Intelligent watch individualized tremor treatment scheme monitoring system and method

Abstract

The invention provides a smartwatch individualized tremor treatment scheme monitoring system and method, and relates to the technical field of scheme monitoring. Physiological data and motion tremor data of a user are collected and analyzed through a smartwatch, and combined analysis is performed according to obtained time sequence user tremor analysis data and time sequence user physiological analysis data; obtaining time sequence user combination analysis feature data; performing electrical stimulation data feature extraction and weight adjustment on the time sequence user combination analysis feature data to obtain updated analysis feature data, training and obtaining an individual response model according to the updated analysis feature data, and performing feedback comparative analysis of multiple electrical stimulation stages according to the individual response model to obtain an electrical stimulation result. And performing effect data monitoring on the tremor treatment scheme according to a feedback comparative analysis result to obtain effect analysis data. According to the method, multi-dimensional information can be deeply fused, the state change of the user is dynamically tracked, and the optimal treatment scheme is automatically recommended.

Description

Technical Field

[0001] This invention proposes a smartwatch-based personalized tremor treatment plan monitoring system and method, which relates to the field of treatment plan monitoring technology, specifically to the field of smartwatch-based personalized tremor treatment plan monitoring technology. Background Technology

[0002] Tremor is a typical symptom of neurological disorders such as Parkinson's disease and essential tremor, severely impacting patients' daily lives. Traditional smartwatches use electrical stimulation for physical intervention. However, the intensity and frequency of tremors are dynamically influenced by the user's physiological state, while electrical stimulation physical interventions mostly use fixed or manually adjustable stimulation parameters, lacking adaptability and failing to meet individual needs. This forces patients to undergo a lengthy trial-and-error process to find temporarily effective parameters, making it difficult to continuously optimize treatment outcomes. Summary of the Invention

[0003] This invention provides a smartwatch-based personalized tremor treatment plan monitoring system and method to solve the above-mentioned problems:

[0004] This invention proposes a smartwatch-based personalized tremor treatment plan monitoring system and method, the method comprising:

[0005] S1. Collect and analyze user physiological data and motion tremor data through smartwatches. Combine and analyze the obtained time-series user tremor analysis data and time-series user physiological analysis data to obtain time-series user combined analysis feature data.

[0006] S2. Extract features and adjust weights from the time-series user combination analysis feature data for electrical stimulation data to obtain updated analysis feature data. Train and obtain a personal response model based on the updated analysis feature data. Perform feedback comparison analysis on multiple electrical stimulation stages based on the personal response model. Monitor the effect data of the tremor treatment plan based on the feedback comparison analysis results to obtain effect analysis data.

[0007] Further, S1 includes:

[0008] The smartwatch collects user hand movement data and user hand physiological data through its built-in sensor group to obtain user movement data and user physiological data.

[0009] Obtain preset acquisition timing information, and perform timing annotation on the user motion acquisition data and user physiological acquisition data according to the preset acquisition timing information to obtain timing motion acquisition data and timing physiological acquisition data;

[0010] Perform tremor data analysis on time-series motion acquisition data to obtain time-series user tremor analysis data;

[0011] Physiological data analysis is performed on time-series physiological data to obtain time-series user physiological analysis data;

[0012] By combining time-series user tremor analysis data with time-series user physiological analysis data, the combined analysis feature data of time-series users is obtained.

[0013] Furthermore, the step of performing tremor data analysis on the time-series motion acquisition data to obtain time-series user tremor analysis data includes:

[0014] Perform regular clustering analysis on time-series motion data to obtain time-series regular motion data and time-series irregular motion data;

[0015] The similarity coefficient is obtained by comparing temporal pattern motion data with preset movement data.

[0016] The tremor similarity coefficient is compared with the motion similarity coefficient to obtain the motion analysis comparison results;

[0017] Based on the motion analysis comparison results, the temporal regularity motion data is used to determine tremor data, thereby obtaining temporal user tremor analysis data.

[0018] Furthermore, the step of performing physiological data analysis on the time-series physiological data acquisition to obtain time-series user physiological analysis data includes:

[0019] Regular clustering analysis was performed on time-series physiological data to obtain both regular and irregular time-series physiological data.

