Cerebral stroke rehabilitation personalized treatment recommendation system based on artificial intelligence

Abstract

The invention belongs to the field of rehabilitation medical treatment and intelligent computing, and particularly discloses a cerebral apoplexy rehabilitation personalized treatment recommendation system based on artificial intelligence. Comprising a multi-modal data acquisition module, a feature extraction and clustering modeling module, a personalized path making module and a recommended path generation and feedback optimization module, and can realize dynamic monitoring and intelligent analysis of rehabilitation data of a stroke patient; a multi-modal aggregation neural network model is constructed by fusing multi-dimensional data such as images, biomarkers, clinical behaviors and cognitive emotions, a rehabilitation path is formulated according to a medical data clustering result, and a recommended path is dynamically adjusted and optimized in combination with treatment feedback of a patient; compared with a traditional method, the method has the advantages that scientificity and adaptability of personalized treatment are improved, the method has the advantages of real-time updating of rehabilitation paths, closed-loop management of a system, high model robustness and the like, rehabilitation intervention efficiency and accuracy can be remarkably improved, and technical support is provided for intelligent development of rehabilitation medicine.

Description

Technical Field

[0001] The present invention belongs to the technical field of information-based healthcare systems, and particularly relates to a personalized treatment recommendation system for stroke rehabilitation based on artificial intelligence. Background Art

[0002] Stroke is one of the major chronic diseases with high mortality and disability rates globally, characterized by high incidence, severe sequelae, and long rehabilitation periods. Most stroke patients still face problems such as varying degrees of motor impairment, speech impairment, dysphagia, and cognitive function impairment after acute-phase treatment, seriously affecting their quality of life. Rehabilitation treatment, as a key link in the late intervention of stroke, runs through the entire disease recovery process and is of great significance for the functional recovery of patients.

[0003] Currently, clinical stroke rehabilitation treatment plans are mostly formulated by rehabilitation physicians based on their experience in judging patients' clinical manifestations, imaging examinations, and scale scores, which has certain subjectivity and uncertainty, and it is difficult to fully explore the deep relationship between patients' individual characteristics and rehabilitation potential. At the same time, due to the highly individualized differences in patients' rehabilitation needs, the traditional "template-based" intervention path is difficult to cover the diverse recovery processes. In addition, the following problems generally exist in the existing rehabilitation treatment process: First, the adjustment of the rehabilitation plan lags behind, and it is difficult to optimize the plan in a timely manner according to the dynamic changes of patients; second, the comprehensive utilization degree of objective information such as multi-dimensional physiological data and behavioral data is insufficient; third, there is a lack of a data-driven precise recommendation mechanism and treatment decision support system.

[0004] With the continuous development of artificial intelligence technology, especially the wide application of methods such as machine learning, deep learning, and knowledge graphs in the medical field, new ideas and tools have been provided for stroke rehabilitation. Artificial intelligence can mine potential patterns from a large amount of patients' rehabilitation data, identify key rehabilitation influencing factors, construct a precise evaluation model, and assist doctors in making treatment strategy recommendations. However, the existing artificial intelligence rehabilitation recommendation applications are still in the primary stage. Most systems only have an evaluation function, lack a deep feedback mechanism, the treatment recommendation lacks a credible reasoning path, and it is difficult to achieve seamless integration with the clinical workflow. Therefore, there is an urgent need to construct a new system that integrates multi-source data modeling, individual feature mining, and intelligent intervention recommendation to improve the scientificity, real-time performance, and individual adaptability of rehabilitation decisions, and realize the intelligent, refined, and personalized upgrade of stroke rehabilitation treatment. Summary of the Invention

