A method and system for closed-loop control of neuroelectrical stimulation based on physiological parameter feedback

Abstract

The present application relates to the technical field of medical monitoring and diagnosis and treatment control, and provides a nerve electric stimulation closed-loop control method and system based on physiological parameter feedback, which comprises: collecting target physiological parameters in at least two different time scales, calculating the change amplitude, change rate and continuous consistency coefficient of the target physiological parameters based on the short-period parameter sequence, matching them with the preset credible interval, determining the effective physiological parameter change based on the short-period parameter sequence, and determining the number of delay periods corresponding to the physiological response in combination with the stimulation parameter historical adjustment record. The present application introduces the multi-time scale acquisition mechanism of short-period parameter sequence and long-period parameter sequence in the process of nerve electric stimulation, so that the transient change characteristics and long-term trend characteristics of physiological parameters are distinguished and processed, and abnormal feedback caused by random noise, transient fluctuation or non-physiological factors is effectively inhibited.

Description

Technical Field

[0001] This invention relates to the field of medical monitoring and treatment control technology, specifically to a closed-loop control method and system for neural electrical stimulation based on physiological parameter feedback. Background Technology

[0002] Transcutaneous electrical nerve stimulation (TENS) has been widely used in the adjunctive treatment of neurological function regulation, pain relief, and autonomic nervous system-related diseases due to its non-invasive nature. With the development of wearable devices and sensing technologies, some TENS systems have begun to incorporate physiological parameters as feedback, dynamically adjusting stimulation parameters through closed-loop control to improve treatment efficacy and reduce the need for manual intervention.

[0003] However, most existing closed-loop control methods for neural electrical stimulation based on physiological parameter feedback assume that changes in physiological parameters can stably reflect the stimulation effect. They typically rely directly on changes in parameters such as heart rate, pulse, or blood flow to continuously adjust the stimulation intensity or rhythm. In practical applications, the human physiological state is highly dynamic and nonlinear, and the response of physiological parameters to stimulation often exhibits time delays, amplitude fluctuations, and periodic drift.

[0004] Under dynamic stimulation conditions, when physiological parameters undergo nonlinear transitions within a short period, existing closed-loop control methods are prone to misinterpreting transient abnormal changes as valid feedback signals. This leads to continuous adjustment of the stimulation parameters in the wrong direction, causing the stimulation output to gradually deviate from the safe range. Simultaneously, the hysteresis effect of the physiological response to the stimulus accumulates continuously during multiple adjustments, easily causing the control system to form a hidden positive feedback loop, triggering self-excited oscillations in the stimulus intensity over multiple cycles.

[0005] Furthermore, during prolonged wear and continuous treatment sessions, the quality of physiological signal acquisition may slowly attenuate due to changes in skin condition, slight shifts in electrode position, and other factors. Existing closed-loop control methods typically cannot distinguish between signal attenuation and a weakening of the actual physiological response, thus unknowingly and continuously increasing the stimulation output. As the treatment cycle extends, the closed-loop control system continuously corrects the stimulation parameter range through multiple adaptive adjustments, gradually eroding the originally set safety boundaries. Ultimately, while still formally meeting the single-adjustment rules, the overall stimulation load approaches or exceeds the safety threshold range.

[0006] The aforementioned problems are not easily detected under normal use conditions, but once they occur in combination under specific physiological conditions or long-term operating scenarios, they may lead to uncontrolled stimulation intensity, amplified physiological inhibitory responses, or even potential safety risks, which seriously restricts the safe application of closed-loop control technology for neurostimulation based on physiological parameter feedback in clinical and home environments. Summary of the Invention

[0007] To address the aforementioned problems, this invention provides a closed-loop control method and system for neural electrical stimulation based on physiological parameter feedback.

[0008] The present invention adopts the following technical solution: a closed-loop control method for neural electrical stimulation based on physiological parameter feedback, comprising: step S1, during the neural electrical stimulation process, acquiring target physiological parameters at at least two different time scales, wherein the target physiological parameters include a short-period parameter sequence for transient change analysis and a long-period parameter sequence for trend stability analysis, and introducing time calibration identifiers for each parameter sequence;

[0009] Step S2: Based on the short-period parameter sequence, calculate the change amplitude, change rate and continuity consistency coefficient of the target physiological parameter, and match them with a preset confidence interval. When the change characteristics simultaneously meet the preset confidence interval, determine that the current physiological parameter change is an effective feedback and mark it as an effective physiological parameter change parameter.

[0010] Step S3: Obtain a short-cycle parameter sequence within each stimulation parameter adjustment cycle; determine the effective physiological parameter changes based on the short-cycle parameter sequence; determine the number of delay cycles D corresponding to the physiological response by combining the historical adjustment records of the stimulation parameters; within the j-th stimulation parameter adjustment cycle, call the effective physiological parameter change parameter corresponding to the jD-th stimulation parameter adjustment cycle as the feedback input quantity and input it into the stimulation parameter prediction model to obtain the predicted stimulation parameter adjustment quantity.

[0011] Step S4: Based on the long-period parameter sequence, calculate the baseline drift trend of physiological parameters, and combine the consistency judgment results of physiological response characteristics before and after stimulation to generate stimulation parameter adjustment instructions.

[0012] As a further description of the above technical solution: the method for collecting target physiological parameters includes:

[0013] Step S11: After the nerve electrical stimulation is initiated, at least two types of physiological parameter sampling channels are established, including short-cycle sampling channels and long-cycle sampling channels.

