validity of the BURD template, e.g., the BURD-related treatment parameters described above. In one implementation, step 1420 uses a 16-bit cyclic redundancy check (CRC) to calculate the hash values ​​of the BURD-related treatment parameters.

[0191] In the next step 1430, CLNS treatment is initiated using the treatment parameters as described above.

[0192] Step 1440 waits for new treatment parameters (or updates to existing treatment parameters). Once the new treatment parameters are received, step 1450 calculates a hash of the values ​​of the new BURD-related treatment parameters in the new treatment parameters using the same method used in step 1420. Step 1460 then compares the calculated hash with the stored hash to determine if the hashes match. If the hashes do not match ("N"), step 1470 derives a new BURD template using the new treatment parameters, for example, using method 900. Step 1480 then stores the hash of the new treatment parameters, and step 1490 begins CLNS treatment using the new treatment parameters, similar to step 1430. Method 1400 then returns to step 1440. If the hashes match ("Y"), method 1400 proceeds directly to step 1490.

[0193] In another implementation, the hash is calculated as part of the programming session and becomes part of the treatment parameters. In one such implementation, steps 1420 and 1450 are omitted. Instead, the hash is simply obtained from the treatment parameters and compared in step 1460.

[0194] With respect to comparing the treatment parameter values ​​themselves, an advantage of method 1400 is that it is more efficient to compare hashes than to compare each value in the list of burd-related treatment parameters individually. The hash is calculated so that a small change in any one of the burd-related treatment parameter values ​​will result in a detectable change in the hash, even though the hash is represented with far fewer bits than the treatment parameter value itself. However, in step 1460, after a match ("Y") is found, the value of each burd-related treatment parameter can optionally be compared to confirm that there has been no change in the value of such parameter. If such confirmation is not obtained, method 1400 can proceed to step 1470.

[0195] Those skilled in the art will appreciate 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]

[0196] 100 stimulator 108 patients 110 Electronic Module 112 Battery 114 Telemetry Module 116 Controller 118 memory 120 Clinical Data 121 Clinical Setting 122 Control Program 124 Pulse Generator 126 Electrode Selection Module 128 Measurement circuit 130 Grounding 150 electrode array 160 Biphasic Stimulation Pulse 170 Neural Response 180 Nerves 190 communication channels 192 External Devices 300 System 302 Clinical Settings Controller 304 Target ECAP Controller 308 Box 309 Box 310 Feedback Controller Box 311 312 Stimulator 313 elements 318 Measurement circuit 319 Signal Window 320 ECAP detector 324 Comparator 336 Gain Element 338 Integrator 402 Activation Plot 404 ECAP Threshold 408 Discomfort Threshold 410 Perceptual Threshold 412 Treatment Range 502 Activation Plot 504 Activation Plot 506 Activation Plot 508 ECAP Threshold 510 ECAP threshold 512 ECAP threshold 520 Target ECAP Amplitude 600 ECAP 700 Neurostimulation System 710 Neuromodulation devices 720 Remote Controller 730 CST, Clinical System Transceiver 740 CI, Clinical Interface 750 charger 800 Data Flow 804 Neuromodulation devices, neuroregulatory devices 810 CPA 812 Clinical Data Log File 814 CDV 816 Clinical Data Uploader 822 Database Loader 824 databases 826 Data Analysis Web API 832 Analysis Module 900 methods 910 steps 915 steps 920 steps 930 steps 940 steps 950 steps 960 steps 1210 BURD Trace 1220 BURD Trace 1230 BURD TRACE 1240 Trace 1300 graphs 1310 line 1320 Legend 1350 graphs 1360 lines 1400 methods 1410 steps 1420 steps 1430 steps 1440 steps 1450 steps 1460 steps 1470 steps 1480 steps 1490 steps

Claims

1. 1. An implantable device for controllably delivering neural stimulation, comprising: a plurality of electrodes including one or more stimulation electrodes and one or more measurement electrodes; a stimulation source configured to provide neural stimulation delivered via the one or more stimulation electrodes to a neural pathway of the patient to elicit a neural response from the neural pathway; measurement circuitry configured to capture a signal window from a signal sensed on the neural pathway via the one or more measurement electrodes following each neural stimulation; a control unit, controlling the stimulation source to provide neural stimulation according to stimulation intensity parameters; measuring the strength of the evoked neural response in a captured signal window following said neural stimulation by correlating said captured signal window with a template; configured to a control unit, wherein the template is orthogonal to a predetermined artifact basis that models artifact components in the captured signal window; An implantable device comprising:

2. The control unit determining a feedback variable from the measured strength of the evoked neural response; adjusting the stimulation intensity parameter using a feedback controller to maintain the feedback variable at a target value; 10. The implantable device of claim 1, further configured to:

3. 3. The implantable device of claim 1, wherein the control unit is further configured to derive the template based on a comparison of new therapy parameters and therapy parameters at which the neurostimulation was provided.

4. The control unit projecting a representative signal including an evoked neural response component onto the predetermined artifact basis; deriving the template by subtracting the projected representative signal from the representative signal; 4. The implantable device of claim 3, configured to derive the template by

5. 5. The implantable device of claim 4, wherein the control unit is further configured to obtain the representative signal from one or more signal windows captured following each stimulus delivered using the new therapy parameters.

6. 1. An automated method for measuring evoked responses to neural stimulation delivered to a neural pathway of a patient, comprising: delivering neural stimulation to the neural pathway of the patient to elicit a neural response from the neural pathway, the neural stimulation being delivered according to stimulation intensity parameters; capturing a signal window sensed on the neural pathway following the delivered neural stimulation; measuring a strength of a neural response evoked by the delivered neural stimulation within the captured signal window by correlating the captured signal window with a template; Including, The method, wherein the template is orthogonal to a predetermined artifact basis that models artifact components in the captured signal window.

7. determining a feedback variable from the measured strength of the evoked neural response; adjusting the stimulation intensity parameter using the feedback variable to maintain the feedback variable at a target value; 7. The method of claim 6, further comprising:

8. 8. The method of claim 6 or 7, further comprising deriving the template based on a comparison of new treatment parameters and treatment parameters under which the neurostimulation was delivered.

9. deriving the template comprises: projecting a representative signal including an evoked neural response component onto the predetermined artifact basis; deriving the template by subtracting the projected representative signal from the representative signal; 9. The method of claim 8, comprising:

10. 10. The method of claim 9, further comprising obtaining the representative signal from one or more signal windows captured following each stimulus delivered using the new treatment parameters.

11. 1. An implantable device for controllably delivering neurostimulation, comprising: a plurality of electrodes including one or more stimulation electrodes and one or more measurement electrodes; a stimulation source configured to provide neural stimulation delivered via the one or more stimulation electrodes to a neural pathway of a patient to elicit a neural response from the neural pathway; measurement circuitry configured to capture a signal window sensed on the neural pathway via the one or more measurement electrodes following each neural stimulation; a control unit configured to control the stimulation source to provide each neural stimulation in accordance with a stimulation intensity parameter; an implantable device comprising: a processor, instructing the control unit to control the stimulation source to provide neural stimulation according to stimulation intensity parameters; measuring the strength of the evoked neural response in a captured signal window following the provided neural stimulation by correlating the captured signal window with a template; a processor configured to Equipped with The neurostimulation system, wherein the template is orthogonal to a predetermined artifact basis that models artifact components in the captured signal window.

12. the processor: determining a feedback variable from the measured strength of the evoked neural response; adjusting the stimulation intensity parameter using a feedback controller to maintain the feedback variable at a target value; The system of claim 11 , further configured to:

13. 13. The system of claim 11 or 12, wherein the processor is further configured to derive the template based on a comparison of new treatment parameters and treatment parameters under which the neurostimulation was delivered.

14. the processor: projecting a representative signal including an evoked neural response component onto the predetermined artifact basis; deriving the template by subtracting the projected representative signal from the representative signal; 14. The system of claim 13, configured to derive the template by:

15. 15. The system of claim 14, wherein the processor is further configured to obtain the representative signal from one or more signal windows captured following each stimulus delivered using the new treatment parameters.