[0005] Aiming at the above problems, the purpose of the present invention is to propose: A personalized treatment recommendation system for stroke rehabilitation based on artificial intelligence, including: A multi-modal data acquisition module, which is used to collect multi-modal medical data from the data sources of stroke patients, and perform synchronization, alignment, truncation, and normalization processing on the data to construct a standardized multi-modal fusion data set; A feature extraction and clustering modeling module, which is used to extract multiple groups of feature sets based on the standardized multi-modal fusion data set, perform high-dimensional feature clustering analysis, construct a multi-modal aggregation neural network model (MA-MNN), and verify the clustering effect; A personalized path formulation module, which is used to formulate a personalized rehabilitation treatment benchmark path that meets the actual rehabilitation needs of patients according to the feature clustering verification results; A recommended path generation and feedback optimization module, which is used to collect multi-modal monitoring data during the patient's rehabilitation process based on the personalized rehabilitation treatment benchmark path, dynamically generate a personalized rehabilitation treatment recommended path, and continuously optimize the recommended path according to real-time feedback.

[0006] Further, the feature extraction and clustering modeling module specifically includes: Construct a fusion feature extraction algorithm to extract multiple feature sets for multi-modal medical data; Adopt a multi-modal clustering algorithm to perform clustering analysis on the extracted high-dimensional feature data, define the feature center and the inter-class distance, and formulate a feature clustering determination criterion; Use the clustering determination criterion as the threshold criterion for neural network training to measure the effectiveness of the clustering results; Integrate multiple high-dimensional feature extraction models to construct a multi-modal aggregation neural network model (MA-MNN) for full-process feature extraction and modeling; Verify the convergence of the model clustering results through a decoupling algorithm. If it converges, the clustering is effective; if it does not converge, return for retraining.

[0007] Further, the system further includes: A data synchronization module, which is used to align the time axes of different data sources; A data standardization module, which is used to unify the formats, units, and scales of multi-modal data; The edge computing unit cooperates with the cloud server to support real-time data uploading, processing, and analysis; A user interaction interface module, which is used to display the patient's rehabilitation data, treatment recommendation paths, and feedback results.

[0008] The present invention also provides an application of an artificial intelligence-based personalized stroke rehabilitation treatment recommendation system, and the application includes the following steps: S1. Data acquisition and preprocessing: Collect multi-modal objective data from the data sources of stroke patients, and perform synchronization, alignment, truncation, and normalization processing on the data to construct a standardized multi-modal fusion data set; S2. Feature extraction and modeling: Based on the multi-modal fusion dataset, use the fusion feature extraction algorithm to cluster medical data, construct a multi-modal aggregation neural network model (MA-MNN), and use this model to verify the clustering effect; S3. Formulation of personalized rehabilitation treatment benchmark path: According to the clustering verification results, formulate a rehabilitation treatment benchmark path that conforms to the actual situation of the patient; S4. Generation of personalized rehabilitation treatment recommended path: Refer to the benchmark path, formulate rehabilitation treatment evaluation indicators; collect multi-modal monitoring data during the patient's rehabilitation treatment process, and compare and analyze it with the evaluation indicators to identify the qualified treatment path segments and form a personalized rehabilitation treatment recommended path; S5. Feedback and optimization of personalized rehabilitation treatment recommended path: According to the recommended path, formulate a specific treatment plan, continuously monitor the data performance during the patient's rehabilitation process, verify the adaptability of the recommended path, and dynamically adjust and integrate the path to achieve path optimization.

[0009] Furthermore, the step S2 specifically includes: S21. Construct a fusion feature extraction algorithm to extract multiple feature sets from multi-modal medical data; S22. Use the multi-modal clustering algorithm to perform clustering analysis on the extracted high-dimensional feature data, define the distance between the feature center and the class, and formulate a feature clustering determination criterion; S23. Measure the effectiveness of feature classification according to the above clustering distance and determination criterion; S24. Use the clustering determination criterion as the threshold criterion for neural network clustering analysis for model training and result classification; S25. Integrate multiple high-dimensional feature extraction models to construct a complete aggregation neural network model as a full-process feature extraction and modeling tool; S26. Verify the model clustering result through the decoupling algorithm. If it converges, the clustering is effective; if it does not converge, return to the previous steps to re-perform clustering training.