[0014] Step S12: Through the short-cycle sampling channel, multiple sampled values ​​of the target physiological parameters are acquired during continuous stimulation, and a short-cycle parameter sequence is constructed in chronological order.

[0015] Step S13: Through the long-period sampling channel, the target physiological parameters are periodically sampled within multiple stimulation cycles, and a long-period parameter sequence is constructed based on the sampling results;

[0016] Step S14: In the process of generating short-period parameter sequences and long-period parameter sequences, time calibration identifiers are introduced into the sampled data in each parameter sequence. The time calibration identifiers include at least a sampling time identifier and a relative stimulus period position identifier.

[0017] As a further description of the above technical solution: the short-cycle sampling channel uses the stimulation output cycle as the time reference, and samples the target physiological parameters at equal time intervals M times within each stimulation cycle, where M≥3;

[0018] Long-cycle sampling involves sampling the target physiological parameters once after P consecutive stimulation cycles (P≥1) during the nerve electrical stimulation process, using the stimulation cycle as the basic unit, thus forming a long-cycle sampling channel.

[0019] As a further description of the above technical solution: the target physiological parameters include physiological signal parameters that reflect the effect of nerve electrical stimulation or changes in physiological safety status.

[0020] As a further description of the above technical solution: the method for obtaining the predicted stimulus parameter adjustment amount includes:

[0021] Step S31: Set the stimulation parameter adjustment cycle as the basic control cycle for stimulation parameter updates, and perform multiple physiological parameter sampling operations within each stimulation parameter adjustment cycle to form a short-cycle parameter sequence consisting of multiple consecutive sampling cycles.

[0022] Step S32: Within each stimulation parameter adjustment cycle, the effective feedback judgment method of step S2 is applied to the acquired short-cycle parameter sequence to screen out effective physiological parameter change parameters that meet the amplitude constraint, rate constraint and consistency constraint.

[0023] Step S33: Statistical analysis is performed on the historical adjustment records of stimulation parameters and the sampling time of effective physiological parameter changes to determine the number of delay cycles D corresponding to the identifiable feedback of physiological parameters from the adjustment of stimulation parameters; the number of delay cycles represents the number of adjustment cycles spanned from the start of the adjustment of stimulation parameters to the first determination of the physiological parameter change as an effective feedback by S2.

[0024] Step S34: During the j-th stimulation parameter adjustment cycle, the effective physiological parameter change parameter corresponding to the jD-th stimulation parameter adjustment cycle is called as the feedback input of the current stimulation parameter adjustment, input into the pre-constructed stimulation parameter prediction model, and the predicted stimulation parameter adjustment amount is output.

[0025] As a further description of the above technical solution: the method for obtaining the baseline drift trend of the physiological parameters includes:

[0026] In the long-period parameter sequence, the trend of physiological parameters changing over time is fitted and analyzed to obtain trend characteristic quantities that characterize the overall direction of change of physiological parameters;

[0027] Preset baseline drift trend parameter gradient thresholds, including positive and negative thresholds;

[0028] When the trend feature value is greater than the preset positive threshold, it indicates that the physiological parameters are generally on an upward trend.

[0029] When the trend feature value is less than the preset negative threshold, it indicates that the physiological parameters are generally showing a downward trend.

[0030] When the trend characteristic is between positive and negative thresholds, it indicates that the physiological parameters are generally stable.

[0031] As a further description of the above technical solution: the method for generating stimulation parameter adjustment instructions by combining the consistency determination results of physiological response characteristics before and after stimulation is as follows:

[0032] Step S41: Before adjusting the stimulation parameters, obtain the long-term physiological parameter characteristics corresponding to the pre-stimulation stage; after adjusting the stimulation parameters, obtain the long-term physiological parameter characteristics corresponding to the post-stimulation stage; and calculate the consistency coefficient of the physiological response characteristics based on the pre-stimulation and post-stimulation physiological parameter characteristics.

[0033] Step S42: Within G consecutive long-cycle analysis windows, where G≥1, comprehensively analyze the baseline drift trend and consistency judgment results of physiological parameters to generate stimulation parameter adjustment instructions.

[0034] As a further description of the above technical solution: the method for generating stimulation parameter adjustment instructions includes:

[0035] When the baseline drift trend shows an upward or stable trend, and the consistency judgment result is greater than or equal to the preset consistency threshold, a stimulus parameter adjustment instruction is generated, and the stimulus parameter is adjusted based on the predicted stimulus parameter adjustment amount.

[0036] In other cases, no stimulus parameter adjustment instructions are generated, and the system enters a pause adjustment state.

[0037] As a further description of the above technical solution: the method for obtaining the consistency coefficient of the physiological response characteristics includes: calculating the mean, variance and variation amplitude of the physiological parameter sequences in the pre-stimulation stage and the post-stimulation stage respectively, which represent the physiological response characteristic vectors of their respective stages;

[0038] The differences in the corresponding physiological response feature vectors before and after stimulation are obtained separately. The differences of each feature are weighted and summed to obtain the consistency coefficient of the physiological response features.

[0039] A closed-loop control system for neural electrical stimulation based on physiological parameter feedback, used to implement the aforementioned closed-loop control method for neural electrical stimulation based on physiological parameter feedback, includes:

[0040] The multi-scale parameter acquisition module acquires target physiological parameters at at least two different time scales during nerve electrical stimulation. The target physiological parameters include a short-period parameter sequence for transient change analysis and a long-period parameter sequence for trend stability analysis, and introduces time calibration labels for each parameter sequence.

