Intelligent assessment method and device for body rehabilitation of stroke patient

Abstract

The invention discloses an intelligent assessment method and device for body rehabilitation of a stroke patient, and relates to the technical field of rehabilitation intelligent assessment, and the method comprises the steps: connecting a medical information management system, and obtaining the stroke medical record information of the current patient; performing identification according to the stroke medical record information; the multi-modal sensing module is connected with the rehabilitation device, performs rehabilitation motion sensing on the limb on the affected side according to the multi-modal sensor, and outputs a multi-modal sensing data set; collecting a multi-modal sensing sample set of a healthy limb corresponding to the limb on the affected side; a rehabilitation evaluation model is constructed, the rehabilitation evaluation model evaluates the multi-modal sensing data set, and rehabilitation level indexes are obtained; the rehabilitation level indexes are fed back to the rehabilitation device for training mode optimization. The technical problems that in the prior art, body rehabilitation evaluation depends on subjective judgment, the limb function recovery condition of the patient is difficult to comprehensively reflect, rehabilitation evaluation efficiency is poor, and the result is inaccurate are solved, and the technical effect of improving rehabilitation efficiency and evaluation accuracy is achieved.

Description

Technical Field

[0001] The present application relates to the technical field related to intelligent rehabilitation assessment, and in particular to a method and device for intelligent physical rehabilitation assessment of stroke patients. Background Art

[0002] Stroke (also known as stroke) is a common and serious cerebrovascular disease. Its primary manifestation is cerebral blood vessel blockage or rupture, which leads to impaired blood supply to the brain, resulting in limb movement disorders, cognitive impairment, and other sequelae. After a stroke, the patient's physical function recovery requires long-term rehabilitation therapy. The effectiveness and progress of rehabilitation therapy are often affected by the patient's specific condition, individual differences, and treatment methods. In the late stages of stroke, patients often experience varying degrees of limb paralysis, especially hemiplegia (usually loss of limb function on one side), which causes significant inconvenience in their daily lives. For stroke patients, rehabilitation therapy is an important means to promote functional recovery and improve their quality of life. Through targeted exercise, physical therapy, and neurological rehabilitation training, it promotes the recovery of affected limb function and improves patients' motor ability. However, current rehabilitation assessment methods often rely on the subjective judgment of clinicians, have long assessment cycles, and lack objectivity and accuracy. Furthermore, during data collection and analysis, they often use a single type of physiological or motor data as the assessment basis, which fails to fully reflect the patient's actual limb function recovery, affecting the accuracy of assessment results and the scientific nature of the patient's rehabilitation process.

[0003] At present, relevant technologies have the technical problem that physical rehabilitation assessment relies on subjective judgment and is difficult to fully reflect the patient's limb function recovery, resulting in poor efficiency of rehabilitation assessment and inaccurate results. Summary of the Invention

[0004] This application solves the technical problems in the prior art that physical rehabilitation assessment relies on subjective judgment and is difficult to fully reflect the patient's limb function recovery, resulting in poor rehabilitation assessment efficiency and inaccurate results by providing an intelligent physical rehabilitation assessment method and device for stroke patients, thereby achieving the technical effect of improving rehabilitation efficiency and assessment accuracy.

[0005] The present application provides an intelligent assessment method for physical rehabilitation of stroke patients, comprising: connecting to a medical information management system to obtain the current patient's stroke medical record information; identifying and determining the affected limb based on the stroke medical record information; connecting to a rehabilitation device, performing rehabilitation movement sensing on the affected limb based on a multimodal sensor, and outputting a multimodal sensing data set, wherein the multimodal sensing data set includes joint activity data, muscle tension data, limb movement data, and physiological sensing data; collecting a multimodal sensing sample set of the affected limb corresponding to the healthy limb; constructing a rehabilitation assessment model, wherein the rehabilitation assessment model evaluates the multimodal sensing data set based on the multimodal sensing sample set to obtain a rehabilitation level indicator; and feeding back the rehabilitation level indicator to the rehabilitation device for training mode optimization.

[0006] In a possible implementation, the rehabilitation level indicator is fed back to the rehabilitation device for training mode optimization, and the following processing is also performed: historical rehabilitation level indicator samples and the training mode corresponding to each historical rehabilitation level indicator sample are collected; rehabilitation level distribution data is constructed based on the rehabilitation level indicator, the historical rehabilitation level indicator samples and the training mode corresponding to each historical rehabilitation level indicator sample; the rehabilitation level distribution data is input into a Markov chain model for prediction, and the training mode is used as a transfer variable to output a training plan that achieves a preset rehabilitation level; and the training mode of the rehabilitation device is optimized according to the training plan.

[0007] In a possible implementation, a Markov chain model is constructed and the following processing is performed: a rehabilitation level state space and a training mode variable are defined, wherein the training mode variable includes a training mode type and a training mode intensity; based on the historical rehabilitation level indicator samples and the training mode corresponding to each historical rehabilitation level indicator sample, the transition probability from the previous time node to the next time node is obtained, and a transition probability matrix is constructed; training is performed according to the rehabilitation level state space, the training mode variables and the transition probability matrix until convergence, and the Markov chain model is output.

[0008] In a possible implementation, the rehabilitation assessment model evaluates the multimodal sensing data set based on the multimodal sensing sample set, and also performs the following processing: performing feature convolution on the multimodal sensing data set to output joint activity features, muscle tension features, movement coordination features, and physiological load features; identifying difference feature vectors corresponding to the joint activity features, muscle tension features, movement coordination features, and physiological load features respectively based on the multimodal sensing sample set; performing information entropy weight evaluation based on the difference feature vectors, and outputting a rehabilitation level indicator.

[0009] In a possible implementation, the intelligent assessment method for physical rehabilitation of stroke patients also performs the following processing: outputting joint activity characteristics, muscle tension characteristics, motion coordination characteristics and physiological load characteristics; wherein, the joint activity characteristics include joint angle, displacement amplitude and trajectory information, the muscle tension characteristics include EMG time domain characteristics and EMG frequency domain characteristics, the motion coordination characteristics include force output distribution and motion trajectory smoothness, and the physiological load characteristics include heart rate changes and fatigue index.

[0010] In a possible implementation, the rehabilitation assessment model evaluates the multimodal sensing data set based on the multimodal sensing sample set, and also performs the following processing: determining a multi-level rehabilitation stage node; connecting the rehabilitation device, and determining a matching rehabilitation stage node for the patient based on the multi-level rehabilitation stage node; screening a corresponding multimodal sensing sample set based on the matching rehabilitation stage node; the rehabilitation assessment model evaluates the multimodal sensing data set according to the screened multimodal sensing sample set to obtain a rehabilitation level indicator.

