System for evaluating the effects of rehabilitation of sports injuries based on a multi-modal sensor chip

Abstract

The application discloses a sports injury rehabilitation effect evaluation system based on a multi-modal sensing chip and belongs to the field of sports injury rehabilitation. The sports injury rehabilitation effect evaluation system based on the multi-modal sensing chip comprises a data acquisition unit, an edge computing unit, an analysis and evaluation unit and a health management unit. The application solves the problem that the prior art mainly relies on subjective judgment and patient self-reports, lacks objective quantitative evaluation means, and leads to low rehabilitation evaluation efficiency, insufficient precision and untimely scheme adjustment. The data acquisition unit of the application synchronously acquires objective data through the multi-modal sensing chip integrated in a limb, realizes real-time preprocessing through the edge computing unit, dynamically analyzes and individually evaluates by using the intelligent model of deep learning and reinforcement learning through the analysis and evaluation unit, and finally, the health management unit adaptively adjusts and visually displays the rehabilitation scheme according to the evaluation result, so as to form a complete sports injury rehabilitation effect evaluation closed loop.

Description

Technical Field

[0001] This invention relates to the field of sports injury rehabilitation technology, specifically a sports injury rehabilitation effect evaluation system based on a multimodal sensor chip. Background Technology

[0002] Sports injuries refer to damage to body tissues or organs caused by various reasons during sports activities. These injuries may involve muscles, tendons, ligaments, joints, bones, nerves, etc., and are usually related to factors such as the intensity, frequency, technique, and individual physical condition of the exercise. The rehabilitation process for sports injuries requires scientific and systematic assessment and guidance. However, traditional rehabilitation assessments mainly rely on the subjective judgment of doctors and the self-feedback of patients, lacking objective and accurate quantitative assessment methods. This assessment method is not only inefficient but also fails to accurately reflect the patient's rehabilitation progress and potential problems, leading to untimely and unpersonalized adjustments to the rehabilitation plan, thus affecting the rehabilitation outcome. Therefore, it does not meet the current needs. To address this, we propose a sports injury rehabilitation effect assessment system based on a multimodal sensor chip. Summary of the Invention

[0003] The purpose of this invention is to provide a sports injury rehabilitation effect evaluation system based on a multimodal sensor chip. By integrating a multimodal sensor chip to synchronously collect objective and quantitative sports physiological data, and using edge computing and intelligent algorithms that integrate deep learning and reinforcement learning for real-time analysis and evaluation, the system achieves accurate, dynamic and personalized evaluation of rehabilitation effects. At the same time, it adaptively adjusts the training program according to the evaluation results, thus solving the problems mentioned in the background art.

[0004] To achieve the above objectives, the present invention provides the following technical solution: a sports injury rehabilitation effect evaluation system based on a multimodal sensor chip, comprising: The data acquisition unit, including a smart chip, is configured to integrate biosensors and physical sensors and is positioned on the limb part of the user to be evaluated, for the purpose of simultaneously collecting multi-dimensional motor physiological data of the user during the rehabilitation training process; The edge computing unit is configured to perform localized real-time preprocessing and initial feature extraction of the collected motion physiological data; The analysis and evaluation unit is configured to receive preprocessed motor physiological data, integrate deep learning algorithms and reinforcement learning algorithms, construct a rehabilitation effect evaluation model, identify abnormal movement patterns and physiological patterns of patients during the rehabilitation process, and use it to evaluate the rehabilitation effect. The health management unit is configured to adjust the user's training plan based on the assessment results of rehabilitation effects, and to display the adjusted training plan and assessment results to rehabilitation therapists and patients in a visual manner.

[0005] Furthermore, the analysis and evaluation unit includes: The multimodal data fusion module, in conjunction with the principal component analysis method, fuses the preprocessed exercise physiological data to generate a comprehensive exercise physiological feature vector. The pattern recognition module is configured to identify abnormal movement patterns in patients during rehabilitation by using a hybrid architecture of recurrent neural networks and convolutional neural networks in deep learning, based on the comprehensive motor physiological feature vector. Also based on the comprehensive motor physiological feature vector, a generative adversarial network architecture is adopted to generate data samples simulating normal physiological states through a generator, and a discriminator distinguishes between the actual collected physiological data and the generated normal samples to identify abnormal physiological patterns in patients during rehabilitation. The model building module is configured to comprehensively analyze the recognition results of the pattern recognition module, use reinforcement learning algorithms to build a rehabilitation effect evaluation model, and dynamically adjust the parameters of the evaluation model according to the preset rehabilitation goals and reward mechanism to achieve the evaluation of rehabilitation effect.

[0006] Furthermore, the analysis and evaluation unit also includes: The adaptive learning module is configured to generate a personalized initial movement baseline model for the user in the early stages of rehabilitation, based on the user's first achievement of standard movement data and incorporating the user's historical injury information. In each subsequent rehabilitation training session, the real-time collected exercise physiological data will be compared with the personalized initial exercise baseline model to calculate the dynamic similarity score. Based on the continuous changing trend of dynamic similarity scores, the core parameters of the rehabilitation effect evaluation model are dynamically fine-tuned through online learning algorithms, so that the evaluation criteria can adapt to the user's slowly recovering physiological functions.

[0007] Furthermore, the adaptive learning module is also configured as follows: Based on the user's personalized initial exercise baseline model and historical rehabilitation training data, a time series prediction algorithm is used to predict the trajectory of changes in the user's exercise physiological data during future rehabilitation training, including the expected change curves of joint range of motion, exercise speed and heart rate. During each rehabilitation training session, the deviation between the actual collected exercise physiological data and the predicted trajectory of exercise physiological data changes is calculated in real time. The deviation includes, but is not limited to, absolute deviation, relative deviation, and root mean square deviation. Based on the calculated deviation value and the preset deviation threshold, it is determined whether the rehabilitation effect evaluation model needs to be adjusted. When the deviation value exceeds the preset threshold, the online learning algorithm is triggered to automatically adjust the core parameters of the rehabilitation effect evaluation model.