[0041] The effective feedback determination module calculates the change amplitude, change rate and continuity consistency coefficient of the target physiological parameter based on the short-period parameter sequence, and matches them with a preset confidence interval. When the change characteristics simultaneously meet the preset confidence interval, the current physiological parameter change is determined to be effective feedback and marked as an effective physiological parameter change parameter.

[0042] The delayed response matching module acquires a short-cycle parameter sequence within each stimulus parameter adjustment cycle, determines the effective physiological parameter changes based on the short-cycle parameter sequence, and determines the number of delayed cycles D corresponding to the physiological response by combining the historical adjustment records of the stimulus parameters; within the j-th stimulus parameter adjustment cycle, it calls the effective physiological parameter change parameter corresponding to the jD-th stimulus parameter adjustment cycle as a feedback input, inputs it into the stimulus parameter prediction model, and obtains the predicted stimulus parameter adjustment amount;

[0043] The baseline drift control module calculates the baseline drift trend of physiological parameters based on long-period parameter sequences, and generates stimulation parameter adjustment instructions by combining the consistency judgment results of physiological response characteristics before and after stimulation.

[0044] Beneficial effects:

[0045] This invention introduces a multi-timescale acquisition mechanism of short-cycle and long-cycle parameter sequences during neural electrical stimulation, enabling the differentiation between transient and long-term trend characteristics of physiological parameters. By calculating the amplitude, rate of change, and continuity consistency coefficient of the short-cycle parameter sequences and performing joint matching judgment in conjunction with preset confidence intervals, it effectively suppresses abnormal feedback caused by random noise, transient fluctuations, or non-physiological factors, avoids misjudging invalid physiological changes as valid feedback, and thus significantly improves the reliability of physiological feedback identification in closed-loop regulation.

[0046] In the closed-loop adjustment of stimulus parameters, a physiological response delay compensation mechanism is introduced. By statistically analyzing the historical adjustment records of stimulus parameters and the occurrence time of effective physiological parameter changes, the number of delay cycles corresponding to the effect of the stimulus parameters is determined. During the adjustment process, physiological feedback aligned with the delay is used as the control input. This approach effectively solves the problem of time lag between physiological response and changes in stimulus parameters, avoiding repeated adjustments or over-adjustment of stimulus parameters due to misjudgment caused by immediate feedback, and improving the stability of the closed-loop control process.

[0047] Analyzing the baseline drift trend of physiological parameters based on long-period parameter sequences and using it as an important basis for stimulus parameter adjustment decisions can identify the overall shift state of physiological parameters over a longer time scale. When a persistent shift in the baseline of physiological parameters or a deviation of the response characteristics from the expected pattern is detected, a consistency judgment mechanism is used to limit further enhancement of stimulus parameters, effectively preventing cumulative drift or instability of closed-loop control during long-term operation and improving the safety of continuous system operation. Attached Figure Description

[0048] The present invention will be further explained below with reference to the accompanying drawings and embodiments:

[0049] Figure 1 A flowchart of a closed-loop control method for neural electrical stimulation based on physiological parameter feedback provided in Embodiment 1 of the present invention;

[0050] Figure 2 This is a flowchart of the method for collecting target physiological parameters provided in Embodiment 1 of the present invention;

[0051] Figure 3 This is a flowchart of the method for obtaining the predicted stimulus parameter adjustment amount provided in Embodiment 1 of the present invention;

[0052] Figure 4 The flowchart of the method for generating stimulation parameter adjustment instructions based on the consistency determination results of physiological response characteristics before and after stimulation, as provided in Embodiment 1 of the present invention;

[0053] Figure 5 This is a module connection diagram of a closed-loop control system for nerve electrical stimulation based on physiological parameter feedback provided in Embodiment 2 of the present invention. Detailed Implementation

[0054] To make the technical means, creative features, objectives, and effects of this invention readily understandable, the invention is further described below with reference to specific illustrations. It should be noted that, unless otherwise specified, the embodiments and features described in these embodiments can be combined with each other.

[0055] Example 1

[0056] Please see Figures 1-4 This invention provides a technical solution: a closed-loop control method for neural electrical stimulation based on physiological parameter feedback, comprising:

[0057] Step S1: During the nerve electrical stimulation process, target physiological parameters are acquired at at least two different time scales. The target physiological parameters include a short-period parameter sequence for transient change analysis and a long-period parameter sequence for trend stability analysis. Time calibration labels are introduced for each parameter sequence.

[0058] Specifically: methods for acquiring target physiological parameters at at least two different time scales include:

[0059] Step S11: After the nerve electrical stimulation is initiated, at least two types of physiological parameter sampling channels are established, including short-cycle sampling channels and long-cycle sampling channels.

[0060] Specifically, the short-cycle sampling channel uses the stimulation output cycle as the time reference and performs M (M≥3) samplings of the target physiological parameters at equal time intervals within each stimulation cycle.

[0061] Long-cycle sampling involves sampling the target physiological parameters once after P (P≥1) consecutive stimulation cycles during the nerve electrical stimulation process, using the stimulation cycle as the basic unit, thus forming a long-cycle sampling channel.