[0020] Feature extraction is performed on irregular time-series physiological data to obtain irregular time-series physiological feature data;

[0021] The irregular physiological characteristic data in the time series is the time series user physiological analysis data.

[0022] Furthermore, the step of combining and analyzing time-series user tremor analysis data with time-series user physiological analysis data to obtain time-series user combined analysis feature data includes:

[0023] The temporal correlation determination is performed on the temporal user tremor analysis data and the temporal user physiological analysis data to obtain temporal correlation determination information;

[0024] Obtain time-series correlated data and time-series non-correlated data based on time-series correlation determination information;

[0025] Physiological impact tremor annotation was performed on time-series correlated data to obtain time-series physiological impact annotation data;

[0026] Non-physiologically affected tremor annotation was performed on time-series unrelated data to obtain time-series non-physiologically affected labeled data.

[0027] The time-series physiological impact labeled data and non-physiological impact labeled data are the time-series user combination analysis feature data.

[0028] Further, S2 includes:

[0029] Feature extraction of electrical stimulation data is performed on time-series user combination analysis feature data to obtain electrical stimulation feature feedback data;

[0030] Weighted feedback data is obtained by adjusting the weights of the electrical stimulation feature feedback data.

[0031] The time-series user combination analysis feature data is updated by using weighted feedback data to obtain updated analysis feature data.

[0032] A training set is generated based on the updated analytical feature data, and a deep learning model is trained based on the training set to obtain a personal response model;

[0033] Based on the individual response model, the effectiveness data of the tremor treatment plan was analyzed to obtain the effectiveness analysis data.

[0034] Based on the effect analysis data, the optimal combination of electrical stimulation parameters is determined, and the optimal parameters are pushed out.

[0035] Further, the step of adjusting the weights of the electrical stimulation feature feedback data to obtain weighted feedback data includes:

[0036] Calculate the proportion of electrical stimulation feature feedback data relative to time-series combined analysis feature data to obtain the data proportion coefficient;

[0037] The weighting baseline data of the electrical stimulation feature feedback data is increased and adjusted according to the data proportion coefficient to obtain the increased and adjusted data.

[0038] The increased adjustment data is the weighted feedback data.

[0039] Furthermore, the effect data analysis of the tremor treatment plan based on the individual response model, to obtain effect analysis data, includes:

[0040] Updated feedback data is obtained by collecting updated data on the user's electrical stimulation characteristics feedback data;

[0041] The updated feedback data is divided into data before, during and after electrical stimulation to obtain first-stage feedback data, second-stage feedback data and third-stage feedback data.

[0042] Obtain the absolute value of the difference between the third-stage feedback data and the second-stage feedback data to obtain the second feedback coefficient;

[0043] Obtain the absolute value of the difference between the second-stage feedback data and the first-stage feedback data to obtain the first feedback coefficient;

[0044] Obtain the absolute value of the difference between the second feedback coefficient and the first feedback coefficient to obtain the comprehensive feedback coefficient;

[0045] The comprehensive feedback coefficient is compared with the preset comprehensive feedback threshold to obtain the comprehensive feedback comparison result;

[0046] Based on the comprehensive feedback comparison results, the effectiveness of the tremor treatment plan is determined, and effectiveness analysis data is obtained.

[0047] Furthermore, the step of determining the effectiveness of the tremor treatment plan based on the comprehensive feedback comparison results and obtaining effectiveness analysis data includes:

[0048] When the comprehensive feedback comparison result is that the comprehensive feedback coefficient is greater than the preset comprehensive feedback threshold, the tremor treatment plan is judged to be qualified.

[0049] When the comprehensive feedback comparison result is that the comprehensive feedback coefficient is less than or equal to the preset comprehensive feedback threshold, the tremor treatment plan is deemed unqualified.

[0050] The criteria for determining whether an effect is satisfactory or unsatisfactory are the data used for effect analysis.

[0051] Furthermore, the system includes:

[0052] The user data analysis module is used to collect and analyze user physiological data and motion tremor data through smartwatches. It combines and analyzes the obtained time-series user tremor analysis data and time-series user physiological analysis data to obtain time-series user combined analysis feature data.

[0053] The effect analysis module is used to extract features and adjust weights of electrical stimulation data from time-series user combination analysis feature data to obtain updated analysis feature data. Based on the updated analysis feature data, a personal response model is trained and obtained. Based on the personal response model, feedback comparison analysis is performed on multiple electrical stimulation stages. Based on the feedback comparison analysis results, the effect data of the tremor treatment plan is monitored to obtain effect analysis data.

