Method and Device for Optimizing Spinal Cord Electrical Stimulation Parameters Based on Electromyographic Phase Consistency

Abstract

This invention discloses a method and apparatus for optimizing spinal cord electrical stimulation parameters based on electromyographic phase consistency. The method includes: automatically detecting stimulation segments and establishing a mapping relationship with a set of stimulation parameters; scoring the signal quality and using cross-conditional in-phase regression pruning to suppress stimulation artifacts; extracting phase consistency indices to quantify the synchronization level between electromyographic activity induced by spinal cord electrical stimulation and the stimulation rhythm; constructing a task completion index based on amplitude characteristics; constructing a candidate interval with the maximum value of the task completion index as the center and the product of the transition interval width and the baseline coefficient as the candidate radius; and extracting a continuous and effective stimulation parameter interval after boundary contraction. This invention solves the problems of inconsistent parameter optimization processes, easily underestimated phase consistency indices, and lack of practical parameter intervals in existing technologies, and can objectively and reliably recommend personalized electrical stimulation parameter sweet spots with clinical guidance significance.

Description

Technical Field

[0001] This invention relates to the field of electrical stimulation technology, and in particular to a method and apparatus for optimizing spinal cord electrical stimulation parameters based on electromyographic phase consistency. Background Technology

[0002] For spinal cord injury (SCI), traditional treatments such as drug intervention, physical rehabilitation, and stem cell therapy have shown significant limitations in motor function reconstruction, resulting in a significant bottleneck in clinical efficacy. However, the emerging spinal cord stimulation (SCS) neuromodulation technology has made groundbreaking progress in recent years, opening up new treatment pathways for motor function reconstruction. This technology achieves targeted modulation of motor neural pathways below the level of spinal cord injury through the synergistic effect of epidural electrical stimulation (EES) and intelligent rehabilitation equipment. EES technology activates residual function in the spinal cord neural network by stimulating the proprioceptive input pathway in the dorsal root region of the spinal cord, thus potentially reconstructing lower limb motor function in patients with spinal cord injury.

[0003] Current clinical research has proposed high- and low-frequency composite stimulation, spatiotemporally encoded stimulation, and brain-spinal interface systems, enabling spinal cord injury patients to achieve weight loss with EES assistance and to stand and walk without weight loss, gradually restoring their voluntary motor abilities and daily living skills. However, the clinical efficacy of EES technology varies greatly among individuals, and its mechanism of action remains unclear. A significant reason for this is the lack of standardized stimulation protocols, controversies surrounding stimulation modes and parameters, and technical bottlenecks in optimizing stimulation targets and parameters. Existing spinal cord electrical stimulation techniques lack unified parameter standards, with significant disagreements in the design of core parameters such as stimulation frequency, spatiotemporal patterns, and closed-loop feedback mechanisms. These technical controversies stem from limitations in understanding the underlying theories—there is currently a lack of systematic theoretical support regarding how electrical stimulation modulates the spinal cord interneuron network and how it synergizes with the central and peripheral nervous systems.

[0004] The most crucial aspect of personalized EES configuration lies in determining the stimulation target and selecting the stimulation parameters. Numerous studies have shown that, for electrode contact configuration, only by spatially and precisely activating the spinal cord functional segments controlling the target muscle groups of the lower limbs, while simultaneously reducing the activation of their antagonistic muscle groups, can a certain aspect of motor function in SCI patients be effectively improved. Simultaneously, multi-dimensional stimulation parameters, including frequency, pulse width, and amplitude, significantly influence EES-assisted motor ability. In particular, when the stimulation intensity is too low, patients cannot achieve voluntary movement; when the stimulation intensity is too high, the EES will induce tetanic contraction of the muscles, resulting in passive movement, which also prevents patients from achieving voluntary motor control. Due to the vast number of parameter combinations, manually probing point by point is typically time-consuming, experience-dependent, and has limited repeatability.

