Time-interference based parkinsonian subthalamic nucleus stimulation parameter optimization system

Abstract

This invention discloses a time-interference-based system for optimizing stimulation parameters of the subthalamic nucleus in Parkinson's disease, belonging to the field of stimulation parameter optimization technology. The system includes: a multimodal data acquisition and preprocessing module for acquiring motor neural data and performing data preprocessing; a gait electromyographic stability assessment module for analyzing motor stability characteristics to construct stability indices and determine stability improvement before and after stimulation; a stimulation dose-response modeling module for identifying the half-effect equivalent stimulation intensity and target stability level, inferring the potential improvement trend of the current stimulation, generating candidate stimulation parameter combinations and improvement predictions; and a comprehensive benefit parameter optimization module for fitting the improvement response time constant, obtaining an evaluation window, assessing the comprehensive benefit, and selecting recommended stimulation parameters. This invention solves the technical problem of existing technologies where large fluctuations and high noise in single-modal data make it difficult to stably reflect the patient's true neuromotor state, leading to unreliable stimulation parameter optimization.

Description

Technical Field

[0001] This invention relates to the field of stimulation parameter optimization technology, and in particular to a system for optimizing stimulation parameters of the subthalamic nucleus in Parkinson's disease based on time-interference. Background Technology

[0002] Parkinson's disease is a common neurodegenerative brain disorder. As the disease progresses, patients often experience gait instability, bradykinesia, and postural control impairment, significantly impacting their daily activities. Clinically, deep brain stimulation has become an important treatment for improving motor symptoms, with subthalamic nucleus stimulation being widely used due to its ability to modulate abnormal neural circuit activity. With the widespread adoption of wearable sensors, physiological signal monitoring, and intelligent medical devices, quantitative assessment and individualized regulation of patient status based on multi-source motor, physiological, and stimulation parameter data are gradually becoming an important direction for the development of precision neuromodulation.

[0003] For example, invention patent CN120392078A discloses a wearable tactile bio-gait feedback system and method for Parkinson's patients, belonging to the field of human gait rehabilitation. It includes the following modules: a posture perception module: using inertial measurement units (IMUs) strapped to the insteps of both feet to acquire accelerometer, magnetometer, and gyroscope information of the foot positions; a data processing module: receiving data from the posture perception module, estimating the foot posture, and using the foot posture to estimate the foot gait phase, finally outputting a signal according to a gait feedback stimulation strategy; and a tactile feedback module: receiving the stimulation signal from the data processing module and providing vibration stimulation signals to the back and calf muscles. This invention employs the aforementioned wearable tactile bio-gait feedback system and method for Parkinson's patients. It can utilize the relevant posture data provided by the IMU, complete foot posture estimation through multi-rate multi-correlation entropy Kalman filtering, estimate the foot phase using an adaptive oscillator algorithm, and implement vibration stimulation based on a set gait feedback strategy.

[0004] For example, invention patent CN116236159B discloses an adaptive feedback system and method for Parkinson's disease. This system includes a central coordinator, a data acquisition system, a stimulus generation system, a stimulation device, and a data processing system. The data acquisition system collects data from the target user, obtains user data, and sends the user data to the data processing system. The data processing system analyzes the user data to obtain user analysis results. The stimulus generation system determines stimulus information based on the user analysis results. The central coordinator determines a stimulus signal based on the stimulus information and controls the stimulation device to provide electrical stimulation to the target user based on the stimulus signal. This invention can automatically determine stimulus information based on user analysis results. By sending the stimulus information to the central coordinator, it can automatically generate a stimulus signal. Based on the stimulus signal, it can automatically control the stimulation device to provide electrical stimulation to the target user, thereby achieving the adaptive feedback effect for Parkinson's disease.

[0005] However, when assessing Parkinson's patients, significant tremors and rhythmic instability are observed in gait evaluations. Electromyographic signals are also easily affected by electrode adhesion quality, skin oil, and postural changes, leading to large fluctuations and high noise in single-modal data. Directly relying on the original gait or electromyographic indicators for stimulation parameter optimization introduces significant errors, making it difficult to stably reflect the patient's true neuromotor state and resulting in a lack of reliable data support for stimulation parameter decisions.

[0006] Therefore, in order to address the above problems, there is an urgent need for a time-intervention-based system for optimizing the stimulation parameters of the subthalamic nucleus in Parkinson's disease. Summary of the Invention

[0007] To address the technical problems in existing technologies for Parkinson's gait assessment, such as large fluctuations and high noise in single-modal data, which make it difficult to stably reflect the patient's true neuromotor state and lead to unreliable stimulation parameter optimization, this invention provides a time-interference-based system for optimizing stimulation parameters of the subthalamic nucleus in Parkinson's disease. The technical solution is as follows:

[0008] A time-interference-based system for optimizing stimulation parameters of the subthalamic nucleus in Parkinson's disease, comprising:

[0009] The system comprises several modules: a multimodal data acquisition and preprocessing module for periodically acquiring patients' motor neuron data in real time, and performing filtering, normalization, and time alignment; a gait electromyography (EMG) stability assessment module for analyzing gait rhythm fluctuations, postural sway amplitude, and EMG morphological similarity based on motor neuron data, and generating stability indices for gait EMG fusion; and a stimulus dose-response modeling module for constructing dose-response relationships by combining historical motor neuron data and corresponding stability improvement records, identifying half-effect equivalent stimulus intensity and target stability level, inferring the potential improvement trend of the current stimulus under different parameters, and generating candidate stimulus parameter combinations and corresponding improvement predictions; and a comprehensive benefit parameter optimization module for fitting the improvement response time constant based on the stability improvement trajectory after historical stimulation and obtaining an evaluation window; inferring the time coverage ratio by combining the improvement response time constant and the evaluation window, and comprehensively evaluating the comprehensive benefit by integrating candidate improvement predictions; selecting recommended stimulus parameters according to the comprehensive benefit ranking, and performing safety constraint verification.

[0010] Optionally, the process of periodically collecting the patient's motor nerve data in real time and performing filtering, normalization, and time alignment on the motor nerve data is as follows: Motor nerve data is collected during the gait cycle, including: stride length, stance phase time, plantar pressure, surface electromyography (EMG) signals of muscle groups, stimulation start time stamp, stimulation end time stamp, stimulation duration, stimulation current, and electrode impedance. The gait cycle is defined as the time interval between two consecutive foot strikes on the same side. The starting point of the gait cycle is the sampling moment when the plantar pressure signal first changes from zero to a non-zero value, and the ending point is the sampling moment when the ipsilateral plantar pressure signal changes from zero to a non-zero value again at subsequent sampling points. The stride length, stance phase time, and time alignment are then processed. Low-pass filtering of plantar pressure is applied to remove high-frequency noise. Band-pass filtering and rectification envelope algorithms are used to perform band-pass filtering, rectification, and envelope extraction on the surface electromyography (EMG) signals of muscle groups within the gait cycle, converting the original surface EMG waveforms of muscle groups into continuous and stable EMG envelope signal curves and identifying the peak values ​​of the EMG envelope. The motor neural data of each gait cycle are time-normalized to a uniform number of sampling points, and the motor neural data of different gait cycles are aligned according to the corresponding time index based on the normalized uniform time sampling points. z-score normalization is performed on the motor neural data to eliminate differences between different cycles. A stimulation parameter optimization database is constructed, and the original and preprocessed motor neural data are written into the stimulation parameter optimization database in units of gait cycle.