[0010] Furthermore, the multi-modal medical data includes one or more of the following: (1) Medical imaging data; at least including one of MRI, CT, and fMRI; (2) Biomarker data; at least including one of gene, protein, metabolomics, and blood sample information; (3) Clinical data; at least including one of medical history, physical examination, laboratory results, and medication records; (4) Behavioral data; at least including one of movement patterns, gait analysis, and rehabilitation records; (5) Cognitive and emotional data; including at least one of psychological assessment, emotional state, and cognitive function evaluation.

[0011] Further, the multi-modal aggregation neural network model (MA-MNN) consists of a backbone network and multiple branch networks. The backbone network includes an input layer, a feature aggregation layer, and an output layer, and the branch network contains an input layer, a feature extraction layer, and an output layer; All data is first input into the backbone network, distributed to each aggregation node via the feature aggregation layer, and passed to the corresponding branch network for processing.

[0012] Further, the downstream network of the MA-MNN model is a medical data clustering network, which performs clustering processing on multi-modal data, determines the key features for the next stage of treatment, and issues control instructions; after receiving the control instructions, the branch network selectively truncates the connection with non-target data to form a minimized topological structure under specific features.

[0013] Further, the medical data clustering network completes the feature clustering of multi-modal medical data based on a high-dimensional clustering model, and different types of data are effectively represented through weighted combination and feature fusion.

[0014] Further, the decoupling algorithm is a neural network fusion method based on a soft discrimination mechanism, which is used to improve the stability and accuracy of clustering determination.

[0015] Further, verifying the model clustering result through the decoupling algorithm includes: S261. Use the MA-MNN model to perform clustering analysis on the sample data to form multiple clustering center vectors; S262. Perform gradient convergence analysis on each type of clustering center vector. If all vectors converge, it indicates that the clustering is effective; S263. If there is a situation where non-convergence exists, it is determined that the clustering is invalid; S264. Perform secondary clustering on the invalid clustering to obtain a new set of center vectors; S265. If the secondary clustering still does not converge, it is fused with the original clustering center to form a new set of feature vectors, and continue the clustering analysis until the result converges. Repeat the above clustering verification steps.

[0016] Further, the fusion algorithm used in the clustering verification is a Gaussian mixture model based on the negative log-likelihood function, which is used to discriminate the class membership.

[0017] Further, after receiving the data output, classify the data based on the Gaussian mixture model, complete the filling of missing data through the sample backfilling method to form a corrected feature data set, and perform clustering analysis again by the downstream network to improve the robustness and recommendation accuracy of the model.

[0018] Furthermore, the missing data in the feature data set will be marked as error terms and distinguished from the complete feature data set for subsequent error repair and analytical modeling.

[0019] Beneficial Effects Compared with the prior art, the stroke rehabilitation personalized treatment recommendation system based on artificial intelligence provided by the present invention has the following beneficial effects: 1. Realized deep fusion and precise modeling of multimodal data: This system collects and processes medical data from multiple sources such as images, biomarkers, clinical behaviors, cognitive emotions, etc., and builds a unified high-dimensional feature data model by integrating feature extraction algorithms, which effectively solves the problems of one-sidedness and poor adaptability of a single data source, and significantly improves the comprehensiveness and accuracy of rehabilitation assessment.

[0020] 2. Improved the personalization and adaptability of rehabilitation pathway recommendations: By constructing a multimodal aggregation neural network model and combining it with a dynamic clustering and feedback mechanism, it is possible to continuously adjust and optimize personalized treatment pathways based on the actual rehabilitation progress of different patients, breaking through the drawbacks of the "fixed" traditional rehabilitation pathway and significantly improving the scientificity and pertinence of rehabilitation interventions.

[0021] 3. An intelligent feedback and adaptive optimization mechanism has been established: The present invention has set up a closed-loop management mechanism from path recommendation to treatment feedback to path correction. Combined with the medical data clustering network and decoupling algorithm, the effectiveness of the recommended path is evaluated in real time during the operation of the system, and dynamic optimization is performed accordingly, making the rehabilitation path more controllable and iterative.