[0005] In response to this, previous studies still have significant shortcomings in terms of automated and reproducible parameter exploration processes, mainly in the following aspects: First, the process focuses on outputting a single electrical stimulation parameter, while rehabilitation training tasks usually require fine-tuning of the EES within a certain parameter range; Second, when phase consistency indicators are used to reflect the phase-locked relationship between stimulation rhythm and electromyographic activity, traditional artifact removal methods such as averaging templates may result in lower estimates due to the removal of common-mode components, thus affecting the judgment of the usable parameter range; Third, there is a lack of standardized implementation that integrates "automatic stimulation segmentation - signal quality control - phase transition interval quantification - extraction of continuous effective parameter intervals - structured output" into a unified technical solution. Summary of the Invention

[0006] This invention provides a method and device for optimizing spinal cord electrical stimulation parameters based on electromyographic phase consistency. For patients with spinal cord injury (SCI) whose electromyographic activity is induced by spinal cord electrical stimulation, it can provide each patient with a personalized range of continuous and effective stimulation parameters to improve the motor function reconstruction effect under epidural electrical stimulation (EES).

[0007] A method for optimizing spinal cord electrical stimulation parameters based on electromyographic phase consistency includes the following steps: (1) Obtain electromyographic signals induced by various spinal epidural electrical stimulations and their corresponding stimulation parameter labels, identify stimulation segments one by one using high artifact reference signals, and establish a mapping relationship between stimulation parameters and signal segments; (2) The quality score of the electromyographic signal in each stimulus segment after mapping is calculated, and the stimulus segments below the preset quality threshold are screened out; the cross-conditional in-phase regression pruning method is used to suppress stimulus artifacts and obtain the denoised electromyographic signal. (3) Extract phase consistency index from the denoised electromyographic signal Extract amplitude features and combine them with phase consistency indices. Construct task completion metrics; plot phase consistency metrics across stimulation frequency, electrode configuration, and stimulation intensity dimensions. The curves showing how the stimulus intensity changes define the initial and stable impact thresholds to delineate the width of the transition range. (4) Taking the stimulus intensity corresponding to the maximum value of the task completion index as the center, the product of the transition interval width and the preset basic coefficient as the candidate radius, and expanding the candidate radius to both sides of the center to construct the candidate stimulus parameter interval, shrinking its boundary according to the preset requirements, and identifying the interval that meets the constraints as the continuous effective stimulus parameter interval.

[0008] In step (1), the high artifact reference signal is selected from the synchronization channel signal, the electrophysiological channel signal near the stimulation electrode, the stimulator housing return path monitoring channel signal, or the reference channel signal that presents a high artifact component after preprocessing in the electromyography channel.

[0009] In step (1), the stimulus segments are identified one by one using a high-artifact reference signal, specifically including: For high artifact reference signals, a bandpass filter combined with the Teager-Kaiser energy operator (TKEO) is used to generate a feature envelope. The start and end times of each stimulus segment are accurately located through adaptive thresholding and peak detection.

[0010] In step (2), the electromyographic signal in each stimulated segment after mapping is scored for quality. The scoring indicators include at least two of the following: signal saturation ratio, number of data drop points, proportion of signal flat segments, and baseline drift. The normalized quality score is obtained by weighted fusion.

[0011] In step (2), the cross-conditional in-phase regression pruning method is used to suppress stimulus artifacts, specifically including: The actual stimulation period of each stimulation segment is accurately estimated using a high-artifact reference signal; For multiple stimulation segments with different stimulation intensities under the same frequency and electrode configuration, all electromyographic signals are phase-aligned according to the estimated actual stimulation cycle; Establish a linear regression model, where the intercept term represents the common mode artifact component and the coefficient term represents the differential response component related to the change in stimulus intensity. The artifact-suppressed electromyographic signal is obtained by subtracting the common-mode artifact component estimated by the model from the original electromyographic signal.

[0012] In step (3), phase consistency indices are extracted from the denoised electromyographic signals. Specifically, it includes: Narrow-band filtering is applied to the denoised electromyographic signals near the stimulation frequency. The instantaneous phase of the filtered electromyographic signal is obtained by Hilbert transform; Construct a reference signal and obtain its instantaneous phase; The circular statistic of the instantaneous phase difference between the electromyographic signal and the reference signal is calculated as an indicator of stimulus-locked phase consistency. .

[0013] The reference signal is either a sinusoidal reference signal consistent with the stimulation frequency, or a reference phase signal obtained by phase alignment and filtering of a high artifact reference signal.

[0014] In step (3), the extracted amplitude features include at least one of the following: average rectified value (ARV), root mean square (RMS), integrated electromyography (iEMG), band energy, short-time energy, and time-frequency energy.