[0008] Furthermore, the data acquisition unit is also configured to perform real-time online calibration of the biosensors and physical sensors, specifically: The calibration is triggered by external commands, generating an electrical reference signal with known amplitude and frequency and injecting it into the front-end circuits of the biosensor and physical sensor; The response data of biosensors and physical sensors to the reference signal are collected. By comparing the response data with the expected standard value of the reference signal, the zero-point drift, sensitivity drift and nonlinear error of each sensor are calculated in real time. The system receives the calculated error data and uses a preset error compensation model to generate calibration parameters in real time. It then performs online correction and compensation on the raw motion physiological data collected by the sensor and outputs the calibrated data to the edge computing unit.

[0009] Furthermore, during the real-time online calibration process, the clock unit and environmental sensors are used to accumulate the power-on working time of each sensor, and the changes in ambient temperature and humidity are monitored in real time. Based on the working time and environmental change data, the calibration trigger frequency is dynamically adjusted through built-in fuzzy control rules.

[0010] Furthermore, the edge computing unit includes: The data processing module is configured to synchronously acquire multi-dimensional motion physiological data from biosensors and physical sensors and perform filtering, noise reduction and signal normalization processing. It is also configured to attach a timestamp and signal-to-noise ratio label to each data frame based on the real-time received multi-dimensional motion physiological data stream. The validity of the data frame is determined based on a preset quality threshold matrix related to the key aspects of the current rehabilitation action. The effective data frames are divided into high-priority physiological data streams and standard-priority motion data streams, and different buffers and computing resources are allocated accordingly. Among them, high-priority physiological data streams are given priority to enter the feature extraction module for feature extraction; The feature extraction module is configured to extract time-domain, frequency-domain, and time-frequency-domain features from preprocessed multi-dimensional motion physiological data. The compression caching module is configured to compress the preprocessed and feature-extracted exercise physiological data, and temporarily store the preprocessed and feature-extracted exercise physiological data locally.

[0011] Furthermore, based on a preset quality threshold matrix related to the key aspects of the current rehabilitation action, the validity of the data frame is determined, including: Obtain the quality threshold matrix that is critical to the current rehabilitation action corresponding to the data frame, and extract the indicator dimensions from the quality threshold matrix; Retrieve historical user data related to the data frame, learn the fluctuation characteristics of the historical user data in the indicator dimension based on time series analysis, compare the threshold fluctuation characteristics with the standard fluctuation characteristics for establishing a quality threshold matrix reference to obtain the fluctuation difference, and determine the individual calibration factor for the indicator dimension based on the fluctuation difference. The system acquires environmental sensor data and real-time physiological state data of the user from the collected data frames. Based on the environmental difference between the environmental sensor data and the standard environmental data, and combined with the influence of the environment on data acquisition, it determines the environmental calibration factor for the indicator dimensions. Based on the physiological difference between the user's real-time physiological state data and healthy physiological data, and combined with the influence of physiological state data on data acquisition, it determines the health calibration factor for the indicator dimensions. The target calibration factor that maximizes the calibration range of the indicator dimension is determined by the individual calibration factor, environmental calibration factor and health calibration factor. The quality threshold matrix is ​​then individually optimized based on the target calibration factor to obtain the optimized quality threshold matrix. Based on the rehabilitation stage in which the data frame is located, a threshold setting leniency coefficient is determined, and the optimized quality threshold matrix is ​​processed in stages based on the leniency coefficient to obtain the target quality threshold matrix. The data frame is compared with the target quality threshold matrix. If the dimensional index value of the data frame is consistent with the target quality threshold matrix, the data frame is determined to be a valid data frame; otherwise, the data frame is determined to be invalid.

[0012] Furthermore, the health management unit includes: The program adjustment module is configured to analyze the patient's current rehabilitation status and needs based on the assessment results, and to make personalized adjustments to the user's training program in combination with the patient's physical condition, rehabilitation goals and sports injury type, including adjusting training intensity, training movements and training frequency. The results display module is configured to provide a human-computer interaction interface to visually present the adjusted training plan and assessment results to rehabilitation therapists and patients. The progress tracking module is configured to track the patient's rehabilitation training progress in real time, record the completion status, training time and training intensity of each training session, compare and analyze the training progress with the expected rehabilitation plan, promptly identify possible deviations or problems in the training process, and provide timely feedback to the rehabilitation therapist for timely adjustment of the training plan. It is also configured to establish a personalized rehabilitation progress prediction model based on the patient's historical rehabilitation training data and current rehabilitation status, predict the trend of key indicators and the expected time to achieve the goals in future rehabilitation stages. After each rehabilitation training session, the actual training data is compared and analyzed with the personalized rehabilitation progress prediction model to calculate the rehabilitation progress deviation index. The deviation index comprehensively considers multiple dimensions such as training intensity completion rate, movement standard execution rate, and physiological parameter improvement. When the rehabilitation progress deviation index exceeds the preset warning threshold, an abnormal progress reminder is sent to the rehabilitation therapist.

[0013] Furthermore, for invalid data frames, the following operations are performed: Perform critical action determination on invalid data frames; When an invalid data frame is determined to be a non-critical action, the most recent valid data frame is automatically retrieved from the local cache, and a data stream is generated to fill the invalid data frame based on the valid data frame using the polynomial interpolation method. When an invalid data frame is determined to be a critical action, a priority score is assigned to the invalid data frame, and a resource allocation coefficient is determined based on the priority score. The invalid data frame is then processed based on the resource allocation coefficient.

[0014] Compared with the prior art, the beneficial effects of the present invention are: This invention achieves efficient, low-latency acquisition and preliminary analysis of multi-dimensional motor physiological data by configuring an intelligent chip integrating biological and physical sensors on the user's limbs and introducing an edge computing unit for localized real-time preprocessing. This overcomes the shortcomings of traditional assessment methods, such as strong subjectivity, single data, and lag, providing an unprecedented objective and continuous data foundation for rehabilitation assessment. Furthermore, by integrating deep learning and reinforcement learning algorithms to construct a rehabilitation effect assessment model based on multimodal data fusion and introducing an adaptive learning module, the system can continuously fine-tune model parameters based on the user's personalized initial baseline and real-time training data to adapt to the user's slowly recovering physiological functions, significantly improving the scientific nature and rehabilitation effect of rehabilitation training. Attached Figure Description

[0015] Figure 1 This is a structural diagram of the sports injury rehabilitation effect evaluation system based on a multimodal sensor chip according to the present invention. Detailed Implementation