[0062] It should be noted that the short-cycle sampling channel is used to capture the rapid changes in physiological parameters under stimulation; the long-cycle sampling channel is used to reflect the overall trend of physiological parameters over multiple stimulation cycles. The sampling period of the short-cycle sampling channel is significantly shorter than that of the long-cycle sampling channel, thus distinguishing the two types of sampling channels in terms of time resolution.

[0063] Step S12: Through the short-cycle sampling channel, multiple sampled values ​​of the target physiological parameters are acquired during continuous stimulation, and a short-cycle parameter sequence is constructed in chronological order. Specifically, the short-cycle parameter sequence is used to reflect the amplitude, rate of change, and fluctuation characteristics of the physiological parameters in adjacent stimulation cycles, so as to capture transient responses and abnormal transition behaviors triggered by stimulation.

[0064] Step S13: Through the long-cycle sampling channel, the target physiological parameters are periodically sampled within multiple stimulation cycles, and a long-cycle parameter sequence is constructed based on the sampling results. Specifically, the long-cycle parameter sequence is used to describe the overall trend of physiological parameters over a longer time range, in order to characterize the stability changes of physiological state, physiological adaptation process and baseline drift characteristics.

[0065] Step S14: During the generation of short-period and long-period parameter sequences, time calibration identifiers are introduced into the sampled data of each parameter sequence. These time calibration identifiers include at least a sampling time identifier and a relative stimulus cycle position identifier. Specifically, through time calibration, the short-period and long-period parameter sequences are aligned on the time axis to support subsequent time-related physiological response analysis.

[0066] The relative stimulation cycle position identifier refers to marking the stimulation cycle in which each data point is located during the sampling process, for example, the first stimulation cycle, the second stimulation cycle, etc.

[0067] The target physiological parameters include physiological signal parameters that reflect the effects of nerve electrical stimulation or changes in physiological safety status, including parameters characterizing cardiovascular status, autonomic nerve activity, local tissue response, and skin electrical properties.

[0068] Specifically, the target physiological parameters include heart rate, pulse, and skin impedance. The heart rate parameter is acquired using a photoplethysmography (PPG) sensor, the pulse parameter is acquired using a PPG sensor, and the skin impedance parameter is acquired using an impedance detection electrode.

[0069] Step S2: Based on the short-period parameter sequence, calculate the change amplitude, change rate and continuity consistency coefficient of the target physiological parameter, and match them with a preset confidence interval. When the change characteristics simultaneously meet the preset confidence interval, determine that the current physiological parameter change is an effective feedback and mark it as an effective physiological parameter change parameter.

[0070] It should be noted that the preset confidence interval is set by those skilled in the art. It is used to limit the reasonable range of changes that physiological parameters may undergo during nerve electrical stimulation. The setting is based on the historical statistical characteristics, physiological stability characteristics, and stimulation condition characteristics of the target physiological parameters.

[0071] For example, suppose the target physiological parameter is heart rate (HR), the sampling period is 1 second, and the stimulation period is 10 seconds.

[0072] Short-period parameter sequence: Heart rate data (unit: beats / minute) obtained by sampling for 10 consecutive seconds:

[0073] [72,74,75,76,77,77,78,79,79,80];

[0074] Variation range = maximum value - minimum value = 8;

[0075] Rate of change = Change in adjacent sample values ​​ / Time interval = (2 + 1 + 1 + 1 + 0 + 1 + 1 + 0 + 1) / 9 ≈ 0.89 times / second;

[0076] Calculate the continuity coefficient. If the ascending or descending sequence is [+,+,+,+,0,+,+,0,+], then the continuity coefficient is approximately 7 / 9 = 0.78.

[0077] If the amplitude, rate of change, and consistency coefficient are all within the preset range, then the heart rate change in that cycle is determined to be a valid physiological parameter change, and the parameter is marked as a valid physiological parameter change parameter.

[0078] Methods for obtaining the magnitude of change include: obtaining the maximum and minimum parameter values ​​within the short-period parameter sequence, and calculating the difference between them as the magnitude of change;

[0079] The method for obtaining the rate of change includes: obtaining the amount of change between adjacent sampled parameter values ​​in a short-period parameter sequence, normalizing the amount of change of each adjacent sampled parameter value with the corresponding short-period sampling time interval, and performing statistical processing on the normalized amount of change, using the average value of the normalized amount of change as the rate of change of the physiological parameter.

[0080] The method for obtaining the continuity consistency coefficient includes: obtaining the direction of change of adjacent sampled parameter values ​​in a short-period parameter sequence, counting the number of times the direction of change of adjacent sampled parameter values ​​is consistent, and calculating the ratio of the number of times the direction of change is consistent to the total number of adjacent sample pairs in the short-period parameter sequence, and using the obtained ratio as the continuity consistency coefficient of physiological parameter changes.

[0081] In this embodiment, by introducing a multi-timescale acquisition mechanism of short-cycle parameter sequences and long-cycle parameter sequences during neural electrical stimulation, the transient change characteristics and long-term trend characteristics of physiological parameters can be distinguished and processed. By calculating the change amplitude, change rate and continuity consistency coefficient of the short-cycle parameter sequences and performing joint matching judgment in combination with preset confidence intervals, abnormal feedback caused by random noise, transient fluctuations or non-physiological factors can be effectively suppressed, and ineffective physiological changes can be avoided from being misjudged as effective feedback, thereby significantly improving the reliability of physiological feedback recognition in closed-loop regulation.