[0015] In step (3), the stimulus initiation effect threshold and the stable effect threshold are defined to define the width of the transition interval, specifically including: Two thresholds are defined on the curve, and the stimulus initiation affects the threshold. This indicates that the stimulus begins to produce a phase-locking effect on electromyographic activity, stabilizing and influencing the threshold. This represents the establishment of stable synergy between the target muscle and the stimulus; Determine the phase lock value minimum strength ,as well as minimum strength ,Will Defined as the width of the transition interval.

[0016] A spinal cord stimulation parameter optimization device based on electromyographic phase consistency includes a memory and one or more processors. The memory stores executable code, and when the one or more processors execute the executable code, they are used to implement the above-mentioned spinal cord stimulation parameter optimization method.

[0017] Compared with the prior art, the present invention has the following beneficial effects: 1. This invention constructs a standardized signal processing workflow to ensure the reproducibility and reliability of the results. Addressing the challenges of strong artifacts, numerous interferences, and low signal-to-noise ratio in electromyographic signals induced by spinal cord electrical stimulation, this invention achieves accurate detection and parameter mapping of the stimulation segment through a high-artifact reference signal, automatically filters out low-quality data through multi-dimensional signal quality assessment, and preferentially employs a cross-conditional in-phase regression pruning method to suppress stimulation artifacts. This workflow not only effectively solves the technical problem of underestimated phase consistency indices caused by traditional artifact suppression methods, but also standardizes and structures the entire process from raw data to parameter recommendation.

[0018] 2. This invention introduces a phase consistency index to achieve a refined assessment of the effects of neuromodulation. Most existing technologies rely solely on the amplitude characteristics of electromyographic signals to evaluate stimulation effects, making it difficult to distinguish whether muscle activity is a passive response directly driven by the stimulus or a synergistic effect of the patient's active control and the stimulus rhythm. This invention extracts and locks the phase consistency index to address this issue. Indicators are used to quantify the synchronization level between stimulation-induced electromyographic signals and stimulation rhythms, which can accurately identify the patient's ability to autonomously coordinate and regulate stimulation rhythms, providing an objective basis for screening stimulation parameters that can induce active movement.

[0019] 3. Unlike traditional methods that only output a single optimal parameter point, this invention centers on the stimulus intensity corresponding to the maximum value of the task completion indicator. It uses the product of the transition interval width and the baseline coefficient to determine the radius of the candidate interval, directly transforming the physiological characteristics of the neural recruitment gradient into geometric constraints for parameter search. After verification by the task completion indicator and boundary contraction, a continuous and effective interval with tolerance for parameter fine-tuning is extracted, thus defining the sweet spot for voluntary motor control. This design fully considers the practical needs in clinical rehabilitation training—parameters need to be dynamically adjusted within a certain range to adapt to scenarios such as fatigue and changes in body position. The output parameter interval not only ensures the stability of the training effect but also significantly reduces the workload and trial-and-error costs of manually searching for stimulus parameters in clinical practice. Attached Figure Description

[0020] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0021] Figure 1 This is a flowchart illustrating the overall process of a spinal cord electrical stimulation parameter optimization method based on electromyographic phase consistency, according to an embodiment of the present invention.

[0022] Figure 2 This is a schematic diagram of multi-channel raw stimulation-induced electromyography signals recorded during multi-parameter scanning of an SCI patient in an embodiment of the present invention.

[0023] Figure 3 This is a schematic diagram of the process for identifying stimulus segments one by one using a high-artifact reference signal in an embodiment of the present invention.

[0024] Figure 4 This is a comparison chart of the artifact suppression effects of cross-conditional in-phase regression in embodiments of the present invention.

[0025] Figure 5 for A graph showing the changes in amplitude characteristics and task completion indicators with normalized stimulus intensity. Detailed Implementation

[0026] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0027] It should be noted that, unless otherwise specified, the features in the following embodiments and implementation methods can be combined with each other.