[0016] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0017] To address the technical problems of existing technologies that rely primarily on subjective judgment and patient self-reporting, lacking objective quantitative assessment methods, resulting in low efficiency, insufficient accuracy, and untimely adjustments to rehabilitation plans, please refer to [link / reference needed]. Figure 1 This embodiment provides the following technical solution: A sports injury rehabilitation effect assessment system based on multimodal sensor chips includes: The data acquisition unit, including a smart chip, is configured to integrate biosensors and physical sensors and is positioned on the limb part of the user to be evaluated, for the purpose of simultaneously collecting multi-dimensional motor physiological data of the user during the rehabilitation training process; In this embodiment, the smart chip adopts a heterogeneous integrated architecture. Its integrated biosensors specifically include a miniaturized surface electromyography (EMG) sensor array and a bioimpedance analysis electrode pair. The physical sensors include a nine-axis inertial measurement unit (IMU) and a flexible micro-strain sensor array. The flexible micro-strain sensor array is distributed in a mesh structure on the skin surface of key muscle groups in the limb area to simultaneously measure the multidimensional strain distribution on the skin surface generated during muscle contraction. The system is configured to synchronously trigger the EMG sensor array, bioimpedance analysis electrode pair, nine-axis IMU, and flexible micro-strain sensor array to acquire data synchronously via the on-chip system of the smart chip, generating a multimodal sensing data stream with strictly aligned timestamps. The edge computing unit is configured to perform localized real-time preprocessing and initial feature extraction of the collected motion physiological data; The analysis and evaluation unit is configured to receive preprocessed motor physiological data, integrate deep learning algorithms and reinforcement learning algorithms, construct a rehabilitation effect evaluation model, identify abnormal movement patterns and physiological patterns of patients during the rehabilitation process, and use it to evaluate the rehabilitation effect. The health management unit is configured to adjust the user's training plan based on the assessment results of rehabilitation effects, and to display the adjusted training plan and assessment results to rehabilitation therapists and patients in a visual manner.

[0018] The technical effects of the above solution are as follows: The data acquisition unit achieves high-precision synchronous acquisition of multimodal physiological and motor data through an integrated architecture smart chip, thereby ensuring the temporal alignment and spatial consistency of the data, providing a high-quality underlying data foundation for subsequent analysis. The edge computing unit performs localized real-time preprocessing and feature extraction, which reduces data transmission latency and cloud load, thus ensuring the real-time performance of the system. The analysis and evaluation unit deeply integrates deep learning and reinforcement learning algorithms, which can accurately identify subtle abnormalities in motor and physiological patterns, thereby achieving quantitative evaluation of rehabilitation effects and effectively overcoming the limitations of traditional subjective evaluation. The health management unit adjusts personalized plans and provides visual feedback based on the evaluation results, which can improve the accuracy, adaptability, and user experience of rehabilitation training, ultimately effectively promoting the overall improvement of rehabilitation efficiency and effectiveness.

[0019] Analysis and evaluation unit, including: The multimodal data fusion module, in conjunction with the principal component analysis method, fuses the preprocessed exercise physiological data to generate a comprehensive exercise physiological feature vector. The pattern recognition module is configured to identify abnormal movement patterns in patients during rehabilitation by using a hybrid architecture of recurrent neural networks and convolutional neural networks in deep learning, based on the comprehensive motor physiological feature vector. Also based on the comprehensive motor physiological feature vector, a generative adversarial network architecture is adopted to generate data samples simulating normal physiological states through a generator, and a discriminator distinguishes between the actual collected physiological data and the generated normal samples to identify abnormal physiological patterns in patients during rehabilitation. Among them, abnormal movement patterns include, but are not limited to, abnormal pattern recognition of features such as joint range of motion, movement speed, and movement trajectory; abnormal physiological patterns include, but are not limited to, abnormal change patterns of physiological parameters such as heart rate, blood pressure, and electromyography. In the hybrid architecture, the recurrent neural network part is used to capture the temporal dependencies of time series data, and the convolutional neural network part is used to extract spatial features and can adaptively learn the differences in movement patterns of different patients; the generative adversarial network architecture can dynamically adjust the training strategies of the generator and discriminator according to the individual physiological characteristics of the patients. The model building module is configured to comprehensively analyze the recognition results of the pattern recognition module, construct a rehabilitation effect evaluation model using reinforcement learning algorithm, and dynamically adjust the parameters of the evaluation model according to the preset rehabilitation goals and reward mechanism to achieve the evaluation of rehabilitation effect. The rehabilitation effect evaluation model outputs multiple dimensions of evaluation results, including rehabilitation progress percentage, rehabilitation quality score and potential risk warning. The rehabilitation effect evaluation model is trained using historical rehabilitation data. The reward mechanism of the reinforcement learning algorithm is dynamically formulated based on the patient's rehabilitation goals and individual physiological characteristics, including but not limited to multiple dimensions such as rehabilitation progress rewards, physiological parameter normalization rewards, and movement pattern improvement rewards, in order to incentivize the model to output evaluation results that better meet the patient's actual rehabilitation needs.

[0020] The technical effects of the above solution are as follows: By using a multimodal data fusion module to reduce the dimensionality of heterogeneous sensor data and fuse it into a comprehensive feature vector, the problem of multi-source data redundancy and information complementarity is effectively solved, thus laying a data foundation for accurate pattern recognition. The pattern recognition module combines a hybrid architecture of recurrent neural networks and convolutional neural networks with a generative adversarial network architecture, which can deeply mine subtle abnormalities in movement and physiological patterns from the spatiotemporal dimension and data distribution level. It not only achieves accurate assessment of kinematic indicators such as joint activity and movement trajectory, but also keenly captures pathological changes in physiological parameters such as heart rate and electromyography, thereby improving the sensitivity and comprehensiveness of abnormality detection. The model building module uses reinforcement learning algorithms to combine multidimensional recognition results with a dynamic and personalized reward mechanism, so that the final rehabilitation effect evaluation model can output multidimensional quantitative results covering progress, quality and risk, thus providing individualized decision support that meets the actual clinical needs for precise adjustment of rehabilitation programs.