[0082] Step S3: Obtain a short-cycle parameter sequence within each stimulation parameter adjustment cycle; determine the effective physiological parameter changes based on the short-cycle parameter sequence; determine the number of delay cycles D corresponding to the physiological response by combining the historical adjustment records of the stimulation parameters; within the j-th stimulation parameter adjustment cycle, call the effective physiological parameter change parameter corresponding to the jD-th stimulation parameter adjustment cycle as the feedback input quantity and input it into the stimulation parameter prediction model to obtain the predicted stimulation parameter adjustment quantity.

[0083] It should be noted that since physiological responses are not instantaneous, directly using physiological changes in the current cycle can lead to prediction bias. By using the jD method, the changes in effective physiological parameters are correctly correlated with the historical adjustment of stimulus parameters, thus achieving closed-loop control based on time correlation.

[0084] The method for obtaining the predicted stimulus parameter adjustment includes:

[0085] Step S31: Set the stimulation parameter adjustment cycle as the basic control cycle for stimulation parameter updates, and perform multiple physiological parameter sampling operations within each stimulation parameter adjustment cycle to form a short-cycle parameter sequence consisting of multiple consecutive sampling cycles.

[0086] Step S32: Within each stimulation parameter adjustment cycle, the effective feedback judgment method of step S2 is applied to the acquired short-cycle parameter sequence to screen out effective physiological parameter change parameters that meet the amplitude constraint, rate constraint and consistency constraint.

[0087] Step S33: Statistical analysis is performed on the historical adjustment records of stimulation parameters and the sampling time of effective physiological parameter changes to determine the number of delay cycles D corresponding to the identifiable feedback of physiological parameters generated by the adjustment of stimulation parameters.

[0088] The number of delay cycles represents the number of adjustment cycles that span from the time the stimulus parameter adjustment takes effect until the physiological parameter change is first determined to be effective feedback in step S2.

[0089] In step S34, during the j-th stimulation parameter adjustment cycle, instead of directly using the physiological parameter change results obtained in the current sampling cycle, the effective physiological parameter change parameters corresponding to the jD-th stimulation parameter adjustment cycle are called as the feedback input of the current stimulation parameter adjustment and input into the pre-built stimulation parameter prediction model, and the predicted stimulation parameter adjustment amount is output.

[0090] The training method for the stimulus parameter prediction model includes:

[0091] Obtain historical stimulation parameter adjustment records, which include effective physiological parameter changes for the number of delay cycles D and the corresponding stimulation parameter adjustment amount for the j-th stimulation regulation cycle. Construct a dataset and divide the dataset into a training set, a validation set, and a test set according to a preset ratio (7:2:1).

[0092] Gradient boosting regression tree was selected as the stimulus parameter prediction model.

[0093] Set the initial hyperparameters, including: number of decision trees: 150-200; maximum depth of a single tree: 4-6; minimum number of samples for node splitting: 15-20; maximum number of features considered during splitting: 2 / 3 of the total number of features; learning rate: 0.05-0.1; loss function: mean squared error; regularization coefficient (L2): 0.1-0.2; single tree splitting criterion: mean squared error.

[0094] The initial predicted values ​​of the gradient boosting regression tree model are set as the training set labels, i.e., the mean of the stimulus parameter adjustment amounts, to provide a benchmark for subsequent iterative training. For each sample in the training set, the difference between the stimulus parameter adjustment amount predicted by the current model and the actual adjustment amount (i.e., the residual) is calculated.

[0095] A new decision tree is constructed with the goal of minimizing the residual. Based on the mean squared error, the optimal splitting feature, such as the physiological parameter change of the delay cycle number D, and the splitting point are selected to divide the samples into different child nodes until the stopping condition is met, such as reaching the maximum depth or the number of samples in the child node is less than the minimum number of samples. The weights of the leaf nodes of the new decision tree are optimized by gradient descent, and the contribution of the new tree to the final prediction result is controlled by the learning rate. The new tree is then integrated into the existing model to correct the previous prediction bias.

[0096] Repeat the residual calculation, fitting, and update steps until a preset number of decision trees are built, or trigger the early stopping mechanism.

[0097] The Bayesian optimization method is used to search for the optimal combination of hyperparameters within a preset range; optimization range: number of decision trees: 120-250; maximum depth of a single tree: 3-7; learning rate: 0.03-0.12; minimum number of samples for node splitting: 10-25; regularization coefficient (L2): 0.05-0.25;

[0098] Every 20 trees are trained in each iteration, the root mean square error of the validation set is calculated. When the root mean square error of the validation set decreases by less than 0.001 for three consecutive iterations, training is stopped to avoid the model overfitting the specific conditions of the training set.

[0099] Save the model parameters (including the splitting rules of all decision trees and the weights of leaf nodes) that have the lowest root mean square error in the validation set during training.

[0100] The model is tested using a test set. When the root mean square error is ≤5% and the mean absolute error is ≤3%, the trained model is exported as a lightweight format, such as pkl or ONNX, and integrated into the stimulus parameter control system for use.

[0101] In this embodiment, a physiological response delay compensation mechanism is introduced during the closed-loop adjustment of stimulation parameters. By statistically analyzing the historical adjustment records of stimulation parameters and the occurrence time of effective physiological parameter changes, the number of delay cycles corresponding to the effect of stimulation parameters is determined. During the adjustment process, physiological feedback after delay alignment is used as the control input. In this way, the problem of time lag between physiological response and changes in stimulation parameters is effectively solved, avoiding repeated adjustments or over-adjustment of stimulation parameters due to misjudgment of immediate feedback, and improving the stability of the closed-loop control process.