[0028] like Figure 1 As shown, a method for optimizing spinal cord electrical stimulation parameters based on electromyographic phase consistency comprises three sequentially connected modules: a signal processing and parameter mapping module, a quality control and artifact suppression module, and a feature extraction and parameter bounding module. The signal processing and parameter mapping module receives multi-channel stimulation-evoked electromyographic signals and their stimulation parameter configuration files. It automatically detects stimulation segments using the TKEO energy operator and adaptive thresholds, and sequentially maps the detected stimulation segments to parameter group indices. The quality control and artifact suppression module performs multi-dimensional signal quality scoring on each mapped stimulation segment and filters out low-quality segments. Subsequently, it employs a cross-conditional in-phase regression pruning method to suppress stimulation artifacts, outputting a denoised electromyographic signal. The feature extraction and parameter bounding module extracts features from the denoised signal. The system uses indicators and amplitude characteristics to construct task completion indicators. It determines the radius of candidate intervals by using the transition interval width and the base coefficient, and constructs candidate intervals centered at the maximum value of the task completion indicator. After boundary shrinkage, it outputs a continuous effective stimulus parameter interval and a structured recommendation configuration file. The specific steps are as follows: S1, Signal Processing and Parameter Mapping Module.

[0029] Electromyographic signals of stimulation-induced electromyographic activity obtained under various spinal epidural electrical stimulation configurations and their corresponding stimulation parameter labels were acquired. Stimulation segments were automatically identified one by one using a high artifact reference signal, and corresponding stimulation parameter group information was assigned to the detected stimulation segments to establish a mapping relationship between "stimulation parameters and signal segments".

[0030] The high artifact reference signal is a specially designed synchronous trigger channel, an electrophysiological recording channel close to the stimulation electrode, or other signals that can clearly capture the periodic characteristics of the stimulation artifact. Its purpose is to accurately locate the moment of stimulation.

[0031] Furthermore, the stimulation scan parameter configuration file bound to the current patient's identity is obtained. This file contains all the preset stimulation parameter groups for the current test. Each parameter group is a specific combination of stimulation parameters (e.g., electrode configuration {E3-, E5+}, frequency 40Hz, intensity 2.5mA) and carries unique index information (such as segment number).

[0032] Furthermore, each stimulus segment of the high-artifact reference signal is automatically identified. Specifically, a bandpass filter combined with the Teager-Kaiser Energy Operator (TKEO) method is used to generate the feature envelope. The formula for calculating TKEO in the discrete-time signal is as follows: ; In the formula, Indicates the current time (the nth time) The signal amplitude corresponding to each sampling point, Indicates that the TKEO operator is in The output value of a point is the instantaneous energy of the signal.

[0033] Furthermore, by using adaptive thresholding and peak detection, the start and end times of each stimulation segment are precisely located.

[0034] Based on this, according to the chronological order or specific identifier of the stimulus segments, the data is compared with the parameter group index information in the parameter scanning configuration. Each detected stimulus segment is assigned a corresponding stimulus parameter group, forming a "stimulus parameter-signal segment" mapping relationship. If the number of detected segments does not match the number of parameter groups, consistency processing such as truncation, interpolation, or timestamp alignment is performed, and a corresponding mapping consistency flag is output as a reference for subsequent analysis.

[0035] S2, Quality Control and Artifact Suppression Module.

[0036] The quality of the electromyographic signals in each mapped stimulation segment is scored, and stimulation segments below the preset quality threshold are removed. The cross-conditional in-phase regression pruning method is used to suppress stimulation artifacts and obtain the denoised electromyographic signals.

[0037] The scoring metrics include, but are not limited to: signal saturation ratio Number of data drops Signal flat segment ratio and the degree of baseline drift And so on. By integrating these indicators into a normalized quality score. And set a threshold. (e.g., 0.7) Automatically filter out low-quality stimulus segments or the entire channel below the threshold, and finally obtain a high-quality set of available stimulus segments.

[0038] Normalized mass fraction The calculation formula is: ; + +...+ =1; In the formula, , ,..., The weighting coefficients represent the weights of each scoring indicator.

[0039] Furthermore, an optimized cross-conditional in-phase regression pruning technique is employed to suppress stimulus artifacts. First, the actual stimulation period of each stimulation segment is accurately estimated using a high-artifact reference signal, providing a benchmark for subsequent alignment. Second, for multiple stimulation segments with different stimulation intensities at the same frequency and electrode configuration, all signals are phase-aligned according to the estimated stimulation period. Finally, a linear regression model is established. The intercept term of the model represents the common-mode artifact component present at all stimulation intensities, while the coefficient term represents the differential response component related to changes in stimulation intensity, i.e., the changes in the actual electromyographic signal caused by neural modulation.