[0011] In a possible implementation, the intelligent assessment method for physical rehabilitation of stroke patients also performs the following processing: the connected medical information management system includes a data sharing module, and the data sharing module is used to store stroke patient samples and rehabilitation training data samples corresponding to the stroke patient samples; the data sharing module determines the identified patient sample based on the current patient's stroke medical record information, wherein the identified patient is a patient sample whose medical record information similarity is greater than a preset similarity; obtains a multimodal identification sample set of the identified patient sample; and updates the multimodal sensing sample set based on the multimodal identification sample set.

[0012] In a possible implementation, after the rehabilitation device is connected, the following processing is also performed: determining whether the rehabilitation device includes an auxiliary connection device; if the rehabilitation device includes an auxiliary connection device, obtaining an auxiliary sensor data set of the auxiliary connection device; performing an evaluation based on the auxiliary sensor data set to obtain an auxiliary rehabilitation level indicator, and updating the rehabilitation level indicator based on the auxiliary rehabilitation level indicator.

[0013] In a possible implementation, after obtaining the multimodal sensing data set, the following processing is further performed: uploading the multimodal sensing data set to a cloud processor, wherein the cloud processor includes a data processing template; and the cloud processor performs data template processing and storage on the multimodal sensing data set according to the data processing template.

[0014] The present application also provides an intelligent physical rehabilitation assessment device for stroke patients, including: a stroke medical record information acquisition module, used to connect to a medical information management system to obtain the stroke medical record information of the current patient; an affected limb determination module, used to identify and determine the affected limb based on the stroke medical record information; a rehabilitation movement sensing module, used to connect to a rehabilitation device, perform rehabilitation movement sensing on the affected limb based on a multimodal sensor, and output a multimodal sensing data set, wherein the multimodal sensing data set includes joint activity data, muscle tension data, limb movement data, and physiological sensing data; a multimodal sensing sample set acquisition module, used to collect a multimodal sensing sample set of the affected limb corresponding to the healthy limb; a rehabilitation assessment model construction module, used to construct a rehabilitation assessment model, wherein the rehabilitation assessment model evaluates the multimodal sensing data set based on the multimodal sensing sample set to obtain a rehabilitation level indicator; and a training mode optimization module, used to feed back the rehabilitation level indicator to the rehabilitation device for training mode optimization.

[0015] The intelligent physical rehabilitation assessment method and device for stroke patients proposed in this application is intended to connect to a medical information management system to obtain the current patient's stroke medical history information; identify the affected limb based on the stroke medical history information; connect to a rehabilitation device, perform rehabilitation motion sensing on the affected limb using a multimodal sensor, and output a multimodal sensor data set; collect a multimodal sensor sample set of the affected limb corresponding to the healthy limb; construct a rehabilitation assessment model, which evaluates the multimodal sensor data set to obtain rehabilitation level indicators; and feed the rehabilitation level indicators back to the rehabilitation device for training mode optimization. This solves the technical problem in the existing technology that physical rehabilitation assessment relies on subjective judgment and is difficult to fully reflect the patient's limb function recovery, resulting in poor rehabilitation assessment efficiency and inaccurate results, thereby achieving the technical effect of improving rehabilitation efficiency and assessment accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings of the embodiments of the present invention are briefly introduced below. Flowcharts are used in this application to illustrate the operations performed by the system according to the embodiments of the present application. It should be understood that the preceding or following operations are not necessarily performed in precise order. Instead, various steps may be processed in reverse order or simultaneously as needed. Furthermore, other operations may be added to these processes, or one or more operations may be removed from these processes.

[0017] Figure 1 A flow chart of the intelligent assessment method for physical rehabilitation of stroke patients provided in an embodiment of the present application.

[0018] Figure 2 This is a schematic diagram of the structure of the intelligent physical rehabilitation assessment device for stroke patients provided in an embodiment of the present application.

[0019] Explanation of the accompanying symbols: stroke medical record information acquisition module 10, affected limb determination module 20, rehabilitation movement sensing module 30, multimodal sensing sample set acquisition module 40, rehabilitation assessment model construction module 50, training mode optimization module 60. DETAILED DESCRIPTION

[0020] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application, it can be implemented in accordance with the contents of the specification. In order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are listed below.

[0021] In order to make the purpose, technical solutions and advantages of this application clearer, the application will be further described in detail below with reference to the accompanying drawings. The described embodiments should not be regarded as limiting this application. All other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of this application.

[0022] In the following description, reference is made to “some embodiments”, which describes a subset of all possible embodiments, but it will be understood that “some embodiments” may be the same subset or different subsets of all possible embodiments, and may be combined with each other without conflict, and the terms “first\second” involved are merely used to distinguish similar objects and do not represent a specific ordering of the objects. The terms “including” and “having” and any variations are intended to cover non-exclusive inclusions. For example, a process, method, product, or server that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or modules that are not clearly listed or that are inherent to these processes, methods, devices, products, or equipment. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the art to which this application belongs. The terms used herein are for the purpose of describing the embodiments of this application only.

[0023] The present application embodiment provides a method for intelligent assessment of physical rehabilitation of stroke patients, such as Figure 1 As shown, the method includes: Step S100: Connect to the medical information management system to obtain the stroke medical record information of the current patient.

[0024] Preferably, the current patient's stroke medical record information (i.e., the patient's relevant stroke medical history data) is obtained by accessing a medical information management system (such as a hospital's information management platform or electronic medical record system). Stroke medical record information typically includes basic patient information (such as name, age, gender, hospitalization number, etc.); stroke medical records, including the type of stroke diagnosis (such as hemorrhagic stroke or ischemic stroke), affected brain area, onset time, and affected limb; previous treatment and rehabilitation records, such as whether surgery has been performed, details of drug treatment, and early rehabilitation training data; imaging and laboratory examination data, such as CT, MRI results, and blood test data, used to determine the possibility of neurological function recovery; other relevant medical history, such as whether the patient has other concomitant conditions (such as diabetes or hypertension). By connecting to the medical information management system, the current patient's medical record information can be obtained to ensure the accuracy and pertinence of subsequent evaluations, while reducing manual intervention in data entry, improving efficiency and the real-time nature of data.

[0025] Step S200: Identify the affected limb based on the stroke medical record information.