[0022] 4. Enhanced the robustness and scalability of the model in complex medical environments: Through a multi-layer aggregation structure, feature selection mechanism and backfill reconstruction module, this system effectively responds to problems such as missing data, category overlap and feature drift, improves the stability of the model in multiple scenarios and across populations, and has good potential for practical promotion and application.

[0023] In summary, the present invention integrates artificial intelligence technology and medical data analysis capabilities to construct a multi-dimensional, feedback-enabled, and highly adaptable stroke rehabilitation recommendation system, which can significantly improve the efficiency and efficacy of rehabilitation treatment and provide a new path for precision rehabilitation medicine. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Figure 1 A schematic diagram of the steps of the application method of the stroke rehabilitation personalized treatment recommendation system based on artificial intelligence of the present invention; Figure 2 This is a schematic diagram of the sub-steps of step S2 of the application method of the artificial intelligence-based stroke rehabilitation personalized treatment recommendation system of the present invention. DETAILED DESCRIPTION

[0025] In order to deepen the understanding of the present invention, the present invention will be further described in detail below in conjunction with examples. The examples are only used to explain the present invention and do not constitute a limitation on the protection scope of the present invention.

[0026] Example 1 according to Figure 1 As shown, this embodiment provides an artificial intelligence-based stroke rehabilitation personalized treatment recommendation system, including: A multimodal data acquisition module is used to collect multimodal medical data from the data source of stroke patients, and synchronize, align, intercept and normalize the data to build a standardized multimodal fusion data set; The feature extraction and clustering modeling module is used to extract multiple sets of features based on standardized multimodal fusion data sets, perform high-dimensional feature clustering analysis, build a multimodal aggregation neural network model (MA-MNN), and verify the clustering effect; The personalized pathway formulation module is used to formulate a personalized rehabilitation treatment benchmark pathway that meets the actual rehabilitation needs of patients based on the feature clustering verification results; The recommended pathway generation and feedback optimization module is used to collect multimodal monitoring data during the patient's rehabilitation process based on the personalized rehabilitation treatment benchmark pathway, dynamically generate personalized rehabilitation treatment recommended pathways, and continuously optimize the recommended pathways based on real-time feedback.

[0027] The application of the stroke rehabilitation personalized treatment recommendation system based on artificial intelligence includes the following steps: S1. Data collection and preprocessing: Collect multimodal objective data from the data source of stroke patients, and perform data synchronization, data alignment, data interception and data normalization on the multimodal objective data to construct a multimodal fusion data set; S2. Feature extraction and modeling: Based on the multimodal fusion data set, a fusion feature extraction algorithm is used to cluster medical data, a multimodal aggregation neural network model MA-MNN is established, and the multimodal aggregation neural network model MA-MNN is used to verify medical data clustering; S3. Development of personalized rehabilitation treatment benchmark pathway: Based on the verified medical data clustering results, develop personalized rehabilitation treatment benchmark pathway; S4. Generation of personalized rehabilitation treatment recommendation pathways: Based on the personalized rehabilitation treatment benchmark pathway, formulate rehabilitation treatment evaluation indicators; monitor the rehabilitation treatment progress of stroke patients and generate multimodal objective monitoring data of the rehabilitation treatment pathway; compare and analyze the multimodal objective monitoring data with the corresponding rehabilitation treatment evaluation indicators to determine the treatment pathway segments that meet the rehabilitation treatment evaluation indicators as the personalized rehabilitation treatment recommendation pathways; S5. Feedback and optimization of personalized rehabilitation treatment recommendation pathways: Formulate treatment plans based on personalized rehabilitation treatment recommendation pathways, monitor the patient's rehabilitation treatment progress, verify the personalized rehabilitation treatment recommendation pathways, and dynamically adjust and integrate the personalized rehabilitation treatment recommendation pathways.