[0040] Furthermore, by subtracting the model-estimated common-mode artifact components from the original signal, the artifact-suppressed electromyographic signal can be obtained. This method preserves the true physiological response to the greatest extent possible, avoiding the underestimation of phase consistency indices that may occur with traditional template subtraction methods.

[0041] S3, Feature Extraction and Parameter Bounding Module.

[0042] First, calculate the amplitude correlation characteristics for each stimulus segment, including the average rectified value. and root mean square These are used to quantify the intensity of muscle activation, and their respective calculation formulas are as follows: ; ; In the formula, Indicates the first The signal amplitude at each sampling point This indicates the total number of sampling points within the current stimulus segment.

[0043] Furthermore, the stimulus-locked phase consistency index was extracted. First, the signal is narrowband filtered near the stimulation frequency. Then, its instantaneous phase is obtained through Hilbert transform, as shown in the formula: ; In the formula, This represents the electromyographic signal after processing by the artifact suppression module and narrowband filtering. For Hilbert transform operators, This represents the transformed orthogonal imaginary part of the signal.

[0044] Furthermore, an ideal sine wave with the same stimulation frequency is constructed as a reference signal, and its instantaneous phase is also obtained. Finally, the phase consistency index between the electromyographic signal phase and the reference signal phase is calculated. Its formula is: ; In the formula, This indicates the total number of sampling points within the current stimulus segment. Represents electromyographic signals in The instantaneous phase difference at a given moment.

[0045] The phase coherence index ranges from 0 to 1. The closer the value is to 1, the more strictly the electromyographic signal is locked to the stimulation rhythm, reflecting the degree of coordination between the spinal cord's rhythmic response to stimulation and the patient's voluntary control and regulation ability.

[0046] Furthermore, with a fixed stimulation frequency and electrode configuration, the phase-locked value was plotted as a function of stimulation intensity. A changing curve. Two thresholds are defined on this curve: the first threshold... (This represents the initial phase-locking effect of the stimulus on electromyographic activity that can be detected) and the second threshold. (This represents the establishment of stable coordination between the target muscle and the stimulus). Determine the phase-lock value. minimum strength ,as well as minimum strength ,Will Defined as the transition interval width, this width quantifies the range of stimulus intensity required from the onset of autonomous coordination to its stable establishment, and is an important indicator for assessing the tolerance of parameter regulation.

[0047] Furthermore, a task completion index F(I) is constructed by combining amplitude characteristics and phase consistency characteristics. The specific construction methods of the task completion index include, but are not limited to, a weighted combination of the amplitude response and phase lock value of the target muscle group, a functional score based on kinematic measurements, or other composite assessment methods.

[0048] Furthermore, determine the task completion indicators. The stimulation intensity corresponding to the maximum value under the current stimulation frequency and electrode configuration. This is used as the center of the candidate stimulus parameter interval. The transition interval width is... With preset base coefficient The product determines the radius of the candidate interval. ,by Expanding from the origin to both sides Construct candidate stimulus parameter ranges Basic coefficient This is a scaling factor pre-set based on the characteristics of the target motor task and clinical calibration data. Its function is to map the width of the transition interval representing the neural recruitment gradient to a clinically operable range of stimulation parameter adjustment.

[0049] Furthermore, it was verified that F(I) at each stimulus intensity within the candidate interval satisfies the preset task completion threshold. If the boundary of the candidate interval is not satisfied, the boundary is contracted along the corresponding direction until the constraint is met, thus obtaining the continuous effective stimulus parameter interval. The continuous effective stimulus parameter interval is obtained by contracting the threshold value with the maximum value of the task completion index as the center. The task completion index under each stimulus intensity within it has met the preset requirements, thereby ensuring the reliable completion of the motor task when fine-tuning the parameters within this interval in clinical applications.

[0050] Furthermore, a structured recommendation configuration file is generated. This file includes, but is not limited to: SCI patient identification, recommended stimulation frequency and electrode configuration, and the start and end range of continuous effective stimulation intensity. Recommended values ​​within the range (preferably the stimulus intensity corresponding to the maximum value of the task completion indicator). ), the width of the transition interval obtained by quantization The basic coefficients used and candidate interval radius The system provides objective, reliable, and reproducible parameter guidance for clinical rehabilitation by assessing the confidence level and various quality indicators of the current treatment process.