[0011] Optionally, the specific process of analyzing gait rhythm fluctuations, postural sway amplitudes, and electromyographic morphological similarities based on motor neural data to form a stability index for gait electromyographic fusion is as follows: Read motor neural data from N preprocessed gait cycles to obtain stride length sequences and stance phase time sequences; calculate the mean and standard deviation of the stride length sequences to obtain the average stride length and standard deviation, and divide the standard deviation by the average stride length to obtain the stride length coefficient of variation; calculate the mean and standard deviation of the stance phase time based on the stance phase time sequences, and divide the standard deviation by the mean of the stance phase time to obtain the stance phase time coefficient of variation; simultaneously, for each gait cycle, obtain the plantar pressure sequence within the current cycle, and calculate the mean and standard deviation to obtain the mean and magnitude of lateral sway of the center of gravity; average the magnitude of the center of gravity sway over all gait cycles. The average center of gravity sway was obtained for all cases. Historical gait cycles in which the historical stride length coefficient of variation, support phase time coefficient of variation, and average center of gravity sway were all within the corresponding stable threshold range were selected from the stimulation parameter optimization database. The corresponding electromyographic envelope signals were extracted and averaged to obtain the reference electromyographic value. The cross-correlation coefficient was calculated by aligning the electromyographic envelope signal of the current gait cycle with the reference electromyographic value point by point to obtain the electromyographic morphological correlation coefficient. The sum of squares of the stride length coefficient of variation and the support phase time coefficient of variation was calculated, and the negative of the sum of squares was used as the exponent for natural exponential operation to obtain the gait temporal stability term. The current electromyographic morphological correlation coefficient was nonnegated and divided by the sum of a constant and the average value of the lateral center of gravity sway to obtain the electromyographic posture coupling term. The gait temporal stability term and the electromyographic posture coupling term were multiplied to obtain the comprehensive value of motion stability.

[0012] Optionally, the specific process of determining the stability improvement before and after stimulation by combining the start and end times of stimulation using stability indices is as follows: Based on the stimulation start time stamp and stimulation end time stamp, the median of the comprehensive motion stability values ​​of the M gait cycles before the stimulation begins is selected as the baseline motion stability value of the stimulation; the median of the comprehensive motion stability values ​​of the M gait cycles after the stimulation ends is selected as the post-stimulation motion stability value of the stimulation; the difference between the post-stimulation motion stability value and the baseline motion stability value is calculated as the stability improvement amount of the stimulation; the baseline motion stability value, post-stimulation motion stability value, and stability improvement amount corresponding to each stimulation are written into the stimulation parameter optimization database.

[0013] Optionally, the specific process of constructing a dose-response relationship and identifying the half-effect equivalent stimulus intensity and target stability level by combining historical motor neural data and corresponding stability improvement records is as follows: Read the baseline motor stability value, post-stimulation motor stability value, and stability improvement amount corresponding to the current stimulus, as well as the stimulation current, electrode impedance, and stimulation duration; simultaneously, based on historical stimulus events in the stimulus parameter optimization database, select the median of historical electrode impedance as the reference impedance, and select the P90 quantile value of historical stability improvement as the historical maximum improvement amount; select historical stimulus events with historical stability improvement amounts greater than the improvement threshold, and statistically analyze the corresponding post-stimulation motor stability values. The median is used as the target motion stability value; the equivalent stimulus intensity value is obtained by adding a constant to the ratio of electrode impedance to reference impedance and dividing the stimulus current by the summation; based on the equivalent stimulus intensity value and the corresponding stability improvement of each historical stimulus event, an exponential saturation growth model is used to construct a dose-effect prediction function, and the minimum and maximum values ​​of the historical equivalent stimulus intensity values ​​are used as the search interval. Each candidate value in the search interval is substituted into the dose-effect prediction function to calculate the historical predicted improvement, and the sum of squared errors between the historical predicted improvement and the corresponding actual stability improvement is calculated. The candidate value that minimizes the sum of squared errors is selected as the half-effect equivalent stimulus intensity value.

[0014] Optionally, the specific process for inferring the potential improvement trend of the current stimulus under different parameters is as follows: subtract the target motion stability value from the baseline motion stability value of the current stimulus and perform non-negation processing to obtain the baseline deviation value; subtract the baseline deviation value from the historical maximum improvement to obtain the improvement potential term; divide the current equivalent stimulus intensity value by the half-effect equivalent stimulus intensity value and use the inverse as the exponent for natural exponent calculation; subtract the natural exponent calculation result from the constant to obtain the stimulus effect term; multiply the improvement potential term and the stimulus effect term to obtain the predicted stability improvement amount of the next stimulus.

[0015] Optionally, the specific process of generating candidate stimulation parameter combinations and corresponding improvement predictions is as follows: Based on the rated output range and rated adjustable step size of the stimulation device, a set of candidate stimulation parameter combinations is generated, including different stimulation currents and stimulation durations; for each set of candidate stimulation parameter combinations, the corresponding equivalent stimulation intensity value is calculated, and the predicted stability improvement amount is calculated to obtain a set of candidate predicted stability improvement amounts; the candidate stimulation parameter combinations and candidate predicted stability improvement amounts are written into the stimulation parameter optimization database.

[0016] Optionally, the specific process of fitting the improved response time constant based on the stability improvement trajectory after historical stimuli and obtaining the evaluation window is as follows: read the time change sequence of the motion stability value after stimulation corresponding to the historical stimulus event from the stimulus parameter optimization database, take the end time stamp of each stimulus as the time origin, combine the corresponding baseline motion stability value and the historical maximum improvement, use the exponential recovery model to fit the change sequence, and solve the parameters of the exponential recovery model by the nonlinear least squares fitting method to obtain the time constant corresponding to each historical stimulus event; statistically analyze the time constants obtained for each historical stimulus event, select the median as the improved response time constant, and use three times the improved response time constant as the evaluation window duration.

[0017] Optionally, the specific process of combining the improved response time constant and the evaluation window to infer the time coverage ratio, and comprehensively evaluating the overall benefits of candidate improvement predictions, is as follows: Divide the improved response time constant by the evaluation window duration to obtain the time weight value; divide the negative of the evaluation window duration by the improved response time constant as the exponent for natural exponentiation, and subtract the result of the natural exponentiation from the constant to obtain the improved time coverage ratio; obtain the candidate predicted stability improvement amount, and multiply the candidate predicted stability improvement amount by the time weight value and the improved time coverage ratio to obtain the average improvement value of the evaluation window; obtain the candidate stimulus parameter combination corresponding to each candidate predicted stability improvement amount, and multiply the square of the corresponding stimulus current by the corresponding stimulus duration to obtain the stimulus load value; add the constant to the stimulus load value and take the natural logarithm as the denominator; divide the average improvement value of the evaluation window by the denominator to obtain the overall benefit value.

[0018] Optionally, the specific process of selecting recommended stimulus parameters according to comprehensive benefit ranking and performing safety constraint verification is as follows: Calculate the comprehensive benefit value corresponding to the improvement in stability of all candidate parameters, sort them in descending order, and select the candidate stimulus parameter with the largest comprehensive benefit value as the prompt parameter combination; perform safety verification on the prompt parameter combination: compare the stimulus current with the maximum stimulus current threshold, and compare the stimulus duration with the maximum stimulus duration threshold. If the stimulus current is greater than the maximum stimulus current threshold or the stimulus duration is greater than the maximum stimulus duration threshold, the parameter exceeding the limit is truncated to the corresponding threshold; at the same time, the change in stimulus current and the change in stimulus duration are obtained through differential calculation, and compared with the maximum change threshold and the maximum change threshold, respectively. If any change exceeds the corresponding change threshold, the corresponding parameter is truncated; push parameter prompts according to the prompt parameter combination that has passed safety verification, and write the prompt parameter combination that has passed safety verification and the corresponding comprehensive benefit value into the stimulus parameter optimization database.

[0019] The beneficial effects of the technical solutions provided in the embodiments of the present invention include at least the following:

[0020] (1) This invention simultaneously collects multimodal motor neural data including stride length, stance phase time, plantar pressure, and electromyographic envelope. Based on processing such as filtering, normalization, time alignment, and posture coupling, it constructs a gait temporal stability term and an electromyographic posture coupling term to form a robust comprehensive index of motor stability. Compared with relying on a single signal, it can significantly reduce the impact of jitter, noise, and electromyographic drift on the assessment, and more stably reflect the patient's true neuromotor state.

[0021] (2) This invention constructs equivalent stimulus intensity, half-effect equivalent stimulus intensity, and target stability level, and uses historical stimulus records to construct an exponential dose-response prediction relationship, which can infer the possible improvement trend of different parameters based on the current stimulus state. This transforms the adjustment of stimulus parameters from an empirical selection to a predictable process based on a quantitative response model, improving the rationality and individualization of stimulus decisions.