[0026] Preferably, the stroke medical record information extracted from the medical information management system is used to identify the affected limb, that is, the limb of the patient affected by the stroke, through analysis and judgment. Specifically, the specific description in the medical record is checked, such as "right limb weakness" and "left limb paralysis" and other clear descriptions to directly determine the affected limb; the anatomical site of the stroke lesion (such as the left cerebral hemisphere or the right cerebral hemisphere) is extracted from the medical record information. Since stroke usually manifests as a mirror image relationship between the dysfunction of the lesion side and the contralateral limb, the lesion in the left brain will cause damage to the right limb, and vice versa; if the medical record information includes neurological Examination results (such as muscle strength scores, reflex tests, etc.) can be analyzed to further confirm the damaged limb. For example, if the record shows "right upper limb muscle strength level 2, right lower limb muscle strength level 3, and the left side is normal", it can be inferred that the right side is the affected limb. If the medical record contains imaging examination results (such as CT, MRI), the location of the stroke lesion can be located, combined with the common distribution pattern of functional disorders after stroke, to assist in confirming the affected limb. For example, damage to the left motor cortex of the brain usually corresponds to movement disorders of the right limb. The affected limb can then be automatically and accurately identified, and the clarity of the goals of subsequent rehabilitation training and evaluation can be ensured.

[0027] Step S300 , connecting a rehabilitation device, performing rehabilitation motion sensing on the affected limb according to a multimodal sensor, and outputting a multimodal sensing data set, wherein the multimodal sensing data set includes joint activity data, muscle tension data, limb movement data, and physiological sensing data.

[0028] Preferably, by connecting to a rehabilitation device, a multimodal sensor is used to perform rehabilitation motion sensing on the affected limb, that is, to collect motion-related data of the patient's affected limb in real time to form a multimodal sensing data set to help evaluate and optimize the rehabilitation effect. Among them, a multimodal sensor refers to a sensor that integrates multiple different functions and can comprehensively monitor the patient's motion and physiological state from different angles. The multimodal sensing data set includes joint activity data, muscle tension data, limb movement data, and physiological sensing data. Specifically, joint activity data refers to parameters such as the angle, amplitude, and speed of each movement of the patient's limb joints during the rehabilitation process. Angle sensors, optical sensors, inertial sensors, etc. are often used for monitoring. By measuring the range of motion of the joints (such as flexion and extension angles, abduction and adduction angles, etc.), the flexibility and mobility of the patient's joints can be understood; muscle tension data refers to the relevant data for measuring the tension and relaxation of the patient's limb muscles. Surface electromyography (EMG) sensors or force sensors can be used to obtain information such as the strength, duration and frequency of muscle contraction. Stroke patients often have abnormal muscle tension, such as excessive muscle tension (such as spasticity) or too low muscle tension (such as muscle weakness). By monitoring changes in muscle tension, the progress of nerve recovery and the functional status of muscles can be evaluated.

[0029] Preferably, limb motion data includes data such as the patient's limb movement pattern, movement speed, and gait analysis, which are usually collected through sensors such as inertial measurement units (IMUs), accelerometers, and gyroscopes. Limb motion data is used to monitor the patient's movement quality and motor function recovery, and can determine the patient's movement coordination, fluency, and stability. For example, an abnormal gait may indicate that the patient still has certain movement disorders; physiological sensor data refers to data collected from other physiological states of the patient's body, such as heart rate, blood pressure, body temperature, blood oxygen saturation, etc., which are usually obtained through physiological sensors (such as heart rate sensors, blood oxygen sensors, etc.). Physiological sensor data can help monitor the patient's physical condition during the rehabilitation process, especially for stroke patients. Controlling physiological parameters such as blood pressure and heart rate is crucial to preventing complications and overtraining. By comprehensively obtaining data at different levels through different types of sensors, it helps to evaluate the patient's rehabilitation status more comprehensively and in real time, thereby supporting smart rehabilitation devices to perform more precise intervention and rehabilitation training.

[0030] Furthermore, step S300 also includes step S310, uploading the multimodal sensing data set to a cloud processor, wherein the cloud processor includes a data processing template; step S320, the cloud processor performs data template processing and stores the multimodal sensing data set according to the data processing template.

[0031] Preferably, the multimodal sensory data (such as joint activity data, muscle tension data, limb movement data, physiological data, etc.) collected by the rehabilitation equipment is uploaded to a remote cloud processing platform via a wireless network (such as Wi-Fi, Bluetooth, 4G, 5G, etc.) for centralized storage and processing, facilitating data sharing across regions and multiple devices. The cloud processor includes a data processing template, which is a standardized data processing rule preset in the cloud platform. It can be regarded as a data processing framework that defines how to convert the raw sensory data into a standardized format, including how to perform data cleaning, organization, classification, standardization and other operations, thereby ensuring data consistency and availability; The cloud processor performs template processing on the uploaded multimodal sensor data set based on the preset data processing template, that is, it structures and standardizes the sensor data, converts the original sensor data into a format that conforms to the template rules and stores it. This includes data cleaning (removing invalid or abnormal data, such as sensor noise or erroneous data), data standardization (unifying data collected by different sensors into the same standard format), data classification (classifying data by type, such as joint activity data, muscle tension data, limb movement data, etc.) and data aggregation (merging data from different sources into one data set), providing a high-quality data foundation for subsequent in-depth analysis, evaluation and intelligent decision-making.

[0032] Step S400: collecting a multimodal sensor sample set of the affected limb corresponding to the healthy limb.

[0033] Preferably, multimodal sensory monitoring is performed on the patient's healthy limb, unaffected by the stroke, to obtain relevant data on the healthy limb's movement, muscle tone, physiological characteristics, and so on. This multimodal sensory sample set is used as a baseline sample for subsequent rehabilitation assessment and comparative analysis. Specifically, a healthy limb refers to a limb on the side not directly affected by the stroke, whose movement ability and physiological function are close to normal. The data from the healthy limb can be used to assess the rehabilitation level and functional recovery of the affected limb. The sensory sample set of the healthy limb is collected using multiple types of sensors, including joint movement data, muscle tone data, limb movement data, and physiological sensor data. By comparing the data from the affected limb with that of the healthy limb, the patient's rehabilitation effect can be more accurately assessed and the rehabilitation plan optimized.

[0034] Furthermore, step S400 also includes step S410, wherein the connected medical information management system includes a data sharing module, and the data sharing module is used to store stroke patient samples and rehabilitation training data samples corresponding to the stroke patient samples; step S20, the data sharing module determines the identified patient sample based on the current patient's stroke medical record information, wherein the identified patient is a patient sample whose medical record information similarity is greater than a preset similarity; step S430, obtaining a multimodal identification sample set of the identified patient sample; step S440, updating the multimodal sensing sample set according to the multimodal identification sample set.