[0102] Step S4: Based on the long-period parameter sequence, calculate the baseline drift trend of physiological parameters, and combine the consistency judgment results of physiological response characteristics before and after stimulation to generate stimulation parameter adjustment instructions.

[0103] The methods for obtaining the baseline drift trend of the physiological parameters include:

[0104] In the long-period parameter sequence, the trend of physiological parameters changing over time is fitted and analyzed to obtain trend characteristic quantities that characterize the overall direction of change of physiological parameters. Specifically, the trend characteristic quantities are used to reflect the direction and degree of change of physiological parameters relative to the baseline reference value over a long time scale. That is, a trend characteristic quantity > 0 indicates that the physiological parameters are generally on an upward trend, a trend characteristic quantity < 0 indicates that the physiological parameters are generally on a downward trend, and a trend characteristic quantity ≈ 0 indicates that the physiological parameters are generally stable.

[0105] Example of obtaining trend features: long-period parameter sequences This represents the set of statistical values ​​of the target physiological parameter obtained within multiple consecutive long-period sampling points, used to analyze the changing trend of the physiological parameter over time.

[0106] The trend function representing the overall direction of change of the fitted physiological parameters is as follows:

[0107] ;

[0108] ;

[0109] In the formula, For the first Statistical values ​​of physiological parameters corresponding to each long-period sampling time. This refers to the number of sampling points within the long-term analysis window. The slope of the fitted line for linear regression is the trend characteristic. The intercept of the fitted line for linear regression is given. For the fitting residual, let represent the th The deviation between the actual physiological parameter values ​​of each sampling point and the linear fitting line is used to measure the fitting error and short-term fluctuations in physiological parameters.

[0110] Preset baseline drift trend parameter gradient thresholds, including positive and negative thresholds;

[0111] When the trend feature value is greater than the preset positive threshold, it indicates that the physiological parameters are generally on an upward trend.

[0112] When the trend feature value is less than the preset negative threshold, it indicates that the physiological parameters are generally showing a downward trend.

[0113] When the trend characteristic is between the positive and negative thresholds, it indicates that the overall change in physiological parameters is not significant, and the baseline drift state is judged to be stable.

[0114] It should be noted that the setting of the gradient threshold for the baseline drift trend parameter is based on the historical statistical characteristics of the target physiological parameter and experimental conditions. For example, the maximum rate of increase and decrease of the physiological parameter under specific stimulus conditions can be used as a reference for setting positive and negative thresholds. Specifically, the positive threshold can be set as a certain percentage increment of the maximum rate of increase, and the negative threshold can be set as a certain percentage decrease of the maximum rate of decrease. Through these settings, significant changes and minor fluctuations in the physiological parameter can be effectively distinguished, ensuring that the system accurately determines changes in the physiological parameter during feedback analysis.

[0115] The method for generating stimulation parameter adjustment instructions by combining the consistency determination results of physiological response characteristics before and after stimulation is as follows:

[0116] Step S41: Before adjusting the stimulation parameters, obtain the long-term physiological parameter characteristics corresponding to the pre-stimulation stage; after adjusting the stimulation parameters, obtain the long-term physiological parameter characteristics corresponding to the post-stimulation stage; and calculate the consistency coefficient of the physiological response characteristics based on the pre-stimulation and post-stimulation physiological parameter characteristics.

[0117] Specifically, the consistency determination results of physiological response characteristics are used to characterize the degree of matching between the actual physiological response pattern and the expected response pattern.

[0118] The method for obtaining the consistency coefficient of the physiological response features includes: calculating the mean, variance and magnitude of change of the physiological parameter sequences in the pre-stimulation stage and the post-stimulation stage, respectively, to represent the physiological response feature vectors of each stage.

[0119] The differences in the corresponding physiological response feature vectors before and after stimulation are obtained respectively. The differences of each feature are weighted and summed to obtain the consistency coefficient of the physiological response features.

[0120] Optionally, the formula for calculating the consistency coefficient is: In the formula The consistency coefficient of physiological response characteristics. The number of features For the first The weight coefficients of each feature For the first The relative difference values ​​of each feature.

[0121] It should be noted that the formulas mentioned above are all numerical calculations after removing dimensions. They are formulas that are closest to the real situation, obtained by software simulation based on a large amount of data. The weight coefficients in the formulas were obtained by those skilled in the art through fitting historical sample data or empirical calibration.

[0122] Step S42: Within G consecutive long-cycle analysis windows, where G≥1, comprehensively analyze the baseline drift trend and consistency judgment results of physiological parameters to generate stimulation parameter adjustment instructions.

[0123] The method for generating stimulus parameter adjustment instructions includes:

[0124] When the baseline drift trend shows an upward or stable trend, and the consistency judgment result is greater than or equal to the preset consistency threshold, a stimulus parameter adjustment instruction is generated, and the stimulus parameter is adjusted based on the predicted stimulus parameter adjustment amount.

[0125] It is required that the consistency threshold be obtained by those skilled in the art through fitting historical sample data or empirical calibration.

[0126] In other cases, no stimulus parameter adjustment instructions are generated, and the system enters a pause adjustment state.