[0022] (3) This invention improves the response time constant by fitting historical improvement trajectories and introduces an evaluation window and time coverage ratio. It normalizes the predicted improvement effect and stimulus load in logarithmic form to construct a comprehensive benefit index, which can achieve an adaptive trade-off between the degree of improvement and the energy burden of stimulation. It avoids excessive stimulation caused by pursuing only high current or long duration, making the recommended parameters safer and more energy-efficient.

[0023] (4) By performing safety checks on the upper limit of current, upper limit of duration, and range of parameter changes before outputting recommended parameters, this invention can automatically truncate abnormal parameters and avoid actual risks caused by stimulation exceeding the threshold or adjustment being too large. This closed-loop constraint mechanism ensures that the system meets the equipment safety boundary while automatically optimizing, improving the feasibility and reliability of the solution in clinical practice. Attached Figure Description

[0024] 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.

[0025] Figure 1 This is a structural diagram of the Parkinson's subthalamic nucleus stimulation parameter optimization system based on time interference provided in an embodiment of the present invention;

[0026] Figure 2 This is a distribution map of the comprehensive benefit values ​​provided in the embodiments of the present invention;

[0027] Figure 3 This is a flowchart of the improvement time constant fitting and evaluation window determination based on historical stimulus response provided in an embodiment of the present invention. Detailed Implementation

[0028] The technical solution of the present invention will now be described with reference to the accompanying drawings.

[0029] In embodiments of the present invention, words such as "exemplarily," "for example," etc., are used to indicate that something is an example, illustration, or description. Any embodiment or design described as "exemplary" in the present invention should not be construed as being more preferred or advantageous than other embodiments or designs. Specifically, the use of the word "exemplary" is intended to present the concept in a concrete manner. Furthermore, in embodiments of the present invention, the meaning expressed by "and / or" can be both, or either one.

[0030] In the embodiments of this invention, the terms "image" and "picture" may sometimes be used interchangeably. It should be noted that, without emphasizing the distinction between them, they convey the same meaning. Similarly, the terms "of," "corresponding (relevant)," and "corresponding" may sometimes be used interchangeably. It should be noted that, without emphasizing the distinction between them, they convey the same meaning.

[0031] In this embodiment of the invention, sometimes a subscript such as W1 may be written in a non-subscript form such as W1. When the difference is not emphasized, the meaning they express is the same.

[0032] To make the technical problems, technical solutions and advantages of the present invention clearer, a detailed description will be given below in conjunction with the accompanying drawings and specific embodiments.

[0033] This invention provides a time-interference-based system for optimizing stimulation parameters of the subthalamic nucleus in Parkinson's disease, such as... Figure 1The diagram shows the structure of a time-interference-based Parkinson's disease subthalamic nucleus stimulation parameter optimization system. This system includes: a multimodal data acquisition and preprocessing module, used to acquire patients' motor neuron data in real-time on a periodic basis, and perform filtering, normalization, and time alignment on the motor neuron data; a gait electromyography stability assessment module, used to analyze gait rhythm fluctuations, postural sway amplitude, and electromyography morphological similarity based on motor neuron data, and to form a stability index for gait electromyography fusion; combining the stimulation start and end times, the stability index is used to determine the stability improvement before and after stimulation; a stimulation dose-response modeling module, used to combine historical motor neuron data and corresponding stability improvement records to construct a dose-response relationship, identify the half-effect equivalent stimulation intensity and target stability level; infer the potential improvement trend of the current stimulation under different parameters, and generate candidate stimulation parameter combinations and corresponding improvement predictions; a comprehensive benefit parameter optimization module, used to fit the improvement response time constant based on the stability improvement trajectory after historical stimulation, and obtain an evaluation window; combine the improvement response time constant and the evaluation window to infer the time coverage ratio, and comprehensively evaluate the comprehensive benefit by integrating candidate improvement predictions; select recommended stimulation parameters according to the comprehensive benefit ranking, and perform safety constraint verification.

[0034] Optionally, the process of periodically collecting the patient's motor nerve data in real time and performing filtering, normalization, and time alignment on the motor nerve data is as follows: The patient's motor nerve data during the gait cycle is collected, including: stride length, stance phase time, plantar pressure, surface electromyography (EMG) signals of muscle groups, stimulation start timestamp, stimulation end timestamp, stimulation duration, stimulation current, and electrode impedance. Stride length and stance phase time are obtained by identifying landing events using the foot IMU of a wearable gait monitoring device; plantar pressure is measured in real time using a pressure pad; surface EMG signals are collected using electrodes attached to the skin of the target muscle groups; stimulation timestamps, current, duration, and electrode impedance are automatically recorded and synchronously written by the stimulation device; all acquisition devices use a uniform sampling frequency to ensure the temporal consistency between gait dynamics signals and EMG signals, thereby avoiding cycle division errors caused by data asynchrony across devices. The gait cycle is defined as the time interval between two consecutive foot strikes on the same side. The starting point of the gait cycle is determined by the sampling moment when the plantar pressure signal first changes from zero to non-zero, and the ending point is the sampling moment when the plantar pressure signal changes from zero to non-zero again at subsequent sampling points. If short-term data loss occurs, linear interpolation is performed on the plantar pressure sequence to restore continuity, ensuring stable detection of the gait cycle boundaries. To avoid misjudgments of zero and non-zero values ​​due to zero drift, baseline drift, or pre-load of the plantar pressure sensor, automatic zero-point calibration is performed before each acquisition, including baseline sampling during the resting period and calculating the average pre-pressure value as a dynamic threshold bias. The start and end points of the gait cycle are determined using a "plantar pressure > baseline pressure + adaptive threshold" method, where the adaptive threshold is set to 3 to 5 times the standard deviation of the noise level, making the cycle detection process robust to factors such as sensor zero drift and pre-contact pressure. Low-pass filtering is applied to stride length, stance phase time, and plantar pressure to remove high-frequency noise. A Butterworth filter of 4 to 8 Hz is preferred for low-pass filtering to match the dominant frequency band of gait dynamics and suppress high-frequency instrument noise. Filter parameters can be calibrated through pre-experimentation to reduce the impact of gait differences among patients on the filtering results. Bandpass filtering and rectification envelope algorithms are used to perform bandpass filtering, rectification, and envelope extraction on the surface electromyography (EMG) signals of muscle groups during the gait cycle, converting the original surface EMG waveforms of muscle groups into continuous and stable EMG envelope signal curves and identifying EMG envelope peaks. Bandpass filtering is preferably performed in the range of 20 to 450 Hz to effectively filter out motion artifacts and baseline drift. The specific cutoff frequency, filtering order, and sampling rate used for filtering are adaptively set by the actual sampling rate of the acquisition device. Full-wave rectification is used for rectification, and the sliding window RMS envelope method is used for envelope extraction to make the EMG activation morphology smoother and more stable, facilitating subsequent alignment and morphological analysis.Motor neural data for each gait cycle are time-normalized to a uniform number of sampling points. Based on these normalized, uniform time sampling points, motor neural data from different gait cycles are aligned according to their corresponding time indices. For optimal time normalization, each gait cycle is resampled to 100 to 200 standard sampling points to ensure comparability in temporal structure across different cycles. Spline interpolation is used during alignment to preserve the morphological characteristics of the original signal, thus avoiding peak shift issues caused by simple linear interpolation. Motor neural data are z-score normalized for each gait cycle, meaning that signals of the same type within each cycle are normalized using their own mean and standard deviation to eliminate differences between cycles. This ensures intra-patient consistency in processed gait and electromyographic signals, enhancing the robustness of subsequent indicator calculations. A stimulation parameter optimization database is constructed, where raw and preprocessed motor neural data are written into the database by gait cycle. A timestamp index structure is used, with data written according to both stimulation events and gait cycles, allowing direct correlation between each stimulus and the corresponding preceding and following gait changes.

[0035] In this implementation scheme, by performing multi-source synchronous acquisition, noise suppression, time normalization, and periodic alignment processing on motor nerve data, the stability and comparability of gait dynamics signals and surface electromyography (EMG) signals can be significantly improved. Through steps such as unified sampling frequency, filter parameter calibration, RMS envelope extraction, and z-score normalization, random fluctuations caused by different devices, different periods, and individual patient differences are effectively reduced, allowing for precise characterization of gait and EMG changes before and after stimulation. Simultaneously, a database structure based on timestamps to associate gait cycles with stimulation events is constructed, enabling complete tracking of the response process for each stimulus. Overall, this improves data quality and temporal consistency, providing reliable and practical basic data support for subsequent stability assessment, stimulus response modeling, and parameter optimization.