[0028] like Figure 2 As shown, the step S2 specifically includes: S21. Construct a fusion feature extraction algorithm for multimodal medical data of stroke. The fusion feature extraction algorithm is used for high-dimensional feature data. Constructing feature dataset ; in, is the sequence number of the high-dimensional feature data, is the total number of high-order feature data, For the The high-dimensional feature center of high-dimensional feature data, For the The high-dimensional feature number of high-dimensional feature data satisfies:

[0029] in, is the feature mode number, The serial number is The number of characteristic patterns of feature patterns, and the total number of features of all feature patterns is ; S22, the high-dimensional feature data Perform cluster analysis and construct high-dimensional clustering functions and distance measurement functions to meet the following requirements:

[0030]

[0031] in, Indicates All high-dimensional feature data High-dimensional clustering function with high-dimensional features; Indicates The first Class and The distance measurement function of the high-dimensional feature center of the class feature; and Respectively The first Class and High-dimensional feature center of class features; Indicates the serial number is The characteristic pattern of Features For the The geometric center of the characteristic mode; Indicates the serial number is The total number of features of the feature pattern; S23. Based on distance metric function Obtain feature clustering decision function ,satisfy:

[0032] in, represents the mean function, represents the variance function.

[0033] S24, clustering the feature determination function As the characteristic distance threshold function of the clustering algorithm, a neural network based on the high-dimensional feature data is completed. Construction and cluster analysis; S25. Co-construction of multimodal medical data for all strokes Feature datasets with high-dimensional feature data , a total of A feature clustering decision function based on different high-dimensional feature data is constructed, and the feature clustering decision functions of all high-dimensional feature data are used to construct an aggregate neural network as a feature extraction algorithm for multimodal medical data of stroke; S26, verification: use the decoupling algorithm to perform clustering verification on the high-dimensional clustering function. If the clustering result converges, the multimodal aggregation neural network model MA-MNN is determined to be a high-dimensional clustering model that meets the requirements; if the clustering result does not converge, return to step S22 and perform clustering again.

[0034] Example 2 This embodiment is further limited on the basis of the embodiment 1: the multimodal medical data includes any one or more of the following data: (1) Imaging data: including magnetic resonance imaging (MRI), computed tomography (CT) and functional MRI (fMRI); (2) Biomarker data: including genomics, proteomics, metabolomics and blood biomarkers; (3) Clinical data: including the patient's medical history, physical examination results, laboratory test results, and medication records.

[0035] (4) Behavioral data: patient's movement patterns, gait analysis, rehabilitation training records, etc.; (5) Cognitive and affective data: including assessments of cognitive abilities, mental health, and emotional states.

[0036] The multimodal aggregate neural network model MA-MNN has a backbone network and A branch network, the backbone network has an input layer, a feature aggregation layer and an output layer, Each branch network has an input layer, a feature extraction layer and an output layer; the multimodal medical data of stroke is input into the input layer of the backbone network; There are aggregation nodes, and each aggregation node corresponds to a branch network.

[0037] The downstream network of the multimodal aggregation neural network model MA-MNN is a medical data clustering network, which performs cluster analysis on the multimodal medical data of stroke, determines the high-dimensional features of the next treatment stage, and sends extraction instructions to the upstream network; the extraction instructions of the self-aggregation nodes are controlled for different high-dimensional features, and the feature extraction layer cuts off the connection with the unextracted medical data to construct a minimum topology that adapts to the high-dimensional features.

[0038] Example 3 This embodiment is further limited on the basis of the embodiment 1: the medical data clustering network performs feature clustering on the multimodal medical data of stroke based on a high-dimensional clustering model, and the high-dimensional clustering model satisfies:

[0039] in, is the output function of the high-dimensional clustering model; is the weight coefficient.

[0040] The decoupling algorithm is a neural network fusion method based on a soft discrimination mechanism, which is used to improve the stability and accuracy of clustering determination.

[0041] Verification of the model clustering results through the decoupling algorithm includes: S261, using the MA-MNN model to perform cluster analysis on the sample data to form multiple cluster center vectors; S262, performing gradient convergence analysis on each type of cluster center vector. If all vectors converge, it means that the clustering is effective; S263, if there is no convergence, the clustering is determined to be invalid; S264, performing secondary clustering on the invalid clusters to obtain a new center vector set; S265. If the secondary clustering still has not converged, it is merged with the original cluster center to form a new set of feature vectors, and the clustering analysis is continued until the result converges, and the above clustering verification steps are repeated.