[0021] The analysis and evaluation unit also includes: The adaptive learning module is configured to generate a personalized initial exercise baseline model for users in the early stages of rehabilitation, based on the standard movement data of the user's first achievement of the target, and by integrating the user's historical injury information (such as the degree of injury and body mass index). In each subsequent rehabilitation training session, the real-time collected exercise physiological data will be compared with the personalized initial exercise baseline model to calculate the dynamic similarity score. Based on the continuous changing trend of dynamic similarity scores, the core parameters of the rehabilitation effect evaluation model are dynamically fine-tuned through online learning algorithms, so that the evaluation criteria can adapt to the user's slowly recovering physiological functions. The adaptive learning module is also configured as follows: Based on the user's personalized initial exercise baseline model and historical rehabilitation training data, a time series prediction algorithm is used to predict the trajectory of changes in the user's exercise physiological data during future rehabilitation training, including the expected change curves of joint range of motion, exercise speed and heart rate. During each rehabilitation training session, the deviation between the actual collected exercise physiological data and the predicted trajectory of exercise physiological data changes is calculated in real time. The deviation includes, but is not limited to, absolute deviation, relative deviation, and root mean square deviation. Based on the calculated deviation value and the preset deviation threshold, it is determined whether the rehabilitation effect evaluation model needs to be adjusted. When the deviation value exceeds the preset threshold, the online learning algorithm is triggered to automatically adjust the core parameters of the rehabilitation effect evaluation model.

[0022] The technical effects of the above solution are as follows: By establishing a personalized initial motion benchmark model and introducing a dynamic similarity assessment mechanism, the system can accurately capture the unique rehabilitation starting point and physiological characteristics of each patient, thereby achieving individualized calibration of the assessment criteria. By continuously tracking the trend of dynamic similarity scores and using online learning algorithms to adaptively fine-tune model parameters, the system ensures that the assessment criteria can evolve in sync with the user's slowly recovering physiological functions, thus maintaining the accuracy and clinical relevance of the assessment. Furthermore, by introducing time-series prediction-based deviation analysis between the expected trajectory and real-time data, the system can not only reflect the current state but also predictively identify deviations between the rehabilitation process and the expected path, and autonomously trigger parameter adjustments when the deviation exceeds the limit. This significantly improves the dynamic adaptability, individualized accuracy, and timely response capability to abnormal fluctuations in the rehabilitation process of the rehabilitation effect assessment model.

[0023] The data acquisition unit is also configured to perform real-time online calibration of biosensors and physical sensors, specifically: The calibration is triggered by external commands, generating an electrical reference signal with known amplitude and frequency and injecting it into the front-end circuits of the biosensor and physical sensor; The response data of biosensors and physical sensors to the reference signal are collected. By comparing the response data with the expected standard value of the reference signal, the zero-point drift, sensitivity drift and nonlinear error of each sensor are calculated in real time. The system receives the calculated error data and uses a preset error compensation model to generate calibration parameters in real time. It then performs online correction and compensation on the raw motion physiological data collected by the sensor and outputs the calibrated data to the edge computing unit. During real-time online calibration, the clock unit and environmental sensors are used to accumulate the power-on working time of each sensor and monitor changes in ambient temperature and humidity in real time. Based on the working time and environmental change data, the calibration trigger frequency is dynamically adjusted through built-in fuzzy control rules.

[0024] The technical effects of the above solution are as follows: By injecting a known electrical reference signal into the sensor front end and performing real-time response comparison, the zero-point drift, sensitivity drift, and nonlinear error of the sensor can be accurately quantified and compensated, thereby improving the accuracy and reliability of multimodal motion physiological data at the source. Combined with a preset error compensation model for online data correction, the data distortion problem caused by sensor characteristic drift or environmental interference is overcome, thus providing a high-quality data foundation for subsequent analysis and evaluation. Furthermore, by utilizing sensor operating time and environmental temperature and humidity data, the calibration frequency can be dynamically adjusted through fuzzy control rules, realizing the intelligent and adaptive calibration process. This ensures data acquisition accuracy while optimizing system energy efficiency and the rational allocation of computing resources.

[0025] Edge computing units include: The data processing module is configured to simultaneously acquire multi-dimensional motion physiological data from biosensors and physical sensors and perform filtering, denoising, and signal normalization processing. The methods for filtering, denoising, and signal normalization processing are existing technologies in this field and are not the inventive solutions of this application, and will not be described in detail here. The feature extraction module is configured to extract time-domain, frequency-domain, and time-frequency-domain features from preprocessed multi-dimensional motion physiological data, specifically: Temporal features: Features such as mean, variance, peak value, and zero-crossing rate are extracted from the temporal signals of exercise physiological data to reflect the statistical characteristics of the data in the time dimension. Frequency domain features: Perform operations such as Fourier transform on exercise physiological data to extract frequency domain features, such as spectral energy and dominant frequency, to analyze the frequency distribution of the data; Time-frequency domain features: By using wavelet transform and other methods, and considering both time and frequency dimensions, the time-frequency domain features of exercise physiological data are extracted, which can better capture the non-stationary characteristics in the data. The compression caching module is configured to compress the preprocessed and feature-extracted exercise physiological data to reduce the data volume and improve data transmission efficiency, while preserving as much key information of the original data as possible so that the subsequent analysis and evaluation unit can process it quickly and accurately. At the same time, it temporarily stores the preprocessed and feature-extracted exercise physiological data locally to provide fast data access support for the analysis and evaluation unit, ensuring the real-time and continuity of the data, and also facilitating data backtracking and further analysis when needed.

[0026] The technical effects of the above solution are as follows: The data processing module improves the reliability of multi-dimensional exercise physiological data by filtering, denoising, and normalizing the signal, thereby ensuring the accuracy of subsequent analysis results. The feature extraction module integrates time-domain, frequency-domain, and time-frequency-domain features to fully capture the static statistical characteristics and dynamic non-stationary characteristics of exercise physiological signals in different dimensions, providing a rich and complementary set of high-value features for subsequent intelligent evaluation. The compression and caching module reduces the amount of data while retaining key information and uses local storage to ensure the real-time performance, continuity, and traceability of historical data in data processing, thus providing a solid foundation for the accurate and real-time decision-making of the analysis and evaluation unit.

[0027] The data processing module is also configured as follows: Based on the real-time received multi-dimensional motion physiological data stream, a timestamp and signal-to-noise ratio label are added to each data frame; The validity of the data frame is determined based on a preset quality threshold matrix related to the key aspects of the current rehabilitation action. The effective data frames are divided into high-priority physiological data streams and standard-priority motion data streams, and different buffers and computing resources are allocated accordingly. Among them, high-priority physiological data streams are given priority to enter the feature extraction module for feature extraction.