[0127] In this embodiment, the baseline drift trend of physiological parameters is analyzed based on long-period parameter sequences and used as an important basis for stimulus parameter adjustment decisions. This enables the identification of the overall deviation state of physiological parameters over a longer time scale. When a persistent deviation of the physiological parameter baseline or a deviation of the response characteristics from the expected pattern is detected, a consistency judgment mechanism is used to limit further enhancement of stimulus parameter adjustment, effectively preventing cumulative drift or instability of closed-loop control during long-term operation and improving the safety of continuous system operation.

[0128] Example 2

[0129] Please see Figure 5 This invention provides a technical solution: a closed-loop control system for neural electrical stimulation based on physiological parameter feedback, used to implement the aforementioned closed-loop control method for neural electrical stimulation based on physiological parameter feedback, comprising:

[0130] The multi-scale parameter acquisition module acquires target physiological parameters at at least two different time scales during nerve electrical stimulation. The target physiological parameters include a short-period parameter sequence for transient change analysis and a long-period parameter sequence for trend stability analysis, and introduces time calibration labels for each parameter sequence.

[0131] The effective feedback determination module calculates the change amplitude, change rate and continuity consistency coefficient of the target physiological parameter based on the short-period parameter sequence, and matches them with a preset confidence interval. When the change characteristics simultaneously meet the preset confidence interval, the current physiological parameter change is determined to be effective feedback and marked as an effective physiological parameter change parameter.

[0132] The delayed response matching module acquires a short-cycle parameter sequence within each stimulus parameter adjustment cycle, determines the effective physiological parameter changes based on the short-cycle parameter sequence, and determines the number of delayed cycles D corresponding to the physiological response by combining the historical adjustment records of the stimulus parameters; within the j-th stimulus parameter adjustment cycle, it calls the effective physiological parameter change parameter corresponding to the jD-th stimulus parameter adjustment cycle as a feedback input, inputs it into the stimulus parameter prediction model, and obtains the predicted stimulus parameter adjustment amount;

[0133] The baseline drift control module calculates the baseline drift trend of physiological parameters based on long-period parameter sequences, and generates stimulation parameter adjustment instructions by combining the consistency judgment results of physiological response characteristics before and after stimulation.

[0134] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of the present invention is defined by the appended claims and their equivalents.

Claims

1. A closed-loop control method for neural electrical stimulation based on physiological parameter feedback, characterized in that, include: Step S1: During the nerve electrical stimulation process, target physiological parameters are acquired at at least two different time scales. The target physiological parameters include a short-period parameter sequence for transient change analysis and a long-period parameter sequence for trend stability analysis. Time calibration labels are introduced for each parameter sequence. Step S2: Based on the short-period parameter sequence, calculate the change amplitude, change rate and continuity consistency coefficient of the target physiological parameter, and match them with a preset confidence interval. When the change characteristics simultaneously meet the preset confidence interval, determine that the current physiological parameter change is an effective feedback and mark it as an effective physiological parameter change parameter. Step S3: Obtain a short-cycle parameter sequence within each stimulation parameter adjustment cycle, determine the effective physiological parameter changes based on the short-cycle parameter sequence, determine the number of delay cycles D corresponding to the physiological response by combining the historical adjustment record of the stimulation parameter, and within the j-th stimulation parameter adjustment cycle, call the effective physiological parameter change parameter corresponding to the jD-th stimulation parameter adjustment cycle as the feedback input quantity and input it into the stimulation parameter prediction model to obtain the predicted stimulation parameter adjustment quantity. Step S4: Based on the long-period parameter sequence, calculate the baseline drift trend of physiological parameters, and combine the consistency judgment results of physiological response characteristics before and after stimulation to generate stimulation parameter adjustment instructions.

2. The closed-loop control method for neural electrical stimulation based on physiological parameter feedback according to claim 1, characterized in that, The method for collecting the target physiological parameters includes: Step S11: After the nerve electrical stimulation is initiated, at least two types of physiological parameter sampling channels are established, including short-cycle sampling channels and long-cycle sampling channels. Step S12: Through the short-cycle sampling channel, multiple sampled values ​​of the target physiological parameters are acquired during continuous stimulation, and a short-cycle parameter sequence is constructed in chronological order. Step S13: Through the long-period sampling channel, the target physiological parameters are periodically sampled within multiple stimulation cycles, and a long-period parameter sequence is constructed based on the sampling results; Step S14: In the process of generating short-period parameter sequences and long-period parameter sequences, time calibration identifiers are introduced into the sampled data in each parameter sequence. The time calibration identifiers include at least a sampling time identifier and a relative stimulus period position identifier.

3. The closed-loop control method for neural electrical stimulation based on physiological parameter feedback according to claim 2, characterized in that, The short-cycle sampling channel uses the stimulation output cycle as the time reference and performs M samplings of the target physiological parameters at equal time intervals within each stimulation cycle, where M≥3. Long-cycle sampling involves sampling the target physiological parameters once after P consecutive stimulation cycles (P≥1) during the nerve electrical stimulation process, using the stimulation cycle as the basic unit, thus forming a long-cycle sampling channel.

4. The closed-loop control method for neural electrical stimulation based on physiological parameter feedback according to claim 2, characterized in that, The target physiological parameters include physiological signal parameters that reflect changes in the effect of nerve electrical stimulation or physiological safety status.