[0042] The fusion algorithm used in the cluster verification is a Gaussian mixture model based on a negative log-likelihood function, which is used to discriminate category affiliation.

[0043] After receiving the data output, the data is classified based on the Gaussian mixture model, and the missing data is supplemented by sample backfilling to form a corrected feature data set. The downstream network then performs cluster analysis again to improve the robustness of the model and the accuracy of recommendations.

[0044] The above shows and describes the basic principles, main features and advantages of the present invention. It should be understood by those skilled in the art that the present invention is not limited to the above embodiments. The above embodiments and descriptions are only for explaining the principles of the present invention. Without departing from the spirit and scope of the present invention, the present invention may have various changes and improvements, which fall within the scope of the present invention. The scope of protection of the present invention is defined by the attached claims and their equivalents.

Claims

1. An AI-based personalized stroke rehabilitation treatment recommendation system, characterized by: include: A multimodal data acquisition module is used to collect multimodal medical data from the data source of stroke patients, and synchronize, align, intercept and normalize the data to build a standardized multimodal fusion data set; The feature extraction and clustering modeling module is used to extract multiple sets of features based on standardized multimodal fusion data sets, perform high-dimensional feature clustering analysis, build a multimodal aggregation neural network model, and verify the clustering effect; The personalized pathway formulation module is used to formulate a personalized rehabilitation treatment benchmark pathway that meets the actual rehabilitation needs of patients based on the feature clustering verification results; The recommended pathway generation and feedback optimization module is used to collect multimodal monitoring data during the patient's rehabilitation process based on the personalized rehabilitation treatment benchmark pathway, dynamically generate personalized rehabilitation treatment recommended pathways, and continuously optimize the recommended pathways based on real-time feedback.

2. The stroke rehabilitation personalized treatment recommendation system based on artificial intelligence according to claim 1 is characterized by: The feature extraction and clustering modeling module specifically includes: Construct a fusion feature extraction algorithm to extract multiple feature sets for multimodal medical data; A multimodal clustering algorithm is used to perform cluster analysis on the extracted high-dimensional feature data, define the feature center and the distance between clusters, and formulate the feature clustering judgment criteria; The clustering judgment standard is used as the threshold standard for neural network training to measure the effectiveness of clustering results; Integrate multiple high-dimensional feature extraction models and build a multimodal aggregation neural network model for feature extraction and modeling throughout the entire process; The convergence of the model clustering results is verified by the decoupling algorithm. If converged, the clustering is effective. If not, it returns to retraining.

3. The stroke rehabilitation personalized treatment recommendation system based on artificial intelligence according to claim 2 is characterized in that: The system further comprises: Data synchronization module, used to align the time axes of different data sources; Data standardization module, used to unify the format, unit and scale of multimodal data; The edge computing unit works with the cloud server to support real-time data uploading, processing and analysis; User interaction interface module, used to display patient rehabilitation data, treatment recommendation path and feedback results.

4. The application of the stroke rehabilitation personalized treatment recommendation system based on artificial intelligence according to any one of claims 1 to 3, characterized in that: The application comprises the following steps: S1. Data collection and preprocessing: Collect multimodal objective data from the data sources of stroke patients, synchronize, align, intercept and normalize the data, and construct a standardized multimodal fusion data set; S2. Feature extraction and modeling: Based on the multimodal fusion data set, the fusion feature extraction algorithm is used to cluster the medical data, a multimodal aggregation neural network model is constructed, and the clustering effect is verified using the model; S3. Development of personalized rehabilitation treatment benchmark pathway: Based on the clustering verification results, develop a rehabilitation treatment benchmark pathway that meets the actual situation of the patient; S4. Generation of personalized rehabilitation treatment recommendation pathways: refer to the benchmark pathway and formulate rehabilitation treatment evaluation indicators; collect multimodal monitoring data during the patient's rehabilitation treatment process, and compare and analyze it with the evaluation indicators to identify the treatment pathway segments that meet the standards and form personalized rehabilitation treatment recommendation pathways; S5. Feedback and optimization of recommended pathways for personalized rehabilitation treatment: Develop specific treatment plans based on the recommended pathways, continuously monitor the patient's data performance during the rehabilitation process, verify the adaptability of the recommended pathways, and dynamically adjust and integrate the pathways to achieve pathway optimization.