[0036] Optionally, the specific process of analyzing gait rhythm fluctuations, postural sway amplitude, and electromyographic morphological similarity based on motor neural data to form a stability index for gait electromyographic fusion is as follows: Motor neural data within N preprocessed gait cycles are read to obtain stride length sequences and stance phase time sequences; where N represents the number of consecutive gait cycles used to calculate the current stability index, preferably adaptively selected within the range of 10 to 20 based on the patient's gait stability, to both cover typical gait fluctuation characteristics and avoid the introduction of non-stationary effects by excessively long windows, and the corresponding N effective cycles are automatically extracted through time stamp windows before and after the stimulation event; the mean and standard deviation of the stride length sequence are calculated to obtain the average stride length and stride length standard deviation, and the stride length variation coefficient is obtained by dividing the stride length standard deviation by the average stride length; the mean and standard deviation of the stance phase time are calculated based on the stance phase time sequence, and the stance phase time variation coefficient is obtained by dividing the stance phase time standard deviation by the stance phase time mean; simultaneously, for each gait cycle, the plantar pressure sequence within the current cycle is obtained, and the mean and standard deviation are calculated to obtain the mean and magnitude of lateral center of gravity sway; where, the lateral center of gravity sway... The average sway was derived from the plantar pressure sequence and obtained based on a conventional plantar pressure center calculation model. Historical gait cycles where the historical stride length coefficient of variation, stance phase time coefficient of variation, and average center of gravity sway were all within their corresponding stable threshold ranges were selected from the stimulation parameter optimization database. The corresponding electromyographic envelope signals were extracted and averaged to obtain reference electromyographic values. These reference values ​​were averaged using a uniform number of time-normalized points to ensure lateral comparability with the current gait cycle's electromyographic curve. The current gait cycle's electromyographic envelope signal and the reference electromyographic values ​​were then cross-aligned point-by-point for cross-validation. The correlation coefficients are obtained by calculating the electromyographic morphology correlation coefficients. The cross-correlation calculation uses normalized cross-correlation coefficients to stabilize the results within the range of 0 to 1, thus avoiding the influence of individual patient differences in electromyographic amplitude on morphological similarity assessment. The sum of squares of the stride length variation coefficient and the support phase time variation coefficient is calculated, and the negative of the sum is used as the exponent for natural exponential calculation to obtain the gait temporal stability term. The exponential form is used to enhance the difference in gait rhythm stability; the smaller the rhythm fluctuation and the lower the variation coefficient, the closer the exponent term is to 1, and the more sensitive it is to reflecting the temporal consistency brought about by a stable gait. The current electromyographic morphology correlation coefficient is non-negatively processed and divided by the sum of a constant and the average value of the center of gravity lateral sway, to obtain the electromyographic posture coupling term. Non-negativity involves comparing the calculated result with zero and taking the larger value, which avoids abnormal cross-correlation values ​​that could lead to index bias. The average value of the center of gravity lateral sway introduced into the denominator automatically suppresses periods with large posture fluctuations, making the electromyographic morphological similarity closer to actual posture control ability. The comprehensive value of motion stability is obtained by multiplying the gait temporal stability term with the electromyographic posture coupling term. This value integrates three types of motion stability elements: gait rhythm, postural sway, and electromyographic morphology. The output range is stable and the weights of different modal signals have physical interpretability, thus providing a reliable basis for judging subsequent stimulus improvement.

[0037] The specific formula for the comprehensive value of motion stability is as follows:

[0038] ;

[0039] In the formula, It represents the overall value of motor stability, used to reflect the robustness of the patient's current motor function status as a whole; The standard deviation of step length represents the degree of dispersion of step length in a continuous gait cycle. The larger the standard deviation, the more unstable the step length changes and the more obvious the gait rhythm fluctuations. This represents the average step size, serving as a benchmark for the standard deviation of the step size. The standard deviation of the support phase time describes the dispersion of the single-leg support time in a continuous cycle. The larger the standard deviation, the more obvious the temporal fluctuation, that is, the unstable gait rhythm. It represents the average time value of the support phase, serving as a benchmark for the time fluctuation of the support phase; The electromyographic morphology correlation coefficient represents the degree of similarity between the current gait cycle electromyographic waveform and the reference electromyographic morphology. This represents the average lateral sway of the center of gravity, reflecting the lateral shift of the center of pressure on the sole of the foot during gait. The larger the value, the more obvious the sway and the weaker the stability.

[0040] This implementation scheme extracts multimodal stability features, including stride length variation, stance phase temporal fluctuations, center of gravity sway, and electromyographic morphological similarity, across multiple consecutive gait cycles. By employing time normalization, cross-correlation normalization, posture compensation, and exponential enhancement strategies, it achieves a more robust and refined quantitative assessment of patient movement stability. This method effectively suppresses interference from single-modal noise, postural deviations, and individual differences in electromyographic amplitude, ensuring consistency and physical interpretability of the obtained comprehensive movement stability values ​​across different cycles. This provides a reliable data foundation for subsequent stimulation effect assessment and parameter optimization, significantly improving the overall feasibility and stability.

[0041] Optionally, the specific process of determining the improvement in stability before and after stimulation by combining the start and end times of stimulation using stability indices is as follows: Based on the stimulation start and end timestamps, the median of the comprehensive motion stability values ​​of the M gait cycles before stimulation is used as the baseline motion stability value, and the median of the comprehensive motion stability values ​​of the M gait cycles after stimulation is used as the post-stimulation motion stability value. Here, M represents the length of the before-and-after control window used to evaluate the stimulation effect, preferably 5 to 10 consecutive gait cycles, which can be adaptively adjusted according to the patient's gait stability. The corresponding cycles are automatically extracted using the stimulation event timestamps to ensure consistency in the before-and-after comparison. If there are short-term missing abnormal stability values ​​within a cycle, they can be corrected using linear interpolation to avoid affecting the accuracy of the baseline level. Gait cycles after stimulation are extracted in time-stamp order. If significant transient fluctuations occur due to the instant the stimulator is turned off, the abnormal cycle is automatically skipped; that is, the comprehensive stability value of the current cycle is not included in the median statistics. Simultaneously, a new effective cycle is added from subsequent cycles to ensure that the samples used for statistics are all effective gait cycles under physiologically stable conditions. The difference between the post-stimulation motor stability value and the baseline motor stability value is calculated as the stability improvement amount of the stimulus, and a prompt is triggered when a reverse improvement occurs; the baseline motor stability value, post-stimulation motor stability value, and stability improvement amount corresponding to each stimulus are written into the stimulus parameter optimization database.

[0042] In this implementation scheme, stability assessment is performed by selecting continuous and modified gait cycles before and after stimulation. A mechanism is employed to automatically eliminate initialization interference, pause gait, and transient fluctuations upon stimulation closure. This ensures that both baseline levels and post-stimulation stability values ​​are statistically derived from true, continuous, and effective gait states. This method ensures that the assessment of stimulation effects is not affected by abnormal cycles, improves the reliability of stability improvement, and provides a more accurate and reliable data foundation for subsequent dose modeling and parameter optimization, thereby significantly enhancing its applicability and robustness in clinical settings.