16. 1. An implantable device for controllably delivering neural stimulation, comprising: a plurality of electrodes including one or more stimulation electrodes and one or more measurement electrodes; a stimulation source configured to provide neural stimulation delivered via the one or more stimulation electrodes to a neural pathway of the patient to elicit a neural response from the neural pathway; measurement circuitry configured to capture a signal window from a signal sensed on the neural pathway via the one or more measurement electrodes following each neural stimulation; a control unit, controlling the stimulation source to provide neural stimulation according to stimulation intensity parameters; a control unit configured to measure a strength of an evoked neural response in a captured signal window following the neural stimulation by correlating the captured signal window with a template, the template being part of a clinical setting of the implantable device; and an implantable device comprising:

1. A processor, comprising: deriving the template to be orthogonal to a predetermined artifact basis that models artifact components in the captured signal window; storing the template in a memory of the implantable device as part of the clinical configuration of the implantable device; a processor configured to A neurostimulation system comprising:

17. the processor: projecting the representative signal including the evoked neural response component onto a predetermined artifact basis; deriving the template by subtracting the projected representative signal from the representative signal; 17. The system of claim 16, configured to derive the template by:

18. 20. The system of claim 17, wherein the processor is further configured to obtain the representative signal by accumulating multiple captured signal windows in which an evoked neural response component is detected.

19. the processor: removing artifact components from the captured signal window; detecting statistically anomalous deviations from an expected noise model in the captured signal window; 20. The system of claim 18, further configured to detect an evoked neural response component in the captured signal window by:

20. 18. The system of claim 17, wherein the representative signal is a four-lobe filter.

21. 21. The system of claim 16, wherein the predetermined artifact basis is a basis derived from a continuous element model of the interface between the electrode and the neural tissue.

22. 21. The system of claim 16, wherein the processor is further configured to derive the predetermined artifact basis from a plurality of captured signal windows that do not include an evoked neural response component.

23. 23. The system of claim 22, wherein the processor is configured to derive the predetermined artifact basis by performing singular value decomposition on the plurality of captured signal windows.

24. 24. The system of claim 22 or 23, wherein the processor is configured to obtain the plurality of captured signal windows by instructing the control unit to control the stimulation source to provide a plurality of neural stimuli according to respective stimulation intensity parameter values ​​below a threshold for eliciting a neural response.

25. 25. The system of any one of claims 22 to 24, wherein the processor is configured to obtain the multiple captured signal windows from multiple patients.

26. 26. The system of claim 16, wherein the processor is further configured to quantize the template into a fixed-point representation so as to maintain orthogonality of the template to a constant value.

27. 1. An automated method for programming an implantable neuromodulation device for a patient, comprising: deriving a template to be orthogonal to a predetermined artifact basis that models artifact components in a signal window captured by the implantable neuromodulation device; storing the template in a memory of the implantable neuromodulation device as part of a clinical setup of the implantable neuromodulation device; A method comprising:

28. deriving the template comprises: projecting a representative signal including a neural response component evoked by neural stimulation provided by the implantable neuromodulation device onto a predetermined artifact basis; deriving a template by subtracting the projected representative signal from the representative signal; 28. The method of claim 27, comprising:

29. 30. The method of claim 28, further comprising obtaining the representative signal by accumulating multiple captured signal windows in which evoked neural response components are detected.

30. removing artifact components from the captured signal window; detecting statistically anomalous deviations from an expected noise model in the captured signal window; 30. The method of claim 29, further comprising detecting an evoked neural response component in the captured signal window by:

31. 29. The method of claim 28, wherein the representative signal is a four-lobe filter.

32. 32. The method of any one of claims 27 to 31, further comprising deriving the predetermined artifact basis from a continuous phase element model of the interface between the electrode and the neural tissue.

33. 32. The method of any one of claims 27 to 31, further comprising deriving the predetermined artifact basis from a plurality of captured signal windows that do not contain an evoked neural response component.

34. 34. The method of claim 33, wherein deriving the predetermined artifact basis comprises performing singular value decomposition on the plurality of captured signal windows.

35. delivering neural stimulation to the patient's neural pathway according to a stimulation intensity parameter value below a threshold for eliciting a neural response; acquiring the signal window sensed on the neural pathway following the delivered neural stimulation; 35. The method of claim 33 or 34, further comprising obtaining the plurality of captured signal windows by:

36. 34. The method of any one of claims 31 to 33, further comprising obtaining the multiple captured signal windows from multiple patients.

37. 37. The method of any one of claims 27 to 36, further comprising quantizing the template into a fixed-point representation so as to maintain orthogonality of the template to a constant value.

Images