[0051] like Figure 2 As shown, taking an AIS-C grade SCI patient as an example, this paper presents the raw multi-channel electromyography (EMG) signals recorded during multi-parameter scanning of an SCI patient. The top row shows the channels of the left and right iliopsoas muscles (IL), and the bottom row shows the channels of the left and right gastrocnemius medial head muscles (Mrk). The horizontal axis represents time, showing the change in signal amplitude with stimulation intensity within different stimulation segments. Under fixed electrode configuration (E3-, shell+) and fixed frequency (20Hz), the stimulation intensity was scanned at equal intervals, with each intensity lasting for 2 seconds and an interval of 0.5 seconds. The figure shows the raw signals of four channels, including the bilateral iliopsoas muscles and the high artifact channel. It can be seen that as the stimulation intensity increases, the signal amplitude of each channel gradually increases, and obvious stimulation artifacts appear in each intensity segment.

[0052] like Figure 3As shown, the feature envelope is obtained by applying bandpass filtering and TKEO operator to the high artifact reference channel (top figure). The start and end times of each stimulus segment are determined by adaptive threshold. After post-processing such as merging adjacent segments and removing excessively short segments, the final stimulus segment label is obtained (bottom figure) and is consistently mapped with the preset parameter group.

[0053] like Figure 4 As shown, the cross-conditional in-phase regression pruning method was used to suppress artifacts. The left column shows the original signal, and the middle column shows the signal after artifact removal. The variance reduction ratio indicates that the artifact energy was effectively removed. The power spectral density comparison in the right column shows that the sharp peaks at the stimulation frequency and its harmonics were significantly suppressed after processing, while the signal components in the physiological frequency band were preserved.

[0054] like Figure 5 As shown, three independent curves were plotted simultaneously on the normalized stimulus intensity axis: The indicator (solid line) rises sharply in an S-shape before reaching a saturation plateau of approximately 0.80, reflecting the degree of phase lock of the nerve to the stimulus rhythm; normalized mean rectified value The (dotted line) rises to a saturation level of approximately 0.97 with varying slopes and inflection points, reflecting the overall degree of muscle activation; task completion indicator (Dashed line) It exhibits a bell-shaped distribution, reaching its peak near moderate stimulus intensity, and then decreasing due to tetanic contractions or non-target muscle contractions caused by overstimulation. the following.

[0055] Figure 5 The paper also demonstrates the process of determining the interval of continuously effective stimulus parameters. First, based on... On the curve and The stimulus intensity corresponding to the two thresholds determines the width of the transition interval. Set the base coefficient Task completion threshold Then, based on task completion metrics... The stimulus intensity corresponding to the maximum value Centered on, with Construct candidate intervals for radius ; The area is Candidate interval Completely falls within its interior and all sampling points satisfy Without the need for boundary contraction, it can be directly identified as a continuous effective stimulus parameter range. (Rounded down according to the sampling grid). The figure shows... The region (the widest light gray region) is significantly larger than the continuous effective interval (the diagonally filled region), reflecting the design intention of this method to achieve conservative selection through the constraint of the width of the transition interval.

[0056] The embodiments described above provide a detailed explanation of the technical solutions and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are not intended to limit the present invention. Any modifications, additions, and equivalent substitutions made within the scope of the principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for optimizing spinal cord electrical stimulation parameters based on electromyographic phase consistency, characterized in that, Includes the following steps: (1) Obtain electromyographic signals induced by various spinal epidural electrical stimulations and their corresponding stimulation parameter labels, identify stimulation segments one by one using high artifact reference signals, and establish a mapping relationship between stimulation parameters and signal segments; (2) The quality score of the electromyographic signal in each stimulus segment after mapping is calculated, and the stimulus segments below the preset quality threshold are screened out; the cross-conditional in-phase regression pruning method is used to suppress stimulus artifacts and obtain the denoised electromyographic signal. (3) Extract phase consistency index from the denoised electromyographic signal Extract amplitude features and combine them with phase consistency indices. Construct task completion metrics; plot phase consistency metrics across stimulation frequency, electrode configuration, and stimulation intensity dimensions. The curves showing how the stimulus intensity changes define the initial and stable impact thresholds to delineate the width of the transition range. (4) Taking the stimulus intensity corresponding to the maximum value of the task completion index as the center, the product of the transition interval width and the preset basic coefficient as the candidate radius, and expanding the candidate radius to both sides of the center to construct the candidate stimulus parameter interval, shrinking its boundary according to the preset requirements, and identifying the interval that meets the constraints as the continuous effective stimulus parameter interval.