[0054] The beneficial effects of this invention are as follows: Instead of using a fixed parameter model, this invention continuously learns each user's unique physiological and tremor response characteristics to obtain a fully customized treatment plan. It automatically completes the entire process of treatment, monitoring, evaluation, and optimization, and can dynamically adjust strategies according to changes in the user's condition or physical adaptability, achieving long-term optimal management. By intelligently recommending optimal initial parameters, it significantly shortens the trial-and-error cycle in finding effective treatment plans. Attached Figure Description

[0055] Figure 1 A schematic diagram of a smartwatch-based monitoring method for personalized tremor treatment;

[0056] Figure 2 This is a schematic diagram for determining temporal correlation. Detailed Implementation

[0057] The preferred embodiments of the present invention will be described below with reference to the accompanying drawings. It should be understood that the preferred embodiments described herein are for illustration and explanation only and are not intended to limit the present invention.

[0058] In one embodiment of the present invention, a smartwatch-based individualized tremor treatment plan monitoring system and method are proposed, the method comprising:

[0059] S1. Collect and analyze user physiological data and motion tremor data through smartwatches. Combine and analyze the obtained time-series user tremor analysis data and time-series user physiological analysis data to obtain time-series user combined analysis feature data.

[0060] S2. Extract features and adjust weights from the time-series user combination analysis feature data for electrical stimulation data to obtain updated analysis feature data. Train and obtain a personal response model based on the updated analysis feature data. Perform feedback comparison analysis across multiple electrical stimulation stages based on the personal response model. Monitor the effectiveness of the tremor treatment plan based on the feedback comparison analysis results to obtain effectiveness analysis data, such as... Figure 1 As shown.

[0061] The working principle and technical effects of the above solution are as follows: The user's hand movement signals and physiological signals are synchronously collected via a smartwatch. After time-series alignment, the tremor and physiological state components are analyzed separately, and then combined and correlated to generate a comprehensive time-series feature dataset that integrates the user's tremor pattern and physiological background. Features related to electrical stimulation therapy events are extracted from the comprehensive data, and their weights are increased to enhance the model's learning of the treatment response. Using this data to train a personal response model, the model can predict the effects of different electrical stimulation parameters and recommend the optimal combination of treatment parameters to the user.

[0062] This invention continuously learns each user's unique physiological and tremor response characteristics to obtain a fully customized treatment plan. It automatically completes the entire process of treatment, monitoring, evaluation, and optimization, dynamically adjusting strategies according to changes in the user's condition or physical adaptation, achieving long-term optimal management. By intelligently recommending optimal initial parameters, it significantly shortens the trial-and-error cycle in finding effective treatment plans.

[0063] In one embodiment of the present invention, S1 includes:

[0064] The smartwatch collects user hand movement data and user hand physiological data through its built-in sensor array. The user movement data includes data on changes in watch position and the frequency of such changes. The user physiological data includes user electrical stimulation feedback data, user heart rate data, and user ductal skin data.

[0065] Obtain preset acquisition timing information, and perform timing annotation on the user motion acquisition data and user physiological acquisition data according to the preset acquisition timing information to obtain timing-series motion acquisition data and timing-series physiological acquisition data; including user motion acquisition data and user physiological acquisition data at each timing node;

[0066] Perform tremor data analysis on time-series motion acquisition data to obtain time-series user tremor analysis data;

[0067] Physiological data analysis is performed on time-series physiological data to obtain time-series user physiological analysis data;

[0068] By combining time-series user tremor analysis data with time-series user physiological analysis data, the combined analysis feature data of time-series users is obtained.

[0069] The working principle and technical effects of the above solution are as follows: The inertial measurement unit continuously and frequently collects acceleration and angular velocity data of the hand, directly capturing the trajectory, speed, and rhythm of movement. Physiological sensors simultaneously collect signals such as heart rate and skin conductance, reflecting the user's autonomic nervous activity. A unified temporal reference is used to precisely timestamp all data streams, achieving accurate alignment of multi-source data and forming a sequence of motion and physiological data with strict temporal relationships, providing the necessary conditions for correlation analysis.