[0035] Preferably, the data sharing module is used to centrally store and manage the data of stroke patients, including patient sample data and rehabilitation training data samples corresponding to the stroke patient samples. The stroke patient sample refers to a data set containing the patient's medical history information, such as diagnosis results, imaging data (CT / MRI), rehabilitation assessment results, etc. The rehabilitation training data sample refers to the sensor data record of the patient during rehabilitation training, including joint movement, muscle tension, movement pattern, physiological indicators, etc. The data sharing module determines the identification of the patient sample based on the current patient's stroke medical record information, that is, by comparing the current patient's stroke medical record information with the patient sample data stored in the database, other patient samples whose similarity with the current patient's condition, medical history, etc. is greater than a preset threshold are screened out. Specifically, according to the patient's medical record information (such as lesion location, stroke type, degree of damage, etc.), the natural language processing (NLP) is used to identify the patient sample. ) algorithm or data matching algorithm to calculate the similarity and compare it with a preset similarity threshold, wherein the preset similarity threshold is used to identify whether the target is a patient, for example, patient samples with a similarity of more than 70% are selected as identified patients; then, from the screened identified patients, the multimodal identification sample set corresponding to the identified patient samples is extracted, that is, multimodal sensing data and rehabilitation training records are obtained, including joint movement, muscle tension, limb movement, and physiological data; finally, the multimodal identification sample set of the identified patient is compared and analyzed with the multimodal sensing sample set of the current patient, and the rehabilitation assessment data of the current patient is updated according to the data of the identified patient, including comparing the current patient's motion data (such as joint range of motion, muscle tension, etc.) with similar data of the identified patient, and updating the multimodal sensing sample set, thereby achieving more accurate rehabilitation assessment and dynamic program adjustment, and improving rehabilitation efficiency and treatment effect.

[0036] Step S500 : constructing a rehabilitation assessment model, wherein the rehabilitation assessment model evaluates the multimodal sensing data set according to the multimodal sensing sample set to obtain a rehabilitation level indicator.

[0037] Preferably, the rehabilitation assessment model may be an assessment model constructed and trained based on a deep learning algorithm, used to analyze and evaluate multimodal sensory data sets collected from the patient's limbs (the affected limb and the healthy limb) to generate a rehabilitation level indicator to reflect the patient's rehabilitation status and progress. Specifically, the assessment model is constructed based on deep learning, and supervised training is performed using data in the multimodal sensory sample set (healthy limb data or identified patient data) as model training samples to learn characteristics such as movement patterns and muscle tension ranges of healthy limbs or identified patients with good recovery. The rehabilitation assessment model is then used to evaluate the multimodal sensory data set based on the multimodal sensory sample set, analyze the degree of deviation between the input data and the training samples, and output a quantitative indicator based on the degree of deviation, namely, a rehabilitation level indicator, which is used to represent the patient's current degree of rehabilitation. Specifically, the rehabilitation assessment model compares the multimodal sensory data of the affected limb with the healthy limb sample data or the identified patient sample data, and generates a rehabilitation level indicator through comparative analysis to represent the degree of proximity of the patient's recovery status to a healthy state or a reference state, such as motor ability, muscle tension recovery (coordination of muscle contraction and relaxation), and physiological state. This makes rehabilitation assessment more accurate and efficient, and can not only dynamically track the patient's rehabilitation progress, but also provide a scientific basis for rehabilitation training.

[0038] Furthermore, step S500 also includes step S510, performing feature convolution on the multimodal sensing data set to output joint activity features, muscle tension features, movement coordination features and physiological load features; step S520, identifying the difference feature vectors corresponding to the joint activity features, muscle tension features, movement coordination features and physiological load features respectively according to the multimodal sensing sample set; step S530, performing information entropy weight evaluation based on the difference feature vectors, and outputting a rehabilitation level index.

[0039] Preferably, feature convolution is performed on the multimodal sensing dataset. Feature convolution is a common operation in deep learning and is usually used in convolutional neural networks (CNNs). The multimodal sensing dataset is filtered through the convolution layer to extract key local features, including joint movement features, muscle tension features, motion coordination features and physiological load features. Specifically, joint movement features refer to motion features such as the range of motion and flexibility of the joints extracted through sensor data (such as angle, speed, acceleration, etc.); muscle tension features refer to muscle tension-related features extracted from electromyography (EMG) data, such as muscle contraction intensity and duration; motion coordination features refer to the analysis of limb movement patterns to evaluate the patient's motion coordination and fine motor skills, such as gait analysis, balance, movement fluency, etc.; physiological load features refer to the extraction of physiological data (such as heart rate, blood pressure, blood oxygen saturation, etc.) to obtain physiological load-related features.

[0040] Preferably, by comparing various features (such as joint movement features, muscle tension features, etc.) obtained from the current patient's multimodal sensory data set with reference data (such as healthy limb data, data of the identified patient sample set), a difference feature vector (i.e., a difference feature vector) is obtained. Each feature (joint movement, muscle tension, movement coordination, physiological load) corresponds to a difference vector, i.e., the difference between the feature in the current patient and the healthy or identified patient. For example, if the current patient's range of joint movement is smaller than that of the healthy sample, then the difference feature vector will show a large numerical difference in the joint movement dimension; then, an information entropy weight evaluation is performed based on the difference feature vector. Specifically, in the rehabilitation assessment, by calculating the information entropy of each feature (such as joint movement, muscle tension, movement coordination, etc.), the contribution of each feature to the overall rehabilitation level is evaluated. Features with higher entropy values indicate that the feature has greater uncertainty and may have stronger diagnostic significance; features with lower entropy values indicate that they are more stable and have less variation, and may have less impact on the rehabilitation assessment; then, a rehabilitation level indicator is output to indicate the patient's current rehabilitation progress or effect, thereby accurately and dynamically evaluating the patient's rehabilitation progress, providing a scientific basis for personalized rehabilitation treatment, and improving treatment effects and patient experience.

[0041] Furthermore, step S510 also includes that the joint movement characteristics include joint angle, displacement amplitude and trajectory information, the muscle tension characteristics include EMG time domain characteristics and EMG frequency domain characteristics, the motion coordination characteristics include force output distribution and motion trajectory smoothness, and the physiological load characteristics include heart rate changes and fatigue index.