2. The method for optimizing spinal cord electrical stimulation parameters based on electromyographic phase consistency according to claim 1, characterized in that, In step (1), the high artifact reference signal is selected from the synchronization channel signal, the electrophysiological channel signal near the stimulation electrode, the stimulator housing return path monitoring channel signal, or the reference channel signal that presents a high artifact component after preprocessing in the electromyography channel.

3. The method for optimizing spinal cord electrical stimulation parameters based on electromyographic phase consistency according to claim 1, characterized in that, In step (1), the stimulus segments are identified one by one using a high-artifact reference signal, specifically including: For high artifact reference signals, a bandpass filter combined with the Teager-Kaiser energy operator (TKEO) is used to generate a feature envelope. The start and end times of each stimulus segment are accurately located through adaptive thresholding and peak detection.

4. The method for optimizing spinal cord electrical stimulation parameters based on electromyographic phase consistency according to claim 1, characterized in that, In step (2), the electromyographic signal in each stimulated segment after mapping is scored for quality. The scoring indicators include at least two of the following: signal saturation ratio, number of data drop points, proportion of signal flat segments, and baseline drift. The normalized quality score is obtained by weighted fusion.

5. The method for optimizing spinal cord electrical stimulation parameters based on electromyographic phase consistency according to claim 1, characterized in that, In step (2), the cross-conditional in-phase regression pruning method is used to suppress stimulus artifacts, specifically including: The actual stimulation period of each stimulation segment is accurately estimated using a high-artifact reference signal; For multiple stimulation segments with different stimulation intensities under the same frequency and electrode configuration, all electromyographic signals are phase-aligned according to the estimated actual stimulation cycle; Establish a linear regression model, where the intercept term represents the common mode artifact component and the coefficient term represents the differential response component related to the change in stimulus intensity. The artifact-suppressed electromyographic signal is obtained by subtracting the common-mode artifact component estimated by the model from the original electromyographic signal.

6. The method for optimizing spinal cord electrical stimulation parameters based on electromyographic phase consistency according to claim 1, characterized in that, In step (3), phase consistency indices are extracted from the denoised electromyographic signals. Specifically, it includes: Narrow-band filtering is applied to the denoised electromyographic signals near the stimulation frequency. The instantaneous phase of the filtered electromyographic signal is obtained by Hilbert transform; Construct a reference signal and obtain its instantaneous phase; The circular statistic of the instantaneous phase difference between the electromyographic signal and the reference signal is calculated as an indicator of stimulus-locked phase consistency. .

7. The method for optimizing spinal cord electrical stimulation parameters based on electromyographic phase consistency according to claim 6, characterized in that, The reference signal is either a sinusoidal reference signal consistent with the stimulus frequency, or a reference phase signal obtained by phase alignment and filtering of a high artifact reference signal.

8. The method for optimizing spinal cord electrical stimulation parameters based on electromyographic phase consistency according to claim 1, characterized in that, In step (3), the extracted amplitude features include at least one of the following: average rectified value (ARV), root mean square (RMS), integrated electromyography (iEMG), band energy, short-time energy, and time-frequency energy.

9. The method for optimizing spinal cord electrical stimulation parameters based on electromyographic phase consistency according to claim 1, characterized in that, In step (3), the stimulus initiation effect threshold and the stable effect threshold are defined to define the width of the transition interval, specifically including: Two thresholds are defined on the curve, and the stimulus initiation affects the threshold. This indicates that the stimulus begins to produce a phase-locking effect on electromyographic activity, stabilizing and influencing the threshold. This represents the establishment of stable synergy between the target muscle and the stimulus; Determine the phase lock value minimum strength ,as well as minimum strength ,Will Defined as the width of the transition interval.

10. A device for optimizing spinal cord electrical stimulation parameters based on electromyographic phase consistency, characterized in that, The device includes a memory and one or more processors, wherein the memory stores executable code, and the one or more processors execute the executable code to implement the spinal cord electrical stimulation parameter optimization method according to any one of claims 1-9.

Images