[0070] Leveraging the convenience of wearable devices, continuous, long-term monitoring of users in their natural state can be achieved, capturing subtle changes and patterns that are difficult to detect during brief clinical consultations. Combining motion dynamics data with intrinsic physiological data provides rich information on the occurrence, development, and regulation of tremor. Precise temporal annotations allow for analysis of potential causal or triggering relationships between tremor events and physiological states.

[0071] In one embodiment of the present invention, the step of performing tremor data analysis on time-series motion acquisition data to obtain time-series user tremor analysis data includes:

[0072] Regular clustering analysis is performed on time-series motion data to obtain time-series regular motion data and time-series irregular motion data; regular clustering analysis includes the analysis of regular and irregular data in time-series motion data.

[0073] The similarity coefficient is obtained by comparing the temporal regularity motion data with the preset tremor regularity motion data; the preset tremor regularity motion data can be historical tremor research data or user's historical tremor data.

[0074] The similarity coefficient is obtained by comparing time-series motion data with preset motion data. The preset motion data can be historical motion research data or user historical motion data, and can include motion speed, motion amplitude, motion rhythm characteristics, etc.

[0075] The tremor similarity coefficient is compared with the motion similarity coefficient to obtain the motion analysis comparison results;

[0076] Based on the motion analysis comparison results, the temporal regularity motion data is classified as tremor data to obtain temporal user tremor analysis data. Data with a tremor similarity coefficient greater than the motion similarity coefficient is classified as tremor data.

[0077] The working principle and technical effects of the above solution are as follows: Pathological tremors typically manifest as rhythmic, oscillatory movement patterns. Regular clustering analysis separates components with significant periodicity from the original motion signal. These periodic components are then compared for similarity with a pre-set library of typical tremor patterns, and also with a pre-set library of normal movement patterns. By comparing tremor similarity and movement similarity, and following the rule that a higher tremor similarity indicates a tremor, tremor signal identification and extraction are achieved. This effectively distinguishes between conscious user movements and involuntary pathological tremors, avoiding misinterpreting normal activity as tremor and inducing unnecessary treatment, thus ensuring the specificity of treatment. It not only detects tremors but also outputs their temporal characteristics.

[0078] In one embodiment of the present invention, the step of performing physiological data analysis on time-series physiological data acquisition to obtain time-series user physiological analysis data includes:

[0079] Regular clustering analysis was performed on time-series physiological data to obtain both regular and irregular time-series physiological data.

[0080] Regular clustering analysis includes the analysis of regular and irregular data in time-series motion data.

[0081] Feature extraction is performed on irregular physiological data over time to obtain irregular physiological feature data over time; feature extraction is used to acquire irregular physiological data and then to analyze abnormal physiological data.

[0082] The irregular physiological characteristic data in the time series is the time series user physiological analysis data.

[0083] The working principle and technical effect of the above technical solution are as follows: Through regular clustering analysis, those periods in physiological signals that significantly deviate from their baseline or average state (such as sudden spikes in heart rate or explosive increases in skin conductance) are identified and separated, and classified as irregular physiological data. Features are extracted from the data of these abnormal event periods, and their intensity, duration, and other attributes are quantified to form temporal feature data characterizing the user's physiological stress events.

[0084] Identifying irregular, sudden physiological changes allows for the precise pinpointing of acute physiological triggers that may directly induce or exacerbate tremors, such as sudden stress, emotional fluctuations, or environmental stimuli. Transforming continuous, analogous physiological signals into discrete, quantifiable sequences of abnormal event characteristics provides clear and computable data for analyzing their association with tremors.

[0085] In one embodiment of the present invention, the step of combining and analyzing time-series user tremor analysis data and time-series user physiological analysis data to obtain time-series user combined analysis feature data includes:

[0086] The time-series user tremor analysis data and time-series user physiological analysis data are subjected to time-series correlation determination to obtain time-series correlation determination information; time-series user tremor analysis data and time-series user physiological analysis data that have both time-series user tremor analysis data and time-series user physiological analysis data in the same time series are correlated; otherwise, they are not correlated. Figure 2 As shown;

[0087] Based on the time-series correlation determination information, obtain time-series correlated data and time-series uncorrelated data; time-series correlated data has a pair of data, while time-series uncorrelated data has only a single data.