[0042] Preferably, the joint activity characteristics reflect the dynamic performance of the patient's joints during movement, including joint angles (rotation angles or flexion and extension angles of the joints during movement, such as the flexion and extension angle range of the elbow joint, the bending angle of the knee joint, etc.), displacement amplitude (the spatial distance moved from one position to another during joint movement, such as the maximum height of the arm raised) and trajectory information (the specific path and trajectory of the joint movement in space, such as whether the trajectory of the patient's fingers is smooth and complete when drawing a circle); muscle tension characteristics are used to evaluate the state of the patient's muscles, including EMG time domain characteristics (statistical characteristics of electromyographic signals on the time axis, including mean, peak, standard deviation, variance, etc.) and EMG frequency domain characteristics (frequency component characteristics of electromyographic signals, such as spectral power, mean frequency rate, median frequency, etc.); movement coordination characteristics are used to assess the patient's ability to coordinate multiple muscle groups and joints during exercise, including force output distribution (the temporal or spatial distribution of force on the limbs during exercise, such as whether the output on both sides is balanced when both legs exert force) and movement trajectory smoothness (the smoothness of the movement trajectory in space or time, such as the curvature of the trajectory curve or the smoothness of the speed change); physiological load characteristics reflect the physiological stress on the patient during rehabilitation exercise, mainly including heart rate changes (real-time fluctuations in heart rate during exercise, such as the increase in heart rate, heart rate recovery time, etc.) and fatigue index (the patient's fatigue level calculated based on physiological and exercise data, such as the assessment of fatigue level through the increase in low-frequency components of the electromyographic signal and the extension of the heart rate recovery time).

[0043] Furthermore, step S500 also includes step S540, determining a multi-level rehabilitation stage node; step S550, connecting the rehabilitation device and determining a matching rehabilitation stage node for the patient based on the multi-level rehabilitation stage node; step S560, screening a corresponding multimodal sensing sample set based on the matching rehabilitation stage node; and step S570, the rehabilitation evaluation model evaluates the multimodal sensing data set according to the screened multimodal sensing sample set to obtain a rehabilitation level indicator.

[0044] Preferably, a multi-level rehabilitation stage node is determined, namely, the hospital stage, the community stage, and the home stage, wherein the hospital stage refers to patients receiving high-intensity, specialized rehabilitation training in a professional medical environment, usually for patients in the acute phase or with severe functional impairment; the community stage refers to patients receiving moderate-intensity rehabilitation training in a rehabilitation institution or community center, suitable for patients in the recovery phase; the home stage refers to patients receiving low-intensity, maintenance rehabilitation training in a home environment, usually suitable for patients in the chronic phase or with better functional recovery; the full-process rehabilitation management of the hospital, community, and home ensures that the rehabilitation needs of patients at different stages are met, while improving the overall rehabilitation efficiency through seamless connection across stages; then, the rehabilitation device is connected, and the patient's matching rehabilitation stage node is determined based on the multi-level rehabilitation stage node, that is, the patient's rehabilitation assessment data and actual conditions (such as the degree of functional recovery, environmental conditions, training intensity requirements, etc.) are used to determine which rehabilitation stage node the patient should currently be in. Specifically, the rehabilitation assessment model is used to assess the patient's rehabilitation level indicators (such as mobility, muscle tone recovery, etc.), and the patient is matched to the most appropriate stage node according to the preset rehabilitation stage classification criteria, ensuring that the patient receives rehabilitation training in an appropriate environment and avoiding switching rehabilitation stages too early or too late.

[0045] Preferably, according to the rehabilitation stage node to which the patient is matched, a multimodal sensor sample set of that stage (including rehabilitation data of other similar patients at that stage) is screened out. The multimodal sensor sample set includes common joint movement characteristics, muscle tension characteristics, movement coordination characteristics, and physiological load characteristics of that stage, thereby providing a stage-by-stage and targeted rehabilitation assessment benchmark, making the assessment results more accurate. For example, in the home stage, the sample set may focus more on the assessment of low-intensity exercise; in the hospital stage, the sample set may focus more on high-intensity training and the recovery of complex motor functions; then the rehabilitation assessment model evaluates the patient's actual multimodal sensor data set based on the screened multimodal sensor sample set, that is, the screened multimodal sensor sample set (as a benchmark) and the current patient's multimodal sensor data set (actual data) are input, the rehabilitation assessment model compares the feature differences between the patient's current data and the sample set, analyzes the importance of each feature in the overall rehabilitation level using information entropy weight evaluation, and outputs a rehabilitation level index, which represents the patient's current functional recovery at the matched rehabilitation stage node. Through the dynamic management of multi-level rehabilitation stage nodes (hospitals, communities, and families), combined with the multimodal sensor sample sets screened for specific stages, the rehabilitation assessment model is used to accurately evaluate the patient's data and generate rehabilitation level indicators to ensure that patients receive appropriate rehabilitation training at the correct stage, thereby achieving efficient, scientific, and personalized full-process rehabilitation management.

[0046] Furthermore, step S500 also includes step S580, determining whether the rehabilitation device includes an auxiliary connection device; step S590, if the rehabilitation device includes an auxiliary connection device, obtaining an auxiliary sensor data set of the auxiliary connection device; step S5100, performing an evaluation based on the auxiliary sensor data set to obtain an auxiliary rehabilitation level indicator, and updating the rehabilitation level indicator based on the auxiliary rehabilitation level indicator.

[0047] Preferably, after connecting the rehabilitation device, it is determined whether the rehabilitation device has an auxiliary connection device, and further the patient's rehabilitation level is evaluated based on the data collected by the auxiliary device, and the original rehabilitation level index is updated. Specifically, it is confirmed whether the rehabilitation device used has additional auxiliary connection devices. The auxiliary connection device is a device used in conjunction with the main rehabilitation device to provide more sensor data and further improve the treatment effect. For example, the rehabilitation device is a manipulator used to help patients perform arm activity training. The auxiliary connection device can be an auxiliary sensor such as a wristband or a belt to monitor other physiological parameters or motion data. If the rehabilitation device includes an auxiliary connection device, the sensors attached to the auxiliary connection device (such as wearable devices, body The auxiliary sensor data set (such as gait, heart rate, body position changes, limb activities, additional joint data, etc.) is obtained through the use of auxiliary sensor data sets; the data in the auxiliary sensor data sets are then analyzed and evaluated to evaluate the patient's rehabilitation level and obtain auxiliary rehabilitation level indicators, which can be combined with the rehabilitation level indicators obtained by the main rehabilitation device to help comprehensively evaluate the patient's overall rehabilitation progress, which may include the patient's motor ability (such as gait balance, walking speed, etc.), physiological recovery (such as blood oxygen level, heart rate recovery, etc.) and functional improvement (such as muscle strength, range of joint motion, etc.). Finally, the auxiliary rehabilitation level indicators are combined with the rehabilitation level indicators obtained by the main rehabilitation equipment to update the patient's overall rehabilitation assessment, thereby improving the accuracy, efficiency and effect of rehabilitation training.