[0028] The technical effects of the above solution are as follows: By attaching a timestamp and signal-to-noise ratio label to each data frame and combining it with a quality threshold matrix related to the key aspects of rehabilitation movements for validity judgment, intelligent quality filtering and screening of massive multimodal data streams can be achieved, thereby ensuring the validity and reliability of input data. By distinguishing valid data into high-priority physiological data streams and standard-priority exercise data streams and implementing differentiated resource allocation strategies, the system can prioritize the real-time processing of key data such as heart rate and electromyography that reflect the user's immediate physiological state and safety, thereby optimizing the overall efficiency of the system with limited computing and caching resources.

[0029] In one embodiment, the validity of a data frame is determined based on a preset quality threshold matrix related to the criticality of the current rehabilitation action, including: Obtain the quality threshold matrix that is critical to the current rehabilitation action corresponding to the data frame, and extract the indicator dimensions from the quality threshold matrix; Retrieve historical user data related to the data frame, learn the fluctuation characteristics of the historical user data in the indicator dimension based on time series analysis, compare the threshold fluctuation characteristics with the standard fluctuation characteristics for establishing a quality threshold matrix reference to obtain the fluctuation difference, and determine the individual calibration factor for the indicator dimension based on the fluctuation difference. The system acquires environmental sensor data and real-time physiological state data of the user from the collected data frames. Based on the environmental difference between the environmental sensor data and the standard environmental data, and combined with the influence of the environment on data acquisition, it determines the environmental calibration factor for the indicator dimensions. Based on the physiological difference between the user's real-time physiological state data and healthy physiological data, and combined with the influence of physiological state data on data acquisition, it determines the health calibration factor for the indicator dimensions. The target calibration factor that maximizes the calibration range of the indicator dimension is determined by the individual calibration factor, environmental calibration factor and health calibration factor. The quality threshold matrix is ​​then individually optimized based on the target calibration factor to obtain the optimized quality threshold matrix. Based on the rehabilitation stage in which the data frame is located, a threshold setting leniency coefficient is determined, and the optimized quality threshold matrix is ​​processed in stages based on the leniency coefficient to obtain the target quality threshold matrix. The data frame is compared with the target quality threshold matrix. If the dimensional index value of the data frame is consistent with the target quality threshold matrix, the data frame is determined to be a valid data frame; otherwise, the data frame is determined to be invalid.

[0030] In this embodiment, the metrics dimensions include, for example, timestamps, signal-to-noise ratio labels, and signal amplitude.

[0031] In this embodiment, the user's historical data consists of valid data streams.

[0032] In this embodiment, the influence of the environment on data acquisition is, for example, that sensor signals are easily interfered with in high-temperature environments, and the signal-to-noise ratio threshold is temporarily lowered by 5-10%.

[0033] In this embodiment, the target calibration factor that has the largest calibration range for the indicator dimension is determined by the combination of individual calibration factor, environmental calibration factor and health calibration factor. The target calibration factor is selected as the one with the largest calibration range among the three.

[0034] In this embodiment, the influence of physiological state data on data acquisition is, for example, when the user's heart rate exceeds the safe range, it affects the stability of the movement, and the corresponding movement stability threshold is appropriately relaxed.

[0035] In this embodiment, the leniency coefficient is set based on the following criteria: in the initial rehabilitation stage, the threshold is set relatively loosely to tolerate the instability of the patient's movements; in the intermediate rehabilitation stage, the threshold is gradually tightened to encourage standardized movements; and in the later rehabilitation stage, a strict threshold is adopted to simulate normal movement patterns.

[0036] The beneficial effects of the above design scheme are as follows: By obtaining the quality threshold matrix and extracting the indicator dimensions, the clear indicator dimensions avoid ambiguity in the judgment process, ensure consistency in the judgment logic of data validity for different scenarios and different operators, and improve the reproducibility of the judgment. By generating individual calibration factors based on user historical data, it avoids the misjudgment of effective individual fluctuation data as invalid due to overly strict general thresholds, or the retention of truly invalid data due to overly lenient thresholds, significantly improving the individual accuracy of the judgment. By generating environmental and health calibration factors, it ultimately ensures that the judgment results focus on the effectiveness of rehabilitation itself, rather than external interference or physiological fluctuations, improving the anti-interference ability of the judgment. By selecting the target calibration factor optimization matrix with the maximum calibration amplitude, it can focus on the most critical interference sources affecting data quality in the current scenario, minimize the impact of key interferences, and further improve the threshold adaptation accuracy. By determining the leniency coefficient processing matrix based on the rehabilitation stage, it ensures that the judgment results serve the current rehabilitation goals without restricting the subsequent process, ultimately providing high-quality and highly reliable input data for subsequent sports injury rehabilitation effect evaluation.

[0037] In one embodiment, for invalid data frames, the following operation is performed: Perform critical action determination on invalid data frames; When an invalid data frame is determined to be a non-critical action, the most recent valid data frame is automatically retrieved from the local cache, and a data stream is generated to fill the invalid data frame based on the valid data frame using the polynomial interpolation method. The method for determining the most recent time period is as follows: in, Indicates a recent time period. This indicates the minimum time window set by the system. This indicates the maximum time window set by the system. Indicates the base time window size. The standard deviation of the data quality metric representing the valid data frames cached locally. This represents the average data quality metric of valid data frames cached locally. When an invalid data frame is determined to be a critical action, a priority score is assigned to the invalid data frame, and a resource allocation coefficient is determined based on the priority score. The invalid data frame is then processed based on the resource allocation coefficient. The formula for calculating the priority score of invalid data frames is as follows: in, The priority score indicating an invalid data frame. Indicates key weight, Key scores indicating invalid data frames Indicates time weighting, The time score indicating an invalid data frame. Indicates spatial weights, Spatial score indicating an invalid data frame. Indicates quality weight, The quality score indicates an invalid data frame; The formula for calculating the resource allocation coefficient of an invalid data frame is: in, The resource allocation coefficient representing an invalid data frame. Represents the basic resource allocation coefficient. This represents the first reference parameter, with a value of 0.2. This represents the second reference parameter, with a value of 0.3. Indicates the time when the invalid data frame was last accessed. Indicates the current time. Represents the time decay constant. This represents the natural exponential function.