5. The closed-loop control method for neural electrical stimulation based on physiological parameter feedback according to claim 1, characterized in that, The method for obtaining the predicted stimulus parameter adjustment includes: Step S31: Set the stimulation parameter adjustment cycle as the basic control cycle for stimulation parameter updates, and perform multiple physiological parameter sampling operations within each stimulation parameter adjustment cycle to form a short-cycle parameter sequence consisting of multiple consecutive sampling cycles. Step S32: Within each stimulation parameter adjustment cycle, the effective feedback judgment method of step S2 is applied to the acquired short-cycle parameter sequence to screen out effective physiological parameter change parameters that meet the amplitude constraint, rate constraint and consistency constraint. Step S33: Statistical analysis is performed on the historical adjustment records of stimulation parameters and the sampling time of effective physiological parameter changes to determine the number of delay cycles D corresponding to the identifiable feedback of physiological parameters from the adjustment of stimulation parameters; the number of delay cycles represents the number of adjustment cycles spanned from the start of the adjustment of stimulation parameters to the first determination of the physiological parameter change as an effective feedback by S2. Step S34: During the j-th stimulation parameter adjustment cycle, the effective physiological parameter change parameter corresponding to the jD-th stimulation parameter adjustment cycle is called as the feedback input of the current stimulation parameter adjustment, input into the pre-constructed stimulation parameter prediction model, and the predicted stimulation parameter adjustment amount is output.

6. The closed-loop control method for neural electrical stimulation based on physiological parameter feedback according to claim 1, characterized in that, The methods for obtaining the baseline drift trend of the physiological parameters include: In the long-period parameter sequence, the trend of physiological parameters changing over time is fitted and analyzed to obtain trend characteristic quantities that characterize the overall direction of change of physiological parameters; Preset baseline drift trend parameter gradient thresholds, including positive and negative thresholds; When the trend feature value is greater than the preset positive threshold, it indicates that the physiological parameters are generally on an upward trend. When the trend feature value is less than the preset negative threshold, it indicates that the physiological parameters are generally showing a downward trend. When the trend characteristic is between positive and negative thresholds, it indicates that the physiological parameters are generally stable.

7. The closed-loop control method for neural electrical stimulation based on physiological parameter feedback according to claim 6, characterized in that, The method for generating stimulation parameter adjustment instructions by combining the consistency determination results of physiological response characteristics before and after stimulation is as follows: Step S41: Before adjusting the stimulation parameters, obtain the long-term physiological parameter characteristics corresponding to the pre-stimulation stage; after adjusting the stimulation parameters, obtain the long-term physiological parameter characteristics corresponding to the post-stimulation stage; and calculate the consistency coefficient of the physiological response characteristics based on the pre-stimulation and post-stimulation physiological parameter characteristics. Step S42: Within G consecutive long-cycle analysis windows, where G≥1, comprehensively analyze the baseline drift trend and consistency judgment results of physiological parameters to generate stimulation parameter adjustment instructions.

8. The closed-loop control method for neural electrical stimulation based on physiological parameter feedback according to claim 7, characterized in that, The method for generating stimulus parameter adjustment instructions includes: When the baseline drift trend shows an upward or stable trend, and the consistency judgment result is greater than or equal to the preset consistency threshold, a stimulus parameter adjustment instruction is generated, and the stimulus parameter is adjusted based on the predicted stimulus parameter adjustment amount. In other cases, no stimulus parameter adjustment instructions are generated, and the system enters a pause adjustment state.

9. The closed-loop control method for neural electrical stimulation based on physiological parameter feedback according to claim 7, characterized in that, The method for obtaining the consistency coefficient of the physiological response features includes: calculating the mean, variance and magnitude of change of the physiological parameter sequences in the pre-stimulation stage and the post-stimulation stage, respectively, to represent the physiological response feature vectors of each stage. The differences in the corresponding physiological response feature vectors before and after stimulation are obtained separately. The differences of each feature are weighted and summed to obtain the consistency coefficient of the physiological response features.

10. A closed-loop control system for neural electrical stimulation based on physiological parameter feedback, used to implement the closed-loop control method for neural electrical stimulation based on physiological parameter feedback as described in any one of claims 1-9, characterized in that, include: The multi-scale parameter acquisition module acquires target physiological parameters at at least two different time scales during nerve electrical stimulation. The target physiological parameters include a short-period parameter sequence for transient change analysis and a long-period parameter sequence for trend stability analysis, and introduces time calibration labels for each parameter sequence. The effective feedback determination module calculates the change amplitude, change rate and continuity consistency coefficient of the target physiological parameter based on the short-period parameter sequence, and matches them with a preset confidence interval. When the change characteristics simultaneously meet the preset confidence interval, the current physiological parameter change is determined to be effective feedback and marked as an effective physiological parameter change parameter. The delayed response matching module acquires a short-cycle parameter sequence within each stimulus parameter adjustment cycle, determines the effective physiological parameter changes based on the short-cycle parameter sequence, and determines the number of delayed cycles D corresponding to the physiological response by combining the historical adjustment records of the stimulus parameters; within the j-th stimulus parameter adjustment cycle, it calls the effective physiological parameter change parameter corresponding to the jD-th stimulus parameter adjustment cycle as a feedback input, inputs it into the stimulus parameter prediction model, and obtains the predicted stimulus parameter adjustment amount; The baseline drift control module calculates the baseline drift trend of physiological parameters based on long-period parameter sequences, and generates stimulation parameter adjustment instructions by combining the consistency judgment results of physiological response characteristics before and after stimulation.

Images