[0048] Step S600: Feedback the rehabilitation level indicator to the rehabilitation device to optimize the training mode.

[0049] Preferably, the rehabilitation level indicator is used to adjust the training program or training parameters of the rehabilitation device in real time to ensure that the training intensity, content and rhythm match the patient's rehabilitation needs and physical condition. That is, the rehabilitation level indicator generated by the rehabilitation assessment model is transmitted to the rehabilitation device through a feedback mechanism, guiding the device to dynamically adjust the training mode according to the patient's condition, which may include adjusting the training intensity, such as increasing the load, speed or number of repetitions; adjusting the training content, such as upgrading from simple joint movement training to complex coordination training; adjusting the training rhythm, such as shortening the training time interval or slowing down the exercise rhythm; providing rest suggestions, and pausing the training or reducing the intensity if the indicator shows that the patient is tired or the physiological load is too high; specifically, the rehabilitation device is the core equipment for executing the training program. Through the feedback of the rehabilitation level indicator, the resistance, load or range of motion in the training can be automatically increased or decreased according to the patient's recovery status. The type of training task can be adjusted to adapt to the patient's rehabilitation needs and recovery stage. The training time, interval and movement rhythm can be dynamically adjusted according to the rehabilitation level indicator. Visual, auditory or tactile feedback can also be combined to help the patient complete the training movements and improve the training effect, thereby achieving comprehensive and accurate rehabilitation training, improving rehabilitation efficiency, and reducing the patient's rehabilitation risk.

[0050] Furthermore, step S600 also includes step S610, collecting historical rehabilitation level indicator samples and the training mode corresponding to each historical rehabilitation level indicator sample; step S620, constructing rehabilitation level distribution data based on the rehabilitation level indicator, the historical rehabilitation level indicator samples and the training mode corresponding to each historical rehabilitation level indicator sample; step S630, inputting the rehabilitation level distribution data into the Markov chain model for prediction, using the training mode as the transfer variable, and outputting a training plan that reaches a preset rehabilitation level; step S640, optimizing the training mode of the rehabilitation device according to the training plan.

[0051] Preferably, historical rehabilitation level indicator samples and the training mode corresponding to each historical rehabilitation level indicator sample are collected from hospital rehabilitation data, community rehabilitation center records, training logs of intelligent rehabilitation devices, rehabilitation data sets stored in the cloud, etc., wherein the historical rehabilitation level indicator samples come from the rehabilitation data of other patients, and include rehabilitation level indicators for each stage (such as joint range of motion, muscle tension recovery, movement coordination, etc.), and each historical rehabilitation level indicator sample corresponds to a training mode, such as training intensity, content, frequency, etc., and then rehabilitation level distribution data is constructed based on the rehabilitation level indicators, historical rehabilitation level indicator samples and the training mode corresponding to each historical rehabilitation level indicator sample. Specifically, the data of the current patient is combined with the historical sample data to form a comprehensive rehabilitation level indicator data set, the relationship between the rehabilitation level indicator and the training mode is analyzed, the influence characteristics of the training mode on the rehabilitation effect are extracted, and the rehabilitation level change trend distribution is generated, that is, rehabilitation level distribution data is constructed to show the distribution of rehabilitation effects under different training modes.

[0052] Preferably, the rehabilitation level distribution data is input into the Markov chain model for prediction, and the possibility of the current rehabilitation level state being transferred to a higher rehabilitation level under different training modes is analyzed, wherein the Markov chain is a statistical model that describes the transition process of the system between different states, based on the memoryless assumption that "the current state determines the next state". Specifically, the state is the patient's rehabilitation level indicator (such as "primary", "intermediate", "advanced", etc.), the transfer variable is the training mode (such as intensity, content, frequency, etc.), and the transfer probability is the probability distribution of the change in rehabilitation level caused by different training modes, that is, the training plan for predicting and outputting the patient to reach the target rehabilitation level (such as the preset "advanced rehabilitation" state) under a specific training mode. The preset rehabilitation level refers to the target rehabilitation state set by the rehabilitation expert, such as restoring normal walking ability, restoring more than 90% of the joint range of motion, etc. The training plan is the prediction result of the Markov chain model (such as a combination of training modes for patients), including training intensity (appropriate resistance, load or training time), training content (targeted sports tasks such as gait training, balance training, joint flexibility training, etc.) and training rhythm (training frequency, time interval between each training). The parameters of the rehabilitation device and the training content are adjusted according to the predicted training plan to match the patient's needs and target rehabilitation level, including gradually increasing or decreasing the training resistance and frequency according to the patient's current rehabilitation level indicators, introducing new training tasks for patients or increasing the diversity of training (such as switching from strength training to coordination training), and adjusting the training time interval or rest time according to the patient's real-time feedback to avoid excessive fatigue, achieve dynamic adjustment and precise management, and effectively improve the patient's rehabilitation effect and reduce risks.

[0053] Furthermore, step S630 also includes step S631, defining the rehabilitation level state space and training mode variables, wherein the training mode variables include the training mode type and the training mode intensity; step S632, obtaining the transition probability from the previous time node to the next time node according to the historical rehabilitation level indicator samples and the training mode corresponding to each historical rehabilitation level indicator sample, and constructing a transition probability matrix; step S633, training until convergence according to the rehabilitation level state space, training mode variables and the transition probability matrix, and outputting the Markov chain model.

[0054] Preferably, the rehabilitation level state space refers to the different possible states of the patient during the rehabilitation process, and the patient's rehabilitation process is converted into discrete states in the mathematical model, so that it can be transferred and predicted between different states. The training mode variable includes two main factors that affect the patient's rehabilitation, namely the training type (such as joint activity training, muscle strength training, coordination training, balance training, etc.) and the training intensity (exercise intensity, training time, frequency, etc.). Specifically, the rehabilitation level state space and the training mode variable are defined, and the state space , indicating the state of rehabilitation level and training mode , representing different training modes (including training mode type and training mode intensity), and then based on the historical rehabilitation level data and the corresponding training mode, the transition probability from one rehabilitation state to another is calculated. The transition probability refers to the probability of transitioning from one rehabilitation level state (current time node) to another rehabilitation level state (next time node). For example, assuming that a patient transitions from the initial state (state 0) to the intermediate state (state 1) more often during high-intensity gait training, it indicates that the patient is more likely to make progress under high-intensity training. The state transition probability matrix is constructed: ; in, Indicates the current status In training mode Transfer to state The probability of and training mode As input, the Markov chain is used to iteratively calculate the future recovery state: ; Then iterate until the prediction result reaches the preset recovery level , convergent output Markov chain model.