[0038] In this embodiment, the data quality metric for locally cached valid data frames is specifically the signal-to-noise ratio, the standard deviation of the data quality metric is used to reflect the volatility of data quality, and the mean of the data quality metric is used to reflect the average level of data quality.

[0039] In this embodiment, the criticality score of an invalid data frame is determined based on the difference between the data frame and the standard data frame of the critical action; the smaller the difference, the higher the score.

[0040] In this embodiment, the time score of an invalid data frame is related to the number of valid adjacent time frames; the larger the number, the higher the score.

[0041] In this embodiment, the spatial score of an invalid data frame is related to the difference between the sensor spatial position vector corresponding to the invalid data frame and the spatial position vector of a reference point (such as the body's center of gravity or the location of the injured joint). The smaller the difference, the higher the score.

[0042] In this embodiment, the quality score of an invalid data frame is related to the comparison result between the invalid data frame and the quality threshold matrix; the closer the frame is to the quality threshold matrix, the higher the score.

[0043] In this embodiment, the first reference parameter and the second reference parameter are used to map the priority scores of invalid data frames.

[0044] In this embodiment, the last access time, current time, and time decay constant of invalid data frames are introduced to ensure that resources are allocated to invalid frames that have not been accessed for a long time, and the latest data is processed first, preventing resources from being wasted on old frames, thereby improving resource utilization efficiency and system real-time performance.

[0045] In this embodiment, the formula for calculating the most recent time period is, for example... =2s, =10s, =5s; Assuming stable data quality =20dB, =5dB, calculate =2s, the system uses the data from the most recent 2 seconds for interpolation to ensure timeliness; For cases where numerical quality fluctuates greatly. =20dB, =15dB, calculate =3.75s, the system expands the time window and uses more data to improve interpolation reliability.

[0046] In this embodiment, priority scoring is performed on invalid data frames, for example... , , , The values ​​are 0.4, 0.3, 0.2, and 0.1 respectively. , , and The values ​​are 0.8, 0.6, 0.9, and 0.5 respectively. =0.73.

[0047] In this embodiment, the resource allocation coefficient for invalid data frames, for example... =110-100=10s, =0.73, =10s, =1, then =1-0.5215=0.4785; =0.3679, calculated as follows =0.176.

[0048] The beneficial effects of the above design scheme are as follows: When an invalid data frame is determined to be a non-critical action, the system automatically retrieves the most recent valid data frame from the local cache and generates a data stream to fill the invalid data frame based on the valid data frame using a polynomial interpolation method. The time window is dynamically adjusted according to data quality fluctuations; when data quality is stable, more recent data is used to ensure interpolation accuracy; when data quality fluctuates greatly, the time window is expanded to improve interpolation reliability. This ensures that the interpolated data reflects recent trends while adapting to changes in data quality, improving the accuracy and robustness of the filled data. When an invalid data frame is determined to be a critical action, a priority score is assigned to the invalid data frame, and a resource allocation coefficient is determined based on the priority score. The invalid data frame is then processed based on the resource allocation coefficient. This comprehensive approach, considering four dimensions—criticality, time, space, and quality—ensures that the score fully reflects the importance of the invalid frame, avoiding bias from a single-dimensional judgment. Through the resource allocation coefficient, the system can dynamically allocate computing resources to high-priority invalid frames, ensuring they receive in-depth processing or repair; while low-priority frames may be downgraded or ignored. This optimizes the overall performance of the system and ensures the core accuracy and safety of rehabilitation assessment.

[0049] The health management unit includes: The program adjustment module is configured to analyze the patient's current rehabilitation status and needs based on the assessment results, and to make personalized adjustments to the user's training program in combination with the patient's physical condition, rehabilitation goals and sports injury type. This includes adjusting the training intensity, training movements and training frequency to ensure the scientific nature and effectiveness of the training program and better promote the patient's rehabilitation process. The results display module is configured to provide a human-computer interaction interface to visually present the adjusted training plan and assessment results to rehabilitation therapists and patients. The training program is presented in various visual ways, including text descriptions, graphic illustrations, and video demonstrations. This allows rehabilitation therapists to accurately understand the adjustments and basis of the training program in order to better guide patients' rehabilitation training. At the same time, it also allows patients to clearly understand their training tasks and goals, thereby enhancing their enthusiasm and compliance in participating in rehabilitation training. The assessment results are presented in various formats, including charts, graphs, and data comparisons, visually demonstrating changes in patients' movement and physiological patterns at different rehabilitation stages, as well as the degree of improvement in rehabilitation outcomes. This allows rehabilitation therapists to understand patients' progress in a timely manner, providing a basis for subsequent rehabilitation treatment decisions; it also allows patients to clearly see their rehabilitation achievements, enhancing their confidence in rehabilitation. The progress tracking module is configured to track the patient's rehabilitation training progress in real time, record the completion status, training time and intensity of each training session, compare and analyze the training progress with the expected rehabilitation plan, promptly identify any deviations or problems that may occur during the training process, and provide timely feedback to the rehabilitation therapist for timely adjustment of the training plan.

[0050] The technical effects of the above-mentioned solution are as follows: the program adjustment module, based on accurate assessment results and combined with individual patient differences and rehabilitation goals, enables dynamic and personalized adjustments to the training program, thereby ensuring that rehabilitation training always matches the patient's actual recovery status. The results display module uses diverse visualization methods to clearly and intuitively present the adjusted program and the phased assessment results, which not only ensures that rehabilitation therapists can efficiently and accurately understand the program's intent and the patient's progress to make professional decisions, but also significantly improves the patient's awareness of their rehabilitation tasks and effectiveness, effectively enhancing their treatment compliance and rehabilitation confidence. The progress tracking module, by recording and comparing training data with the expected plan in real time, constructs a continuous progress monitoring and feedback mechanism, which can promptly identify training deviations and issue warnings, thereby supporting rehabilitation therapists to quickly intervene and optimize the program.