[0088] Physiological impact tremor was labeled on the time-series correlated data to obtain time-series physiological impact labeled data; the time-series correlated data contained outlier groups at the same time, so they may be physiological impact tremors;

[0089] Non-physiologically affected tremor annotation was performed on time-series unrelated data to obtain time-series non-physiologically affected labeled data.

[0090] The time-series physiological impact labeled data and non-physiological impact labeled data are the time-series user combination analysis feature data.

[0091] The working principle and technical effect of the above-mentioned technical solution are as follows: Correlation determination is performed on tremor analysis data (when the tremor occurred) and physiological analysis data (when the physiological abnormality occurred). When both exist simultaneously at the same time point or within the same time period, they are marked as time-series correlated data. Correlated data is labeled as physiologically affected tremor, indicating that the tremor attack was likely triggered or modulated by the abnormal physiological state at that time. Tremors not associated with physiological abnormalities are labeled as non-physiologically affected tremors, and are determined to be more likely to originate from primary pathological activity of the nervous system.

[0092] Tremor has evolved from a single motor symptom into a physiological event closely related to internal states. Contextual features of individual response models are obtained through labeled information. The model can learn that when a user is under physiological influence, different treatment strategies may be needed compared to when they are not under physiological influence, thus enabling more refined and intelligent parameter recommendations.

[0093] In one embodiment of the present invention, S2 includes:

[0094] Feature extraction of electrical stimulation data is performed on the time-series user combination analysis feature data to obtain electrical stimulation feature feedback data; the electrical stimulation data includes any data related to electrical stimulation, specifically including time-series user tremor analysis data before, during and after electrical stimulation and time-series user physiological analysis data;

[0095] The weights of the electrical stimulation feature feedback data are increased to obtain weighted feedback data; increasing the weights is used to make the model training more in line with the needs of electrical stimulation effect evaluation.

[0096] The time-series user combination analysis feature data is updated by weighted feedback data to obtain updated analysis feature data; the time-series user combination analysis feature data is then adjusted by weighting the electrical stimulation feature feedback data.

[0097] A training set is generated based on the updated analytical feature data, and a deep learning model is trained based on the training set to obtain a personal response model; the personal response model is the model obtained by training a deep learning model on the updated analytical feature data.

[0098] Based on the individual response model, the effectiveness data of the tremor treatment plan was analyzed to obtain the effectiveness analysis data.

[0099] Based on the effect analysis data, the optimal combination of electrical stimulation parameters is determined, and the optimal parameters are pushed out.

[0100] The working principle and technical effects of the above solution are as follows: Data collected before and after an electrical stimulation treatment event is obtained from comprehensive feature data, and electrical stimulation feature feedback data is extracted. The weights of key data are increased during model training, making the output of the artificial intelligence model more closely resemble the treatment response pattern. The processed data is used to train a deep learning model, enabling it to learn to map the user's current state features to the electrical stimulation treatment effect. The trained personal response model can simulate and compare the expected effects of different combinations of stimulation parameters and proactively push the parameters with the optimal predicted effect.

[0101] Deep learning models can fit the highly complex, nonlinear mapping between tremor physiology, user state, and electrical stimulation parameters, exhibiting stronger expressive power compared to traditional methods. The models can not only evaluate tried treatments but also predict the potential effects of new, unused parameter combinations, thus proactively exploring optimal treatment areas and achieving self-evolution of treatment plans. Weighting treatment event data ensures the model's core capabilities. The rapid formation and continuous optimization of treatment decision-making capabilities improve the accuracy and reliability of the recommendation system.

[0102] In one embodiment of the present invention, the step of adjusting the weights of the electrical stimulation feature feedback data to obtain weighted feedback data includes:

[0103] The proportion of electrical stimulation feature feedback data relative to time-series combined analysis feature data is calculated to obtain the data proportion coefficient; by performing proportion analysis, the influence weight of electrical stimulation on the overall training data can be understood, and the weight can be further adjusted.

[0104] The weighted baseline data of the electrical stimulation feature feedback data is increased and adjusted according to the data proportion coefficient to obtain the increased adjustment data; the weighted baseline data is preset weighted data, which is set by the electrical stimulation influence annotation data.

[0105] The increased adjustment data is the weighted feedback data.