[0055] In the above, refer to Figure 1 The intelligent assessment method for physical rehabilitation of stroke patients according to an embodiment of the present invention is described in detail. Figure 2The intelligent physical rehabilitation assessment device for stroke patients according to an embodiment of the present invention is described.

[0056] The intelligent physical rehabilitation assessment device for stroke patients, according to an embodiment of the present invention, is designed to address the technical issues in existing technologies where physical rehabilitation assessments rely on subjective judgment, fail to fully reflect the patient's limb function recovery, and result in poor rehabilitation assessment efficiency and inaccurate results. This device achieves the technical effect of improving rehabilitation efficiency and assessment accuracy. The intelligent physical rehabilitation assessment device for stroke patients includes: a stroke medical record information acquisition module 10, an affected limb determination module 20, a rehabilitation motion sensing module 30, a multimodal sensor sample set acquisition module 40, a rehabilitation assessment model construction module 50, and a training mode optimization module 60.

[0057] A stroke medical record information acquisition module 10 is used to connect to a medical information management system to obtain the stroke medical record information of the current patient; an affected limb determination module 20 is used to identify and determine the affected limb based on the stroke medical record information; a rehabilitation movement sensing module 30 is used to connect to a rehabilitation device, perform rehabilitation movement sensing on the affected limb based on a multimodal sensor, and output a multimodal sensing data set, wherein the multimodal sensing data set includes joint activity data, muscle tension data, limb movement data, and physiological sensing data; a multimodal sensing sample set acquisition module 40 is used to collect a multimodal sensing sample set of the healthy limb corresponding to the affected limb; a rehabilitation assessment model construction module 50 is used to construct a rehabilitation assessment model, wherein the rehabilitation assessment model evaluates the multimodal sensing data set based on the multimodal sensing sample set to obtain a rehabilitation level indicator; a training mode optimization module 60 is used to feed back the rehabilitation level indicator to the rehabilitation device for training mode optimization.

[0058] The specific configuration of the training mode optimization module 60 will be described in detail below. The training mode optimization module 60 may further include: collecting historical rehabilitation level indicator samples and the training mode corresponding to each historical rehabilitation level indicator sample; constructing rehabilitation level distribution data based on the rehabilitation level indicator, the historical rehabilitation level indicator samples, and the training mode corresponding to each historical rehabilitation level indicator sample; inputting the rehabilitation level distribution data into a Markov chain model for prediction, using the training mode as a transfer variable, and outputting a training plan that achieves a preset rehabilitation level; and optimizing the training mode of the rehabilitation device according to the training plan.

[0059] The specific configuration of the training mode optimization module 60 will be described in detail below. The training mode optimization module 60 may further include: defining a rehabilitation level state space and training mode variables, wherein the training mode variables include a training mode type and a training mode intensity; obtaining the transition probability from the previous time node to the next time node based on the historical rehabilitation level indicator samples and the training mode corresponding to each historical rehabilitation level indicator sample, and constructing a transition probability matrix; training according to the rehabilitation level state space, training mode variables, and the transition probability matrix until convergence, and outputting the Markov chain model.

[0060] The specific configuration of the rehabilitation assessment model construction module 50 will be described in detail below. The rehabilitation assessment model construction module 50 may further include: performing feature convolution on the multimodal sensor data set to output joint activity features, muscle tension features, movement coordination features, and physiological load features; identifying difference feature vectors corresponding to the joint activity features, muscle tension features, movement coordination features, and physiological load features based on the multimodal sensor sample set; performing information entropy weight evaluation based on the difference feature vectors to output a rehabilitation level indicator.

[0061] The specific configuration of the rehabilitation assessment model construction module 50 will be described in detail below. The rehabilitation assessment model construction module 50 may further include outputting joint activity characteristics, muscle tension characteristics, motion coordination characteristics, and physiological load characteristics; wherein the joint activity characteristics include joint angle, displacement amplitude, and trajectory information; the muscle tension characteristics include EMG time domain characteristics and EMG frequency domain characteristics; the motion coordination characteristics include force output distribution and motion trajectory smoothness; and the physiological load characteristics include heart rate variability and fatigue index.

[0062] The specific configuration of the rehabilitation assessment model construction module 50 will be described in detail below. The rehabilitation assessment model construction module 50 may further include: determining a multi-level rehabilitation stage node; connecting the rehabilitation device and determining a matching rehabilitation stage node for the patient based on the multi-level rehabilitation stage node; screening a multimodal sensor sample set corresponding to the matching rehabilitation stage node; and evaluating the multimodal sensor data set using the screened multimodal sensor sample set by the rehabilitation assessment model to obtain a rehabilitation level indicator.

[0063] The specific configuration of the multimodal sensing sample set acquisition module 40 will be described in detail below. The multimodal sensing sample set acquisition module 40 may further include: the connection to the medical information management system includes a data sharing module, the data sharing module is used to store stroke patient samples and rehabilitation training data samples corresponding to the stroke patient samples; the data sharing module determines an identified patient sample based on the current patient's stroke medical record information, wherein the identified patient is a patient sample whose medical record information similarity is greater than a preset similarity; obtains a multimodal identified sample set of the identified patient sample; and updates the multimodal sensing sample set based on the multimodal identified sample set.

[0064] The specific configuration of the rehabilitation assessment model construction module 50 will be described in detail below. The rehabilitation assessment model construction module 50 may further include: determining whether the rehabilitation device includes an auxiliary connection device; if the rehabilitation device includes an auxiliary connection device, obtaining an auxiliary sensor data set from the auxiliary connection device; performing an assessment based on the auxiliary sensor data set to obtain an auxiliary rehabilitation level indicator; and updating the rehabilitation level indicator based on the auxiliary rehabilitation level indicator.

[0065] The specific configuration of the rehabilitation movement sensing module 30 will be described in detail below. The rehabilitation movement sensing module 30 may further include: uploading the multimodal sensing dataset to a cloud processor, wherein the cloud processor includes a data processing template; and the cloud processor performs data template processing and storage on the multimodal sensing dataset according to the data processing template.