[0043] Optionally, the specific process of constructing a dose-response relationship and identifying the half-effect equivalent stimulus intensity and target stability level by combining historical motor neuron data and corresponding stability improvement records is as follows: Read the baseline motor stability value, post-stimulation motor stability value, and stability improvement amount corresponding to the current stimulus, as well as the stimulation current, electrode impedance, and stimulation duration; simultaneously, optimize historical stimulus events in the stimulus parameter database, selecting 50 to 100 historical stimulus event samples. A sample size of at least 50 can cover different intensity ranges such as low dose, near half-effect dose, and near historical maximum dose, ensuring sufficient fitting constraints for the exponential saturation curve across the entire dose range, thereby guaranteeing the stability of the estimated half-effect equivalent stimulus intensity value and historical maximum improvement amount; a sample size of no more than 100 can control the interference of long-term factors such as patient disease progression, changes in medication regimens, and fine-tuning of electrode positions on the model, avoiding the downgrading of the model's accuracy to the current state by outdated data. This approach ensures representativeness of the model's state while reducing the computational complexity of parameter fitting, facilitating real-time model updates in clinical settings. The median of historical electrode impedance is selected as the reference impedance, and the P90 quantile of historical stability improvement is selected as the maximum historical improvement. Historical stimulation events with historical stability improvement exceeding the improvement threshold are selected, and the median of the corresponding post-stimulation motor stability values ​​is calculated as the target motor stability value. The reference impedance is used to offset impedance fluctuations caused by electrode adhesion quality and differences in skin conductance, allowing for comparison of stimulation intensities across different stimulation events on the same equivalent scale. The P90 quantile avoids model bias caused by individual abnormally high values, more accurately reflecting the upper limit of improvement achievable by patients in historical stimuli. The improvement threshold is adaptively set based on the patient's long-term average improvement level and can be set to the median of the historical distribution, preventing ineffective or negative stimulation data from entering the modeling process and ensuring statistical robustness of the target stability level. The equivalent stimulation intensity value is obtained by adding a constant to the ratio of electrode impedance to reference impedance and dividing the stimulation current by the sum. This equivalent stimulation intensity value is used to standardize the current value according to impedance, ensuring that the actual stimulation dose reflects the effective amount of real biocurrent entering the tissue, rather than the device output. Based on the equivalent stimulation intensity values ​​and corresponding stability improvements of each historical stimulation event, an exponential saturation growth model is used to construct a dose-response prediction function, with the following form:

[0044] ;

[0045] in, This indicates that the equivalent stimulus intensity value is The historical forecast improvement amount; This indicates the largest historical improvement. This represents the half-effect equivalent stimulus intensity value. It reflects the dose-response characteristics of neural stimulation, where low doses lead to slow improvement, medium doses to rapid improvement, and high doses tend to saturate, which better aligns with the clinical patterns of deep brain stimulation therapy. The minimum and maximum historical equivalent stimulus intensity values ​​are used as the search interval. Each candidate value within the search interval is substituted into the dose-response prediction function to calculate the historical predicted improvement. The sum of squared errors between the historical predicted improvement and the corresponding actual stable improvement is then calculated. The candidate value that minimizes the sum of squared errors is selected as the half-effect equivalent stimulus intensity value. The half-effect equivalent stimulus intensity value corresponds to the stimulation dose at which the improvement reaches 50% of the maximum improvement. It is a key parameter of the dose-response curve, reflecting the patient's sensitivity to stimulation. The solution is based on nonlinear least squares optimization to ensure the objectivity and uniqueness of the model fit.

[0046] In this implementation scheme, by introducing reference impedance, historical maximum improvement, and data adaptive threshold, the dose information of different stimulus events is made comparable on a uniform scale, improving the stability of the modeling. The constructed exponential saturation growth model can accurately characterize the typical dose-response pattern of neural stimulation, making the improvement prediction closer to clinical reality. By performing a point-by-point search within the historical equivalent stimulus intensity range and solving for the half-effect equivalent stimulus intensity with the minimum sum of squared errors as the criterion, the obtained key parameters are ensured to have statistical optimality and physiological rationality. Overall, this method achieves the joint inference of individual patient dose sensitivity, self-improvement upper limit, and target stability level, making the subsequent stimulation parameter recommendations have reliable dose basis and interpretability, thereby significantly improving the accuracy and safety of stimulation parameter optimization.

[0047] Optionally, the specific process for inferring the potential improvement trend of the current stimulus under different parameters is as follows: Subtract the target motion stability value from the baseline motion stability value of the current stimulus and perform non-negativity processing to obtain the baseline deviation value. The non-negativity operation involves setting the value to zero if the difference is negative to avoid meaningless negative deviations when the target value has already been exceeded, ensuring that the subsequent improvement potential term only reflects the portion that can still be improved. Subtract the baseline deviation value from the historical maximum improvement to obtain the improvement potential term, reflecting the remaining improvement space achievable at the patient's previous best improvement level, and ensuring that the predicted value does not exceed the patient's historically achievable physiological improvement limit. Divide the current equivalent stimulus intensity value by... The half-effect equivalent stimulus intensity value is used as the inverse to perform natural exponentiation. The stimulus effect term is obtained by subtracting the natural exponentiation result from a constant. The stimulus effect term is essentially derived from the exponential saturation characteristic of the dose-response curve, avoiding nonlinear amplification caused by the infinite increase of the stimulus dose, thus maintaining the physiological rationality of the predicted improvement. The improvement potential term is multiplied by the stimulus effect term to obtain the predicted stable improvement amount of the next stimulus. The product structure ensures that the predicted value simultaneously meets the constraints of the maximum space that the patient can achieve and the actual effect range that the current dose can produce, thus making the predicted improvement amount physiologically reasonable, monotonic, and continuous.

[0048] The specific formula for predicting the improvement in stability is as follows:

[0049] ;

[0050] In the formula, It represents the predicted amount of stability improvement, used to infer the expected amount of stability improvement under a given patient state and stimulation parameters, and to provide a quantitative basis for subsequent optimization decisions of stimulation parameters; This represents the historical maximum improvement, used to limit the maximum achievable improvement and provide the physiological boundary of the improvement range; Indicates the baseline motion stability value; The target motion stability value, along with the baseline motion stability value, reflects the gap between the patient's current state and the target stable state, and is used to determine the potential for release of improvement. Indicates the stimulation current; This indicates electrode impedance. The stimulation current and electrode impedance together determine the effective stimulation intensity of the current in the tissue. This represents the reference impedance, which is used to normalize and compensate for the stimulation efficiency under different impedance conditions, so that the stimulation dose under different electrode contact states is comparable. This represents the half-effect equivalent stimulus intensity value, used to describe the equivalent stimulus intensity required to achieve half the improvement effect, reflecting stimulus dose sensitivity.

[0051] In this implementation plan, by simultaneously utilizing the patient's stability baseline deviation before and after the current stimulation, historical maximum improvement capacity, and dose-response characteristics, the prediction results not only reflect the patient's actual improvement potential but also incorporate the biological dose-response law between equivalent stimulation intensity and half-effect intensity, thereby obtaining a physically meaningful and individualized predicted improvement amount for the next stimulation. The prediction model can adapt to the sensitivity differences among different patients, avoid over- or under-stimulation caused by individual differences, and provide a quantifiable, continuous, and stable basis for judging the improvement trend for subsequent parameter optimization.

[0052] Optionally, the specific process for generating candidate stimulation parameter combinations and corresponding improvement predictions is as follows: Based on the rated output range and rated adjustable step size of the stimulation device, a set of candidate stimulation parameter combinations is generated, including different stimulation currents and stimulation durations. The candidate stimulation current range is constrained by the upper limit of the safe current and the minimum effective current of the stimulation device, and can be automatically generated within a range of 1.0 to 5.0 mA with a device step size of 0.1 mA. The stimulation duration range is based on commonly used clinical stimulation durations, such as 10 to 60 seconds, and is automatically generated by gridding according to device step sizes, such as 1 second or 2 seconds. A Cartesian combination of these two parameters constitutes a complete candidate parameter search space. To avoid an excessively large search space leading to increased computational load, stratified sampling of the stimulation current and stimulation duration can be performed, for example, prioritizing dense sampling of the low-dose interval to improve resolution of dose-sensitive regions. For each set of candidate stimulation parameter combinations, the corresponding equivalent stimulation intensity value is calculated, and the predicted stability improvement is calculated to obtain a set of candidate predicted stability improvement values, enabling interpretable prospective estimation of candidate combinations that have not yet been implemented. The candidate stimulation parameter combinations and candidate predicted stability improvement values ​​are then written into the stimulation parameter optimization database.