5. The application of the stroke rehabilitation personalized treatment recommendation system based on artificial intelligence according to claim 4 is characterized by: The step S2 specifically includes: S21. Construct a fusion feature extraction algorithm to extract multiple feature sets from multimodal medical data; S22, using a multimodal clustering algorithm to perform cluster analysis on the extracted high-dimensional feature data, defining the distance between feature centers and classes, and formulating feature clustering judgment criteria; S23, measuring the effectiveness of feature classification according to the above clustering distance and judgment criteria; S24, using the clustering judgment standard as a threshold standard for neural network clustering analysis for model training and result classification; S25. Integrate multiple high-dimensional feature extraction models to build a complete aggregate neural network model as a full-process feature extraction and modeling tool; S26. Verify the clustering results of the model through the decoupling algorithm. If convergence occurs, the clustering is effective. If not, return to the previous steps and perform clustering training again.

6. The application of the stroke rehabilitation personalized treatment recommendation system based on artificial intelligence according to claim 5 is characterized by: The multimodal medical data includes one or more of the following: Medical imaging data; including at least one of MRI, CT, and fMRI; Biomarker data; including at least one of gene, protein, metabolomics and blood sample information; Clinical data: at least one of medical history, physical examination, laboratory results, and medication records; Behavioral data; including at least one of movement pattern, gait analysis, and rehabilitation record; Cognitive and affective data; including at least one of psychological assessment, emotional state, and cognitive function assessment.

7. The application of the stroke rehabilitation personalized treatment recommendation system based on artificial intelligence according to claim 6 is characterized by: The multimodal aggregation neural network model (MA-MNN) consists of a backbone network and multiple branch networks. The backbone network includes an input layer, a feature aggregation layer and an output layer, and the branch network includes an input layer, a feature extraction layer and an output layer. All data are first input into the backbone network, distributed to each aggregation node through the feature aggregation layer and passed to the corresponding branch network for processing.

8. The application of the stroke rehabilitation personalized treatment recommendation system based on artificial intelligence according to claim 7 is characterized by: The downlink network of the multimodal aggregation neural network model is a medical data clustering network, which performs clustering processing of multimodal data to determine the key features of the next stage of treatment and issue control instructions; after receiving the control instructions, the branch network selectively cuts off the connection with non-target data to form a minimized topological structure under specific features.

9. The application of the stroke rehabilitation personalized treatment recommendation system based on artificial intelligence according to claim 8 is characterized by: The medical data clustering network completes the feature clustering of multimodal medical data based on a high-dimensional clustering model, and different types of data are effectively represented through weighted combination and feature fusion; the decoupling algorithm is a neural network fusion method based on a soft discrimination mechanism, which is used to improve the stability and accuracy of clustering judgment.

10. The application of the stroke rehabilitation personalized treatment recommendation system based on artificial intelligence according to claim 9 is characterized by: Verification of the model clustering results through the decoupling algorithm includes: S261. Perform cluster analysis on the sample data using a multimodal aggregation neural network model to form multiple cluster center vectors; S262, performing gradient convergence analysis on each type of cluster center vector. If all vectors converge, it means that the clustering is effective; S263, if there is no convergence, the clustering is determined to be invalid; S264, performing secondary clustering on the invalid clusters to obtain a new center vector set; S265. If the secondary clustering still has not converged, it is merged with the original cluster center to form a new set of feature vectors, and the clustering analysis is continued until the result converges, and the above clustering verification steps are repeated.

Images