[0106] The working principle and technical effect of the above technical solution are as follows: The proportion coefficient of electrical stimulation feature data in the overall feature data is calculated. Based on the proportion coefficient, the preset weight benchmark of the electrical stimulation feature data is adjusted and increased in a targeted manner. Adjusting the weight of data samples that directly reflect the treatment response in the total data amplifies their contribution to the model training loss function, making the model pay more attention to and learn from these scarce but highly valuable samples, thereby more effectively building the stimulus-response prediction capability.

[0107] The strategy enables the model training process to automatically adapt to the data distribution characteristics of different users, ensuring that treatment response learning remains a high priority even when massive background data is generated by 24 / 7 monitoring. By focusing on learning from high-value samples, the model can more quickly grasp individualized treatment patterns and achieve higher accuracy in predicting treatment effects, thus shortening the time required for model maturation.

[0108] In one embodiment of the present invention, the step of analyzing the effect data of tremor treatment plans based on individual response models to obtain effect analysis data includes:

[0109] Updated data of the user's electrical stimulation characteristic feedback data is collected to obtain updated feedback data; the purpose of obtaining updated feedback data is to obtain the user's latest physiological and motor abnormal data.

[0110] The updated feedback data is divided into three stages: before, during, and after electrical stimulation, to obtain first-stage feedback data, second-stage feedback data, and third-stage feedback data. The three stages are used to differentiate the impact data of electrical stimulation.

[0111] Obtain the absolute value of the difference between the third-stage feedback data and the second-stage feedback data to obtain the second feedback coefficient;

[0112] Obtain the absolute value of the difference between the second-stage feedback data and the first-stage feedback data to obtain the first feedback coefficient;

[0113] Obtain the absolute value of the difference between the second feedback coefficient and the first feedback coefficient to obtain the comprehensive feedback coefficient; obtaining the comprehensive feedback coefficient is to understand the data changes after electrical stimulation compared to before electrical stimulation, in order to understand the effect;

[0114] The comprehensive feedback coefficient is compared with the preset comprehensive feedback threshold to obtain the comprehensive feedback comparison result;

[0115] Based on the comprehensive feedback comparison results, the effectiveness of the tremor treatment plan is determined, and effectiveness analysis data is obtained.

[0116] When the comprehensive feedback comparison result is that the comprehensive feedback coefficient is greater than the preset comprehensive feedback threshold, the tremor treatment plan is judged to be qualified.

[0117] When the comprehensive feedback comparison result is that the comprehensive feedback coefficient is less than or equal to the preset comprehensive feedback threshold, the tremor treatment plan is deemed unqualified.

[0118] The criteria for determining whether an effect is satisfactory or unsatisfactory are the data used for effect analysis.

[0119] The working principle and technical effects of the above solution are as follows: Three consecutive stages of a treatment process are captured: before treatment, during treatment, and after treatment. A comprehensive feedback coefficient is obtained by calculating the change during treatment relative to before treatment (first feedback coefficient) and the change after treatment relative to during treatment (second feedback coefficient). This coefficient is used to comprehensively evaluate the immediate and sustained effects of the treatment. A binary judgment of "qualified" or "unqualified" is made by comparing the result with a preset threshold.

[0120] By transforming vague, subjective perceptions of treatment effects into precise, calculable objective indicators, a clear and unified optimization objective was achieved. Incorporating post-treatment observation data not only assessed the immediate intensity of the treatment but also its persistence, which is crucial for optimizing treatment parameters (such as dosage and duration) to balance immediate effects with long-term tolerability. The resulting effect analysis data directly serves as training labels (supervisory signals) for the individual response model or as rewards for reinforcement learning, driving the model to continuously optimize treatment plans.

[0121] According to one embodiment of the present invention, the system includes:

[0122] The user data analysis module is used to collect and analyze user physiological data and motion tremor data through smartwatches. It combines and analyzes the obtained time-series user tremor analysis data and time-series user physiological analysis data to obtain time-series user combined analysis feature data.

[0123] The effect analysis module is used to extract features and adjust weights of electrical stimulation data from time-series user combination analysis feature data to obtain updated analysis feature data. Based on the updated analysis feature data, a personal response model is trained and obtained. Based on the personal response model, feedback comparison analysis is performed on multiple electrical stimulation stages. Based on the feedback comparison analysis results, the effect data of the tremor treatment plan is monitored to obtain effect analysis data.