[0053] In this implementation scheme, candidate stimulation parameter combinations are automatically generated based on the adjustable range of the device, and the corresponding predicted improvement is calculated using an equivalent stimulation intensity and dose-response model. This allows for prospective evaluation of multiple stimulation protocols without direct intervention in patients. The method can cover a sufficiently rich set of stimulation parameter combinations within a safe boundary and write them into a database in a structured manner. This provides a comprehensive, traceable, and clinically feasible set of candidate protocols for subsequent comprehensive benefit ranking and optimal parameter selection, thereby significantly improving the efficiency and reliability of stimulation parameter optimization.

[0054] Optionally, the specific process of fitting the improved response time constant based on the stability improvement trajectory after historical stimuli and obtaining the evaluation window is as follows: Figure 3The diagram shows the flowchart for fitting and evaluating the improvement time constant based on historical stimulus responses. The time variation sequence of post-stimulus motion stability values ​​corresponding to historical stimulus events is read from the stimulus parameter optimization database. This variation sequence is generated based on gait cycle data at a uniform sampling frequency, preferably resampled to 100 to 200 normalized time points per cycle to ensure comparability of time-series data between different stimulus events. For missing points in the variation sequence, linear interpolation is used for continuity repair to maintain the temporal coherence of the fitted data. Using the end timestamp of each stimulus as the time origin, combined with the corresponding baseline motion stability value and the historical maximum improvement, an exponential recovery model is used to fit the variation sequence. The parameters of the exponential recovery model are solved using a nonlinear least squares fitting method to obtain the time constant corresponding to each historical stimulus event. When performing nonlinear least squares fitting, an optimization method with a robust loss function is preferred. The robust loss function can be Huber loss to suppress the influence of occasional outliers on the time constant solution, ensuring a more stable and reliable estimation result for the time constant. The time-varying sequence of post-stimulus motor stability values ​​for each historical stimulus event is composed of continuous gait cycle stability composite values ​​arranged in timestamp order. Abnormal cycles are automatically filtered out to ensure the fitted data is continuous and reflects the true neural recovery process. The exponential recovery model takes the following form: ;in, This represents the post-stimulus stability value at time t after the stimulus ends; This represents the baseline motion stability value corresponding to the stimulus; This indicates the largest historical improvement. This represents the time after the stimulus ends, with the stimulus end timestamp as the time origin, and calculates the duration from the time corresponding to each value to the origin. The time constant is represented to depict the recovery trend of stability after stimulation, from rapid increase to gradual stabilization, which is consistent with the dynamic characteristics of the improvement process of deep brain stimulation in clinical practice. The nonlinear least squares fitting uses a robust loss function to reduce the impact of occasional outliers on the time constant solution. The time constants obtained from each historical stimulation event are statistically analyzed, and the median is selected as the improvement response time constant. Three times the improvement response time constant is used as the evaluation window length. This three-times time constant corresponds to the physiological timescale at which the model recovers to approximately 95% steady state, serving as a reliable observation interval for improvement. It can fully cover the main improvement stages after stimulation while avoiding the introduction of gait non-stationarity factors due to an excessively long window.

[0055] In this implementation plan, by performing exponential recovery fitting on the stability change process after historical stimulation, an improved response time constant that can truly reflect the patient's stimulus response speed is obtained. The median is selected as the individualized recovery parameter using a robust statistical method, so that the subsequent evaluation window has clear physiological significance and statistical stability. At the same time, using three times the time constant as the improvement observation period can cover the main improvement interval after stimulation, avoid the influence of occasional abnormal gait and long-term drift, and improve the accuracy and repeatability of stimulation effect assessment.

[0056] Optionally, the specific process of combining the improvement response time constant and the evaluation window to infer the time coverage ratio and comprehensively assess the overall benefit of candidate improvement predictions is as follows: Divide the improvement response time constant by the evaluation window duration to obtain a time weight value, which describes the proportion of the improvement response speed relative to the evaluation window. By scaling the time factor, the recovery rhythms of different patients are made comparable. Perform a natural exponentiation operation by dividing the negative of the evaluation window duration by the improvement response time constant, and subtract the natural exponentiation result from the constant to obtain the improvement time coverage ratio. The improvement time coverage ratio measures the effective improvement ratio that the predicted improvement can actually achieve within the evaluation window, reflecting the time occupancy of the recovery model. When the evaluation window is too short, the coverage ratio automatically decreases, thus avoiding overestimation of the recovery process. Obtain the candidate predicted stability improvement amount, and multiply the candidate predicted stability improvement amount by the time weight value and the improvement time coverage ratio to obtain the average improvement value of the evaluation window. The average improvement value of the evaluation window can comprehensively reflect the dual influence of the improvement magnitude and the time coverage degree, avoiding short-term stimulus bias caused by ranking solely by the improvement amount, and ensuring that the screening results take into account both improvement effect and time efficiency. For each candidate predicted stability improvement, a combination of candidate stimulation parameters is obtained. The stimulation load value is obtained by multiplying the square of the corresponding stimulation current by the corresponding stimulation duration. Using the square of the stimulation current more accurately reflects the actual burden of the stimulation dose on neural tissue. Combined with the stimulation duration, the total load helps to constrain unnecessary high-intensity or long-duration stimulation. The stimulation load value is added to a constant and the natural logarithm is used as the denominator. The constant ensures that the expression within the denominator remains positive, and the natural logarithm compresses the stimulation load to a reasonable range, making the penalty for high-load stimulation smoother and avoiding distortion of the overall benefit value due to an excessively large denominator, while maintaining the differentiability and numerical stability of the overall function. The average improvement value of the evaluation window is divided by the denominator to obtain the overall benefit value. The overall benefit value is used to simultaneously measure the predicted improvement effect and stimulation cost. A larger value indicates a more significant average improvement per unit stimulation load, and it is the core evaluation metric for the final recommended stimulation parameters. The overall benefit value automatically forms a unified evaluation scale among different candidate parameters, making the stimulation parameter optimization process interpretable and robust.

[0057] The specific formula for the comprehensive benefit value is as follows:

[0058] ;

[0059] In the formula, This represents the overall benefit value, used to measure the cost-effectiveness of different combinations of stimulus parameters in terms of improvement and stimulus consumption. This represents the improvement in the stability of the candidate prediction, reflecting the potential for stability improvement that the candidate stimulus plan can achieve within the evaluation window, and is the main source of the benefit improvement; This represents the improvement response time constant, which indicates the speed at which the patient responds to a stimulus. Indicates the duration of the evaluation window, which is the observation period used to evaluate the improvement effect; Indicates the stimulation current; The stimulation duration is indicated by the stimulation current and the stimulation duration, which together determine the energy load of the stimulation protocol.

[0060] In this embodiment, Table 1 is a comprehensive benefit value data table. The improvement response time constant was set to 10, and the evaluation window duration was 30. The table details the improvement in candidate predicted stability, stimulation current, stimulation duration, and overall benefit value for five parameter combinations. Specifically, combination 1 corresponds to an improvement in candidate predicted stability of 0.24, stimulation current of 1.8, stimulation duration of 22, and an overall benefit value of 0.0178; combination 2 corresponds to an improvement in candidate predicted stability of 0.31, stimulation current of 2.2, stimulation duration of 27, and an overall benefit value of 0.0201; combination 3 corresponds to an improvement in candidate predicted stability of 0.37, stimulation current of 2.5, stimulation duration of 33, and an overall benefit value of 0.0220; combination 4 corresponds to an improvement in candidate predicted stability of 0.42, stimulation current of 3.0, stimulation duration of 29, and an overall benefit value of 0.0239; and combination 5 corresponds to an improvement in candidate predicted stability of 0.39, stimulation current of 3.3, stimulation duration of 35, and an overall benefit value of 0.0208.

[0061] Table 1. Comprehensive Benefit Value Data Table

[0062] like Figure 2 The figure shows the distribution of comprehensive benefit values. The horizontal axis represents the parameter combination number, and the vertical axis represents the comprehensive benefit value. Combination 4, with the highest comprehensive benefit value, is marked in red for comparison with other combinations. (Refer to Table 1 and...) Figure 2It can be seen that combination 4 has the highest overall benefit value, indicating that the stimulation current and duration of combination 4 can achieve better stability improvement potential under relatively acceptable costs. The overall benefit values ​​of combinations 1, 2, 3, and 5 are relatively similar, showing a trend of gradually increasing from combination 1 to combination 4 and slightly decreasing from combination 5, indicating that stimulation parameters are not necessarily better the larger they are, and there is an optimal range. Overall, this reflects that the optimization of stimulation parameters requires a comprehensive consideration of the balance between improvement effect and stimulation load.