[0066] The above specific embodiments do not constitute a limitation to the scope of protection of this application. It should be understood by those skilled in the art that various modifications, combinations and substitutions can be made according to design requirements and other factors. Any modifications, equivalent replacements and improvements made within the spirit and principles of this application should be included in the scope of protection of this application. In some cases, the actions or steps recorded in this application can be performed in an order different from that in the embodiments and can still achieve the desired results. In addition, the processes depicted in the accompanying drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

Claims

1. An intelligent assessment method for physical rehabilitation of stroke patients, characterized in that: The method comprises: Connect to the medical information management system to obtain the current patient's stroke medical record information; Identify the affected limb based on the stroke medical record information; Connecting to a rehabilitation device, performing rehabilitation motion sensing on the affected limb according to a multimodal sensor, and outputting a multimodal sensing data set, wherein the multimodal sensing data set includes joint movement data, muscle tension data, limb movement data, and physiological sensing data; Collecting a multimodal sensor sample set of the affected limb corresponding to the healthy limb; Constructing a rehabilitation assessment model, wherein the rehabilitation assessment model evaluates the multimodal sensing data set according to the multimodal sensing sample set to obtain a rehabilitation level indicator; The rehabilitation level indicator is fed back to the rehabilitation device to optimize the training mode.

2. The intelligent assessment method for physical rehabilitation of stroke patients according to claim 1, characterized in that: Feeding back the rehabilitation level indicator to the rehabilitation device to optimize the training mode, the method further includes: Collect historical rehabilitation level indicator samples and the training mode corresponding to each historical rehabilitation level indicator sample; constructing rehabilitation level distribution data according to the rehabilitation level indicator, the historical rehabilitation level indicator samples, and the training mode corresponding to each historical rehabilitation level indicator sample; Inputting the rehabilitation level distribution data into a Markov chain model for prediction, using the training mode as a transfer variable, and outputting a training plan that achieves a preset rehabilitation level; The training mode of the rehabilitation device is optimized according to the training plan.

3. The intelligent assessment method for physical rehabilitation of stroke patients according to claim 2, characterized in that: Methods for constructing Markov chain models include: defining a rehabilitation level state space and training mode variables, wherein the training mode variables include a training mode type and a training mode intensity; According to the historical rehabilitation level indicator samples and the training mode corresponding to each historical rehabilitation level indicator sample, obtaining the transition probability from the previous time node to the next time node, and constructing a transition probability matrix; The training is performed according to the rehabilitation level state space, the training pattern variables and the transition probability matrix until convergence, and the Markov chain model is output.

4. The intelligent assessment method for physical rehabilitation of stroke patients according to claim 1, wherein: The rehabilitation assessment model evaluates the multimodal sensing data set according to the multimodal sensing sample set, and the method includes: Performing feature convolution on the multimodal sensing data set to output joint activity features, muscle tension features, movement coordination features, and physiological load features; Identifying, based on the multimodal sensing sample set, difference feature vectors corresponding to the joint activity feature, the muscle tension feature, the movement coordination feature, and the physiological load feature; An information entropy weight evaluation is performed based on the difference feature vector, and a rehabilitation level index is output.

5. The intelligent assessment method for physical rehabilitation of stroke patients according to claim 4, characterized in that: Output joint activity characteristics, muscle tension characteristics, movement coordination characteristics and physiological load characteristics; Among them, the joint movement characteristics include joint angle, displacement amplitude and trajectory information, the muscle tension characteristics include EMG time domain characteristics and EMG frequency domain characteristics, the movement coordination characteristics include force output distribution and movement trajectory smoothness, and the physiological load characteristics include heart rate changes and fatigue index.

6. The intelligent assessment method for physical rehabilitation of stroke patients according to claim 1, wherein: The rehabilitation assessment model evaluates the multimodal sensing dataset based on the multimodal sensing sample set, and the method further includes: Identify multi-level rehabilitation phase nodes; connecting the rehabilitation device and determining a matching rehabilitation stage node for the patient based on the multi-level rehabilitation stage nodes; Filtering a multimodal sensing sample set corresponding to the matching rehabilitation stage node; The rehabilitation assessment model evaluates the multimodal sensing data set according to the screened multimodal sensing sample set to obtain a rehabilitation level indicator.

7. The intelligent assessment method for physical rehabilitation of stroke patients according to claim 1, wherein: The connected medical information management system includes a data sharing module, and the data sharing module is used to store stroke patient samples and rehabilitation training data samples corresponding to the stroke patient samples; The data sharing module determines an identified patient sample based on the stroke medical record information of the current patient, wherein the identified patient is a patient sample whose medical record information similarity is greater than a preset similarity; Acquire a multimodal identified sample set of the identified patient samples; The multimodal sensing sample set is updated according to the multimodal identification sample set.

8. The intelligent assessment method for physical rehabilitation of stroke patients according to claim 1, wherein: After connecting the rehabilitation device, the method further includes: determining whether the rehabilitation device includes an auxiliary connection device; If the rehabilitation device includes an auxiliary connection device, obtaining an auxiliary sensor data set of the auxiliary connection device; An evaluation is performed according to the auxiliary sensor data set to obtain an auxiliary rehabilitation level indicator, and the rehabilitation level indicator is updated according to the auxiliary rehabilitation level indicator.

9. The intelligent assessment method for physical rehabilitation of stroke patients according to claim 1, wherein: After acquiring the multimodal sensing data set, uploading the multimodal sensing data set to a cloud processor, wherein the cloud processor includes a data processing template; The cloud processor performs data template processing and storage on the multimodal sensing data set according to the data processing template.

10. Intelligent physical rehabilitation assessment device for stroke patients, characterized by: The device is used to implement the intelligent assessment method for physical rehabilitation of stroke patients according to any one of claims 1 to 9, and the device comprises: The stroke medical record information acquisition module is used to connect to the medical information management system to obtain the current patient's stroke medical record information; an affected limb determination module, configured to identify and determine the affected limb based on the stroke medical record information; a rehabilitation motion sensing module, configured to connect to a rehabilitation device, perform rehabilitation motion sensing on the affected limb using a multimodal sensor, and output a multimodal sensing data set, wherein the multimodal sensing data set includes joint movement data, muscle tension data, limb movement data, and physiological sensing data; A multimodal sensor sample set acquisition module, used to acquire a multimodal sensor sample set of the affected limb corresponding to the healthy limb; a rehabilitation assessment model construction module, configured to construct a rehabilitation assessment model, wherein the rehabilitation assessment model evaluates the multimodal sensing data set according to the multimodal sensing sample set to obtain a rehabilitation level indicator; The training mode optimization module is used to feed back the rehabilitation level indicator to the rehabilitation device to optimize the training mode.

Images