[0051] The progress tracking module is also configured as follows: Based on the patient's historical rehabilitation training data and current rehabilitation status, a personalized rehabilitation progress prediction model is established to predict the changing trends of key indicators and the expected time to achieve the goals in future rehabilitation stages. After each rehabilitation training session, the actual training data is compared and analyzed with the personalized rehabilitation progress prediction model to calculate the rehabilitation progress deviation index. The deviation index comprehensively considers multiple dimensions such as training intensity completion rate, movement standard execution rate, and physiological parameter improvement. When the rehabilitation progress deviation index exceeds the preset warning threshold, an abnormal progress reminder is sent to the rehabilitation therapist.

[0052] The technical effects of the above-mentioned solution are as follows: By establishing a personalized rehabilitation progress prediction model, it is possible to prospectively predict the future rehabilitation trajectory and target achievement time based on the patient's historical data and current status, thereby providing a scientific basis for rehabilitation therapists to formulate long-term plans and set reasonable expectations. By comprehensively calculating the rehabilitation progress deviation index through multi-dimensional indicators, it is possible to achieve a holistic quantitative assessment of training intensity, movement quality, and physiological improvement, thereby accurately capturing subtle deviations in the rehabilitation process. When the deviation index exceeds the limit, it automatically sends an early warning to the rehabilitation therapist, enabling the therapist to promptly identify problems such as progress lag or potential rehabilitation plateaus, and thus quickly adjust the treatment plan, thereby effectively avoiding rehabilitation delays caused by untimely problem detection.

[0053] Working principle: The data acquisition unit, through a smart chip integrated into the limb, simultaneously collects motion and physiological data. The edge computing unit performs local preprocessing and initial feature extraction, effectively reducing data transmission latency and cloud load. The analysis and evaluation unit integrates multimodal data, uses a deep learning model to accurately identify abnormalities in motion and physiological patterns, and combines reinforcement learning to build a dynamic evaluation model that can accurately identify subtle abnormalities in motion and physiological patterns, thereby achieving a quantitative evaluation of rehabilitation effects. The health management unit can personalize the training plan based on the evaluation results and display the progress through a visual interface. Based on the above design, rehabilitation evaluation has been transformed from subjective experience to objective quantification, improving the accuracy and efficiency of the evaluation. Furthermore, real-time feedback and dynamic adjustment of the plan enhance the personalization level and final effect of rehabilitation training.

[0054] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0055] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention.

Claims

1. A sports injury rehabilitation effect evaluation system based on a multimodal sensor chip, characterized in that, include: The data acquisition unit, including a smart chip, is configured to integrate biosensors and physical sensors and is positioned on the limb part of the user to be evaluated, for the purpose of simultaneously collecting multi-dimensional motor physiological data of the user during the rehabilitation training process; The edge computing unit is configured to perform local real-time preprocessing and initial feature extraction on the collected motion physiological data; The analysis and evaluation unit is configured to receive preprocessed motor physiological data, integrate deep learning algorithms and reinforcement learning algorithms, construct a rehabilitation effect evaluation model, identify abnormal movement patterns and physiological patterns of patients during the rehabilitation process, and use it to evaluate the rehabilitation effect. The health management unit is configured to adjust the user's training plan based on the assessment results of rehabilitation effects, and to display the adjusted training plan and assessment results to rehabilitation therapists and patients in a visual manner; The edge computing unit includes: The data processing module is configured to synchronously acquire multi-dimensional motion physiological data from biosensors and physical sensors, and perform filtering, noise reduction, and signal normalization processing. It is also configured to attach a timestamp and signal-to-noise ratio label to each data frame based on the real-time received multi-dimensional motion physiological data stream; determine the validity of the data frame based on a preset quality threshold matrix related to the key aspects of the current rehabilitation movement; and distinguish valid data frames into high-priority physiological data streams and standard-priority motion data streams, allocating different caches and computing resources accordingly. High-priority physiological data streams are given priority for feature extraction in the feature extraction module. The validity of a data frame is determined based on a preset quality threshold matrix related to the key aspects of the current rehabilitation action, including: Obtain the quality threshold matrix that is critical to the current rehabilitation action corresponding to the data frame, and extract the indicator dimensions from the quality threshold matrix; Retrieve historical user data related to the data frame, learn the fluctuation characteristics of the historical user data in the indicator dimension based on time series analysis, compare the threshold fluctuation characteristics with the standard fluctuation characteristics for establishing a quality threshold matrix reference to obtain the fluctuation difference, and determine the individual calibration factor for the indicator dimension based on the fluctuation difference. The system acquires environmental sensor data and real-time physiological state data of the user from the collected data frames. Based on the environmental difference between the environmental sensor data and the standard environmental data, and combined with the influence of the environment on data acquisition, it determines the environmental calibration factor for the indicator dimensions. Based on the physiological difference between the user's real-time physiological state data and healthy physiological data, and combined with the influence of physiological state data on data acquisition, it determines the health calibration factor for the indicator dimensions. The target calibration factor that maximizes the calibration range of the indicator dimension is determined by the individual calibration factor, environmental calibration factor and health calibration factor. The quality threshold matrix is ​​then individually optimized based on the target calibration factor to obtain the optimized quality threshold matrix. Based on the rehabilitation stage in which the data frame is located, a threshold setting leniency coefficient is determined, and the optimized quality threshold matrix is ​​processed in stages based on the leniency coefficient to obtain the target quality threshold matrix. The data frame is compared with the target quality threshold matrix. If the dimension index value of the data frame is consistent with the target quality threshold matrix, the data frame is determined to be a valid data frame; otherwise, the data frame is determined to be invalid. Among these, critical actions are performed to determine invalid data frames; When an invalid data frame is determined to be a non-critical action, the most recent valid data frame is automatically retrieved from the local cache, and a data stream is generated to fill the invalid data frame based on the valid data frame using the polynomial interpolation method. The method for determining the most recent time period is as follows: in, Indicates a recent time period. This indicates the minimum time window set by the system. This indicates the maximum time window set by the system. Indicates the base time window size. The standard deviation of the data quality metric representing the valid data frames cached locally. This represents the average data quality metric of valid data frames cached locally. When an invalid data frame is determined to be a critical action, a priority score is assigned to the invalid data frame, and a resource allocation coefficient is determined based on the priority score. The invalid data frame is then processed based on the resource allocation coefficient. The formula for calculating the priority score of invalid data frames is as follows: in, The priority score indicating an invalid data frame. Indicates key weight, Key scores indicating invalid data frames Indicates time weight, The time score indicating an invalid data frame. Indicates spatial weights, Spatial score indicating an invalid data frame. Indicates quality weight, The quality score indicates an invalid data frame; The formula for calculating the resource allocation coefficient of an invalid data frame is: in, The resource allocation coefficient representing an invalid data frame. Represents the basic resource allocation coefficient. This represents the first reference parameter, with a value of 0.