[0124] The working principle and technical effects of the above solution are as follows: The system synchronously collects the user's hand movement signals and physiological signals via a smartwatch. After time-series alignment, it analyzes the tremor and physiological state components separately, and then combines and correlates these components to generate a comprehensive time-series feature dataset that integrates the user's tremor pattern and physiological background. From this comprehensive data, the system focuses on extracting features related to electrical stimulation therapy events and strengthens the model's learning of the treatment response by increasing their weights. Using this data, a personal response model is trained. This model can predict the effects of different electrical stimulation parameters, thereby recommending the optimal combination of treatment parameters to the user.

[0125] The system breaks away from fixed-parameter models, providing fully customized treatment plans by continuously learning each user's unique physiological and tremor response characteristics. The system automates the entire process of treatment, monitoring, evaluation, and optimization, dynamically adjusting strategies according to changes in the user's condition or physical adaptation to achieve long-term optimal management. By intelligently recommending optimal initial parameters, it significantly shortens the trial-and-error cycle in finding effective treatment options, enabling users to achieve symptom relief more quickly and improving treatment experience and adherence.

[0126] The system's decision-making is based on the objective analysis of multimodal time-series data, which builds the data foundation for precision medicine and makes the treatment process more scientific, quantifiable, and traceable.

[0127] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A method for monitoring personalized tremor treatment plans using a smartwatch, characterized in that, The method includes: S1. Collect and analyze user physiological data and motion tremor data through smartwatches. Combine and analyze the obtained time-series user tremor analysis data and time-series user physiological analysis data to obtain time-series user combined analysis feature data. S2. Extract features and adjust weights from the time-series user combination analysis feature data for electrical stimulation data to obtain updated analysis feature data. Train and obtain a personal response model based on the updated analysis feature data. Perform feedback comparison analysis on multiple electrical stimulation stages based on the personal response model. Monitor the effect data of the tremor treatment plan based on the feedback comparison analysis results to obtain effect analysis data.

2. The method for monitoring a personalized tremor treatment plan using a smartwatch according to claim 1, characterized in that, S1 includes: The smartwatch collects user hand movement data and user hand physiological data through its built-in sensor group to obtain user movement data and user physiological data. Obtain preset acquisition timing information, and perform timing annotation on the user motion acquisition data and user physiological acquisition data according to the preset acquisition timing information to obtain timing motion acquisition data and timing physiological acquisition data; Perform tremor data analysis on time-series motion acquisition data to obtain time-series user tremor analysis data; Physiological data analysis is performed on time-series physiological data to obtain time-series user physiological analysis data; By combining time-series user tremor analysis data with time-series user physiological analysis data, the combined analysis feature data of time-series users is obtained.

3. The method for monitoring a personalized tremor treatment plan using a smartwatch according to claim 2, characterized in that, The process of performing tremor data analysis on time-series motion acquisition data to obtain time-series user tremor analysis data includes: Perform regular clustering analysis on time-series motion data to obtain time-series regular motion data and time-series irregular motion data; The similarity coefficient of tremor is obtained by comparing the temporal regularity motion data with the preset tremor regularity motion data. The similarity coefficient is obtained by comparing temporal pattern motion data with preset movement data. The tremor similarity coefficient is compared with the motion similarity coefficient to obtain the motion analysis comparison results; Based on the motion analysis comparison results, the temporal regularity motion data is used to determine tremor data, thereby obtaining temporal user tremor analysis data.

4. The method for monitoring a personalized tremor treatment plan using a smartwatch according to claim 2, characterized in that, The step of performing physiological data analysis on time-series physiological data to obtain time-series user physiological analysis data includes: Regular clustering analysis was performed on time-series physiological data to obtain both regular and irregular time-series physiological data. Feature extraction is performed on irregular time-series physiological data to obtain irregular time-series physiological feature data; The irregular physiological characteristic data in the time series is the time series user physiological analysis data.