[0063] In this implementation plan, by constructing time weights, recovery coverage ratios, and stimulus loads, the improvement prediction and stimulus cost are incorporated into the quantitative system. At the same time, logarithmic penalties are used to suppress excessive stimulus loads, thereby improving the safety and stability of stimulus parameter screening. The comprehensive benefit value can fully reflect the magnitude of improvement, time efficiency, and stimulus cost, making the recommendation results more consistent with clinical use logic and improving the reliability and individualization of stimulus parameter optimization.

[0064] Optionally, the specific process of selecting recommended stimulation parameters according to comprehensive benefit ranking and performing safety constraint verification is as follows: Calculate the comprehensive benefit value corresponding to the improvement in stability of all candidate parameters, sort them in descending order, and select the candidate stimulation parameter with the largest comprehensive benefit value as the prompt parameter combination. Before ranking, the comprehensive benefit value is uniformly formatted in floating-point format and its numerical integrity is verified to avoid ranking distortion due to missing or outlier values. The ranking process uses a stable ranking algorithm to ensure output consistency. Safety verification is performed on the prompt parameter combination: The stimulation current is compared with the maximum stimulation current threshold, and the stimulation duration is compared with the maximum stimulation duration threshold. If the stimulation current is greater than the maximum stimulation current threshold or the stimulation duration is greater than the maximum stimulation duration threshold, the parameters exceeding the limit are truncated to the corresponding threshold. The maximum stimulation current threshold and the maximum stimulation duration threshold are derived from the rated safety upper limit of the stimulation device, preset according to the device's technical specifications, ensuring that the stimulation dose does not exceed the physiological safety range required by regulations. The truncation adopts an upper limit saturation strategy, ensuring that the parameters maintain an optimization trend without exceeding the safety boundary. Simultaneously, the changes in stimulation current and duration are calculated using differential methods and compared with the maximum current change threshold and the maximum duration change threshold, respectively. If either change exceeds the corresponding threshold, the corresponding parameter is truncated. The maximum current change threshold and the maximum duration change threshold are adaptively set based on patient tolerance and historical stimulation events to avoid adverse reactions caused by excessively rapid changes in stimulation dose. The differential calculation is based on the previous stimulation parameter recorded in the stimulation parameter optimization database, and timestamp matching ensures that the change comparison corresponds to the correct previous stimulation event. If an abnormal change is detected, it will be automatically corrected to a value within the maximum allowable range, ensuring the smoothness and safety of parameter adjustment. Parameter prompts are pushed based on the safety-verified prompt parameter combination, and the safety-verified prompt parameter combination and the corresponding comprehensive benefit value are written into the stimulation parameter optimization database. A transactional record structure with timestamps is used during writing to ensure a traceable link for the generation, verification, and implementation of each stimulation parameter.

[0065] In this implementation plan, by comprehensively ranking candidate stimulation parameters based on their benefits and conducting multi-level safety checks in conjunction with the safety boundaries of the stimulation device and patient tolerability, this method ensures that the recommended parameters not only maximize the expected stability improvement but also strictly comply with clinical safety thresholds. Simultaneously, by constraining the magnitude of parameter changes, the method ensures that the stimulation dose adjustment process is smooth and controllable, avoiding adverse reactions caused by excessively rapid changes. Furthermore, by implementing a transactional database writing mechanism to achieve traceable closed-loop recording, the optimized parameters possess high reliability in terms of safety, effectiveness, and feasibility.

[0066] The above embodiments can be implemented, in whole or in part, by software, hardware (such as circuits), firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, as a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, all or part of the processes or functions described in the embodiments of the present invention are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that includes one or more sets of available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium. A semiconductor medium can be a solid-state drive.

[0067] It should be understood that the term "and / or" in this article is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A alone, A and B simultaneously, or B alone. A and B can be singular or plural. Additionally, the character " / " in this article generally indicates an "or" relationship between the preceding and following related objects, but it can also represent an "and / or" relationship. Please refer to the context for a more accurate understanding.

[0068] In this invention, "at least one" means one or more, and "more than one" means two or more. "At least one of the following" or similar expressions refer to any combination of these items, including any combination of a single item or a plurality of items. For example, at least one of a, b, or c can represent: a, b, c, ab, ac, bc, or abc, where a, b, and c can be a single item or multiple items.

[0069] It should be understood that, in various embodiments of the present invention, the sequence number of each process does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0070] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0071] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the devices, apparatuses, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0072] In the several embodiments provided by this invention, it should be understood that the disclosed devices, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another device, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between devices or units may be electrical, mechanical, or other forms.

[0073] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0074] In addition, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0075] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0076] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A system for optimizing stimulation parameters of the subthalamic nucleus in Parkinson's disease based on temporal interference, characterized in that, The system includes: The multimodal data acquisition and preprocessing module is used to acquire patients' motor nerve data in real time on a periodic basis, and to perform filtering, normalization and time alignment on the motor nerve data; The gait electromyography stability assessment module is used to analyze gait rhythm fluctuations, postural sway amplitudes, and electromyographic morphological similarities based on motor neural data, and to generate stability indices for gait electromyography fusion; combined with the start and end times of stimulation, the stability indices are used to determine the improvement in stability before and after stimulation. The stimulus dose-response modeling module is used to combine historical motor neural data and corresponding stability improvement records to construct dose-response relationships, identify half-effect equivalent stimulus intensity and target stability level, infer the potential improvement trend of the current stimulus under different parameters, and generate candidate stimulus parameter combinations and corresponding improvement predictions. The comprehensive benefit parameter optimization module is used to fit the improved response time constant based on the stability improvement trajectory after historical stimuli and obtain the evaluation window; combine the improved response time constant and the evaluation window to infer the time coverage ratio, and comprehensively evaluate the comprehensive benefit by comprehensively predicting candidate improvements; select recommended stimulus parameters according to the comprehensive benefit ranking, and perform safety constraint verification.

2. The Parkinson's subthalamic nucleus stimulation parameter optimization system based on time-interference according to claim 1, characterized in that, The specific process of periodically collecting the patient's motor nerve data in real time, and performing filtering, normalization, and time alignment on the motor nerve data is as follows: Motor neural data were collected from patients during the gait cycle, including stride length, stance phase time, plantar pressure, surface electromyography (EMG) signals of muscle groups, stimulation start time stamp, stimulation end time stamp, stimulation duration, stimulation current, and electrode impedance. The gait cycle was defined as the time interval between two consecutive ipsilateral plantar impact events. The starting point of the gait cycle was determined by the sampling moment when the plantar pressure signal first changed from zero to non-zero, and the ending point was determined by the sampling moment when the ipsilateral plantar pressure signal changed from zero to non-zero again at subsequent sampling points. Low-pass filtering was applied to stride length, stance phase time, and plantar pressure to remove high-frequency noise. Band-pass filtering and rectification envelope algorithms were used to perform band-pass filtering, rectification, and envelope extraction on the surface electromyography (EMG) signals of muscle groups within the gait cycle, converting the original EMG waveforms into continuous and stable EMG envelope signal curves and identifying EMG envelope peaks. Motor neural data for each gait cycle were time-normalized to a uniform number of sampling points, and motor neural data from different gait cycles were aligned according to their corresponding time indices based on the normalized uniform time sampling points. z-score normalization was applied to the motor neural data to eliminate differences between different cycles. A stimulation parameter optimization database was constructed, in which the original and preprocessed motor neural data were written into the database, with gait cycles as the unit.