2. This represents the second reference parameter, with a value of 0.

3. Indicates the time when the invalid data frame was last accessed. Indicates the current time. Represents the time decay constant. This represents the natural exponential function.

2. The sports injury rehabilitation effect evaluation system based on a multimodal sensor chip according to claim 1, characterized in that, The analysis and evaluation unit includes: The multimodal data fusion module, in conjunction with the principal component analysis method, fuses the preprocessed exercise physiological data to generate a comprehensive exercise physiological feature vector. The pattern recognition module is configured to identify abnormal movement patterns in patients during rehabilitation by using a hybrid architecture of recurrent neural networks and convolutional neural networks in deep learning, based on the comprehensive motor physiological feature vector. Also based on the comprehensive motor physiological feature vector, a generative adversarial network architecture is adopted to generate data samples simulating normal physiological states through a generator, and a discriminator distinguishes between the actual collected physiological data and the generated normal samples to identify abnormal physiological patterns in patients during rehabilitation. The model building module is configured to comprehensively analyze the recognition results of the pattern recognition module, use reinforcement learning algorithms to build a rehabilitation effect evaluation model, and dynamically adjust the parameters of the evaluation model according to the preset rehabilitation goals and reward mechanism to achieve the evaluation of rehabilitation effect.

3. The sports injury rehabilitation effect evaluation system based on a multimodal sensor chip according to claim 2, characterized in that, The analysis and evaluation unit also includes: The adaptive learning module is configured to generate a personalized initial movement baseline model for the user in the early stages of rehabilitation, based on the user's first achievement of standard movement data and incorporating the user's historical injury information. In each subsequent rehabilitation training session, the real-time collected exercise physiological data will be compared with the personalized initial exercise baseline model to calculate the dynamic similarity score. Based on the continuous changing trend of dynamic similarity scores, the core parameters of the rehabilitation effect evaluation model are dynamically fine-tuned through online learning algorithms, so that the evaluation criteria can adapt to the user's slowly recovering physiological functions.

4. The sports injury rehabilitation effect evaluation system based on a multimodal sensor chip according to claim 3, characterized in that, The adaptive learning module is further configured as follows: Based on the user's personalized initial exercise baseline model and historical rehabilitation training data, a time series prediction algorithm is used to predict the trajectory of changes in the user's exercise physiological data during future rehabilitation training, including the expected change curves of joint range of motion, exercise speed and heart rate. During each rehabilitation training session, the deviation between the actual collected exercise physiological data and the predicted trajectory of exercise physiological data changes is calculated in real time. The deviation value is selected from absolute deviation, relative deviation, and root mean square deviation. Based on the calculated deviation value and the preset deviation threshold, it is determined whether the rehabilitation effect evaluation model needs to be adjusted. When the deviation value exceeds the preset threshold, the online learning algorithm is triggered to automatically adjust the core parameters of the rehabilitation effect evaluation model.

5. The sports injury rehabilitation effect evaluation system based on a multimodal sensor chip according to claim 1, characterized in that, The data acquisition unit is also configured to perform real-time online calibration of the biosensors and physical sensors, specifically: The calibration is triggered by external commands, generating an electrical reference signal with known amplitude and frequency and injecting it into the front-end circuits of the biosensor and physical sensor; The response data of biosensors and physical sensors to the reference signal are collected. By comparing the response data with the expected standard value of the reference signal, the zero-point drift, sensitivity drift and nonlinear error of each sensor are calculated in real time. The system receives the calculated error data and uses a preset error compensation model to generate calibration parameters in real time. It then performs online correction and compensation on the raw motion physiological data collected by the sensor and outputs the calibrated data to the edge computing unit.

6. The sports injury rehabilitation effect evaluation system based on a multimodal sensor chip according to claim 5, characterized in that, During real-time online calibration, the clock unit and environmental sensors are used to accumulate the power-on working time of each sensor and monitor changes in ambient temperature and humidity in real time. Based on the working time and environmental change data, the calibration trigger frequency is dynamically adjusted through built-in fuzzy control rules.

7. The sports injury rehabilitation effect evaluation system based on a multimodal sensor chip according to claim 1, characterized in that, The edge computing unit further includes: The feature extraction module is configured to extract time-domain, frequency-domain, and time-frequency-domain features from preprocessed multi-dimensional motion physiological data. The compression caching module is configured to compress the preprocessed and feature-extracted exercise physiological data, and temporarily store the preprocessed and feature-extracted exercise physiological data locally. 8.The multi-modal sensor chip based motion impairment rehabilitation outcome assessment system according to claim 1, wherein, The health management unit includes: The program adjustment module is configured to analyze the patient's current rehabilitation status and needs based on the assessment results, and to make personalized adjustments to the user's training program in combination with the patient's physical condition, rehabilitation goals and sports injury type, including adjusting training intensity, training movements and training frequency. The results display module is configured to provide a human-computer interaction interface to visually present the adjusted training plan and assessment results to rehabilitation therapists and patients. The progress tracking module is configured to track the patient's rehabilitation training progress in real time, record the completion status, training time and training intensity of each training session, compare and analyze the training progress with the expected rehabilitation plan, promptly identify possible deviations or problems in the training process, and provide timely feedback to the rehabilitation therapist for timely adjustment of the training plan. It is also configured to establish a personalized rehabilitation progress prediction model based on the patient's historical rehabilitation training data and current rehabilitation status, predict the trend of key indicators and the expected time to achieve the goals in future rehabilitation stages. After each rehabilitation training session, the actual training data is compared and analyzed with the personalized rehabilitation progress prediction model to calculate the rehabilitation progress deviation index. The deviation index comprehensively considers multiple dimensions such as training intensity completion rate, movement standard execution rate, and physiological parameter improvement. When the rehabilitation progress deviation index exceeds the preset warning threshold, an abnormal progress reminder is sent to the rehabilitation therapist.

Images