5. The method for monitoring a personalized tremor treatment plan using a smartwatch according to claim 2, characterized in that, The step of combining and analyzing time-series user tremor analysis data with time-series user physiological analysis data to obtain time-series user combined analysis feature data includes: The temporal correlation determination is performed on the temporal user tremor analysis data and the temporal user physiological analysis data to obtain temporal correlation determination information; Obtain time-series correlated data and time-series non-correlated data based on time-series correlation determination information; Physiological impact tremor annotation was performed on time-series correlated data to obtain time-series physiological impact annotation data; Non-physiologically affected tremor annotation was performed on time-series unrelated data to obtain time-series non-physiologically affected labeled data. The time-series physiological impact labeled data and non-physiological impact labeled data are the time-series user combination analysis feature data.

6. The method for monitoring a personalized tremor treatment plan using a smartwatch according to claim 1, characterized in that, S2 includes: Feature extraction of electrical stimulation data is performed on time-series user combination analysis feature data to obtain electrical stimulation feature feedback data; Weighted feedback data is obtained by adjusting the weights of the electrical stimulation feature feedback data. The time-series user combination analysis feature data is updated by using weighted feedback data to obtain updated analysis feature data. A training set is generated based on the updated analytical feature data, and a deep learning model is trained based on the training set to obtain a personal response model; Based on the individual response model, the effectiveness data of the tremor treatment plan was analyzed to obtain the effectiveness analysis data. Based on the effect analysis data, the optimal combination of electrical stimulation parameters is determined, and the optimal parameters are pushed out.

7. The method for monitoring a personalized tremor treatment plan using a smartwatch according to claim 6, characterized in that, The step of adjusting the weights of the electrical stimulation feature feedback data to obtain weighted feedback data includes: Calculate the proportion of electrical stimulation feature feedback data relative to time-series combined analysis feature data to obtain the data proportion coefficient; The weighting baseline data of the electrical stimulation feature feedback data is increased and adjusted according to the data proportion coefficient to obtain the increased and adjusted data. The increased adjustment data is the weighted feedback data.

8. The method for monitoring a personalized tremor treatment plan using a smartwatch according to claim 6, characterized in that, The effect data analysis of the tremor treatment plan based on the individual response model is used to obtain effect analysis data, including: Updated data of the user's electrical stimulation characteristic feedback data is collected to obtain updated feedback data; the updated feedback data is divided into data before electrical stimulation, during electrical stimulation and after electrical stimulation to obtain first-stage feedback data, second-stage feedback data and third-stage feedback data. Obtain the absolute value of the difference between the third-stage feedback data and the second-stage feedback data to obtain the second feedback coefficient; Obtain the absolute value of the difference between the second-stage feedback data and the first-stage feedback data to obtain the first feedback coefficient; Obtain the absolute value of the difference between the second feedback coefficient and the first feedback coefficient to obtain the comprehensive feedback coefficient; The comprehensive feedback coefficient is compared with the preset comprehensive feedback threshold to obtain the comprehensive feedback comparison result; Based on the comprehensive feedback comparison results, the effectiveness of the tremor treatment plan is determined, and effectiveness analysis data is obtained.

9. The method for monitoring a personalized tremor treatment plan using a smartwatch according to claim 8, characterized in that, The process of determining the effectiveness of the tremor treatment plan based on the comprehensive feedback comparison results and obtaining effectiveness analysis data includes: When the comprehensive feedback comparison result is that the comprehensive feedback coefficient is greater than the preset comprehensive feedback threshold, the tremor treatment plan is judged to be qualified. When the comprehensive feedback comparison result is that the comprehensive feedback coefficient is less than or equal to the preset comprehensive feedback threshold, the tremor treatment plan is deemed unqualified. The criteria for determining whether an effect is satisfactory or unsatisfactory are the data used for effect analysis.

10. A smartwatch-based personalized tremor treatment monitoring system, characterized in that, The system includes: The user data analysis module is used to collect and analyze user physiological data and motion tremor data through smartwatches. It combines and analyzes the obtained time-series user tremor analysis data and time-series user physiological analysis data to obtain time-series user combined analysis feature data. The effect analysis module is used to extract features and adjust weights of electrical stimulation data from time-series user combination analysis feature data to obtain updated analysis feature data. Based on the updated analysis feature data, a personal response model is trained and obtained. Based on the personal response model, feedback comparison analysis is performed on multiple electrical stimulation stages. Based on the feedback comparison analysis results, the effect data of the tremor treatment plan is monitored to obtain effect analysis data.

Images