3. The Parkinson's subthalamic nucleus stimulation parameter optimization system based on time-interference according to claim 1, characterized in that, The specific process of analyzing gait rhythm fluctuations, postural sway amplitudes, and electromyographic morphological similarities based on motor neural data to form a stability index for gait electromyographic fusion is as follows: Read the motor neural data within N preprocessed gait cycles to obtain the stride length sequence and the support phase time series; calculate the mean and standard deviation of the stride length sequence to obtain the mean stride length and the standard deviation of the stride length, and divide the standard deviation of the stride length by the mean stride length to obtain the stride length variation coefficient. The support phase time mean and support phase time standard deviation are calculated based on the support phase time series. The support phase time variation coefficient is obtained by dividing the support phase time standard deviation by the support phase time mean. Simultaneously, for each gait cycle, the plantar pressure sequence within the current cycle is obtained, and the average value and standard deviation are calculated to obtain the average value of lateral sway and the amount of sway of the center of gravity; the average amount of sway of the center of gravity is obtained by averaging the amount of sway of the center of gravity over all gait cycles. Historical gait cycles in which the coefficient of variation of historical stride length, the coefficient of variation of support phase time, and the average center of gravity sway are all within the corresponding stable threshold range are selected from the stimulation parameter optimization database. The corresponding electromyographic envelope signals are extracted and averaged to obtain the reference electromyographic value. The electromyographic envelope signal of the current gait cycle is aligned point by point with the reference electromyographic value to calculate the cross-correlation coefficient and obtain the electromyographic morphological correlation coefficient. The sum of squares of the stride length variation coefficient and the support phase time variation coefficient is calculated, and the negative of the sum of squares is used as the exponent for natural exponential operation to obtain the gait temporal stability term. The current electromyographic morphological correlation coefficient is nonnegated and divided by the sum of a constant and the average value of the lateral sway of the center of gravity to obtain the electromyographic posture coupling term. The gait temporal stability term and the electromyographic posture coupling term are multiplied to obtain the comprehensive value of motion stability.

4. The Parkinson's subthalamic nucleus stimulation parameter optimization system based on time-interference according to claim 1, characterized in that, The specific process of determining the improvement in stability before and after stimulation by combining the start and end times of the stimulus and using stability indicators is as follows: Based on the stimulus start time stamp and stimulus end time stamp, the median of the comprehensive motion stability values ​​of the M gait cycles before the stimulus start is selected as the baseline motion stability value of the stimulus, and the median of the comprehensive motion stability values ​​of the M gait cycles after the stimulus end is selected as the post-stimulus motion stability value of the stimulus. The difference between the post-stimulation motor stability value and the baseline motor stability value is calculated as the measure of stability improvement after stimulation. The baseline motion stability value, post-stimulation motion stability value, and stability improvement for each stimulus are written into the stimulus parameter optimization database.

5. The Parkinson's subthalamic nucleus stimulation parameter optimization system based on time-interference according to claim 1, characterized in that, The specific process of combining historical motor neural data and corresponding stability improvement records to construct dose-response relationships and identify half-effect equivalent stimulus intensity and target stability level is as follows: The baseline motor stability value, post-stimulation motor stability value, and stability improvement amount corresponding to the current stimulus, as well as the stimulation current, electrode impedance, and stimulation duration, are read. Simultaneously, based on historical stimulation events in the stimulation parameter optimization database, the median of historical electrode impedance is selected as the reference impedance, and the P90 quantile of historical stability improvement is selected as the historical maximum improvement amount. Historical stimulation events with historical stability improvement amounts greater than the improvement threshold are selected, and the median of the corresponding post-stimulation motor stability value is calculated as the target motor stability value. The equivalent stimulus intensity value is obtained by adding a constant to the ratio of electrode impedance to reference impedance and dividing the stimulation current by the summation. Based on the equivalent stimulus intensity value and corresponding stability improvement of each historical stimulus event, an exponential saturation growth model is used to construct a dose-effect prediction function. The minimum and maximum values ​​of the historical equivalent stimulus intensity values ​​are used as the search interval. Each candidate value in the search interval is substituted into the dose-effect prediction function to calculate the historical predicted improvement. The sum of squared errors between the historical predicted improvement and the corresponding actual stability improvement is calculated. The candidate value that minimizes the sum of squared errors is selected as the half-effect equivalent stimulus intensity value.

6. The Parkinson's subthalamic nucleus stimulation parameter optimization system based on time-interference according to claim 1, characterized in that, The specific process for inferring the potential improvement trend of the current stimulus under different parameters is as follows: The baseline deviation is obtained by subtracting the target motion stability value from the baseline motion stability value of the current stimulus and then nonnegating it. The improvement potential term is obtained by subtracting the baseline deviation value from the historical maximum improvement. The current equivalent stimulus intensity value is divided by the half-effect equivalent stimulus intensity value, and the negative number is used as the exponent for natural exponentiation. The stimulus effect term is obtained by subtracting the natural exponentiation result from the constant. The improvement potential term and the stimulus effect term are multiplied to obtain the predicted stability improvement for the next stimulus.

7. The Parkinson's subthalamic nucleus stimulation parameter optimization system based on time-interference according to claim 1, characterized in that, The specific process for generating candidate stimulus parameter combinations and corresponding improved predictions is as follows: Based on the rated output range and rated adjustable step size of the stimulation device, a set of candidate stimulation parameter combinations is generated, including different stimulation currents and stimulation durations; for each set of candidate stimulation parameter combinations, the corresponding equivalent stimulation intensity value is calculated, and the predicted stability improvement amount is calculated to obtain a set of candidate predicted stability improvement amounts. The candidate stimulus parameter combinations and the improvement in candidate prediction stability are written into the stimulus parameter optimization database.

8. The Parkinson's subthalamic nucleus stimulation parameter optimization system based on time-interference according to claim 1, characterized in that, The specific process of fitting the improved response time constant based on the stability improvement trajectory after historical stimuli and obtaining the evaluation window is as follows: The time-varying sequence of post-stimulation motion stability values ​​corresponding to historical stimulus events is read from the stimulus parameter optimization database. Using the end timestamp of each stimulus as the time origin, and combining the corresponding baseline motion stability value and the historical maximum improvement, an exponential recovery model is used to fit the variation sequence. The parameters of the exponential recovery model are solved by a nonlinear least squares fitting method to obtain the time constant corresponding to each historical stimulus event. The time constants obtained for each historical stimulus event are statistically analyzed, and the median is selected as the improvement response time constant. Three times the improvement response time constant is used as the evaluation window duration.

9. The Parkinson's subthalamic nucleus stimulation parameter optimization system based on time-interference according to claim 1, characterized in that, The specific process of combining the improved response time constant and the evaluation window to infer the time coverage ratio, and comprehensively evaluating the overall benefits by integrating candidate improvement predictions, is as follows: The time weight value is obtained by dividing the improved response time constant by the evaluation window duration; the natural exponent is calculated by dividing the negative of the evaluation window duration by the improved response time constant, and the improvement time coverage ratio is obtained by subtracting the natural exponent from the constant; the improvement amount of candidate prediction stability is obtained, and the improvement amount of candidate prediction stability is multiplied by the time weight value and the improvement time coverage ratio to obtain the average improvement value of the evaluation window. Obtain the candidate stimulus parameter combination corresponding to each candidate predicted stability improvement amount, multiply the corresponding stimulus current squared by the corresponding stimulus duration to obtain the stimulus load value; add the stimulus load value to the constant and take the natural logarithm as the denominator; divide the average improvement value of the evaluation window by the denominator to obtain the comprehensive benefit value.

10. The Parkinson's subthalamic nucleus stimulation parameter optimization system based on time-interference according to claim 1, characterized in that, The specific process of selecting recommended stimulus parameters according to comprehensive benefit ranking and performing safety constraint verification is as follows: Calculate the comprehensive benefit value corresponding to the improvement in stability of all candidate predictions, sort them in descending order, and select the candidate stimulus parameter with the largest comprehensive benefit value as the cue parameter combination; A safety check is performed on the combination of prompt parameters: the stimulation current is compared with the maximum stimulation current threshold, and the stimulation duration is compared with the maximum stimulation duration threshold. If the stimulation current is greater than the maximum stimulation current threshold or the stimulation duration is greater than the maximum stimulation duration threshold, the parameters exceeding the limit are truncated to the corresponding thresholds. At the same time, the changes in stimulation current and stimulation duration are calculated differentially and compared with the maximum current change threshold and the maximum duration change threshold, respectively. If either change exceeds the corresponding change threshold, the corresponding parameter is truncated. Based on the verified combination of prompt parameters, parameter prompts are pushed out, and the verified combination of prompt parameters and the corresponding comprehensive benefit value are written into the stimulus parameter optimization database.

Images