CN-122025000-A - Rehabilitation system and method based on big data and AI drive

CN122025000ACN 122025000 ACN122025000 ACN 122025000ACN-122025000-A

Abstract

The invention discloses a rehabilitation system and a method based on big data and AI drive, which relate to the technical field of rehabilitation medical treatment, the invention collects multi-source heterogeneous data based on different service scenes, collects and accesses virtual data lakes for storage and treatment, further carries out data screening, feature extraction and quality control treatment through Spark tools to obtain multi-source heterogeneous data feature sets, builds a rehabilitation model through machine learning and reinforcement learning algorithms, the training mode, the difficulty parameters and the training scene are matched by combining individual differences of patients, further, the rehabilitation data of the patients are monitored in real time, the training immersion is promoted through an interaction technology, the abnormality in the rehabilitation data is identified, an intervention mechanism is triggered, a therapist dynamically adjusts a rehabilitation scheme to generate a rehabilitation evaluation report, the adaptation logic of the rehabilitation model and the immersion training scene is optimized by combining rehabilitation data shared by a cross mechanism, support is provided for intelligent development of industries, and the rehabilitation efficiency and the medical service quality are promoted by aid of assistance.

Inventors

CHEN LINGJIA
YU JINGWEN
Meng Feifan
GUAN TONG
XU YAO
Gu Shuangni

Assignees

南京交通职业技术学院

Dates

Publication Date: 20260512
Application Date: 20260113

Claims (8)

1. The rehabilitation method based on big data and AI drive is characterized by comprising the following steps: S1, acquiring multi-source heterogeneous data based on different service scenes, and summarizing the acquired multi-source heterogeneous data and accessing the collected multi-source heterogeneous data into a virtual data lake for unified storage and treatment; the multi-source heterogeneous data comprises structured data, unstructured data, time sequence data and text records; s2, based on the stored multi-source heterogeneous data, performing data screening, feature extraction and quality control processing through a Spark tool to obtain a standardized multi-source heterogeneous data feature set; S3, constructing a rehabilitation model based on a standardized multi-source heterogeneous data feature set through a machine learning and reinforcement learning algorithm, and intelligently matching a training mode, a difficulty parameter and a training scene by combining individual differences of patients; S4, monitoring patient rehabilitation data in real time, improving training immersion through an interactive technology, identifying abnormality in the rehabilitation data and triggering an intervention mechanism, and dynamically adjusting a rehabilitation scheme by a therapist according to the abnormality data in the patient rehabilitation data; S5, generating a data recovery evaluation report based on the full-flow patient recovery data and the recovery scheme adjustment record, and continuously optimizing the adaptation logic of the recovery model and the immersive training scene by combining the recovery data shared by the cross-institutions.
2. The rehabilitation method based on big data and AI driving according to claim 1, wherein the steps of collecting multi-source heterogeneous data based on different service scenes, and collecting the collected multi-source heterogeneous data to be connected into a virtual data lake for unified storage and management comprise the following steps: Constructing a multi-source heterogeneous data access channel based on data generated by a clinical evaluation system, various rehabilitation training devices and different business scenes recorded by a patient daily, and designing an adaptive acquisition interface to acquire multi-source heterogeneous data; setting a classification rule, accessing the collected multi-source heterogeneous data into a virtual data lake according to the classification rule, and carrying out classified storage and treatment through a virtual data lake storage architecture.
3. The rehabilitation method based on big data and AI driving according to claim 1, wherein the step of obtaining a standardized multi-source heterogeneous data feature set by performing data screening, feature extraction and quality control processing on the basis of the stored multi-source heterogeneous data through a Spark tool comprises the following steps: S21, reading the multi-source heterogeneous data stored in the virtual data lake through a Spark tool, carrying out format normalization on the multi-source heterogeneous data of different types, setting screening rules, screening according to the set screening rules, and reserving the multi-source heterogeneous data conforming to the rules; S22, extracting key features related to the rehabilitation state and training effect of the patient from the screened multi-source heterogeneous data by adopting a sliding window method and a Z-score normalization method to obtain multi-source heterogeneous data features; Z-score normalization is as follows: ; The sliding window method formula is as follows: ; Wherein the method comprises the steps of For the normalized multi-source heterogeneous data, A data value that is a feature in multi-source heterogeneous data, The average value and standard deviation of all multi-source heterogeneous data under the rehabilitation characteristic dimension are respectively, For rehabilitation profile statistics extracted within a single sliding window, For the time step of the sliding window, Is the first in the sliding window Key characteristic data values corresponding to the time steps; S23, performing quality control on the multi-source heterogeneous data characteristics, correcting the missing values and the abnormal values, checking the multi-source heterogeneous data characteristics, and integrating to obtain a standardized multi-source heterogeneous data characteristic set.
4. The rehabilitation method based on big data and AI driving according to claim 1, wherein the constructing a rehabilitation model based on the standardized multi-source heterogeneous data feature set through machine learning and reinforcement learning algorithm, and intelligently matching training modes, difficulty parameters and training scenes in combination with individual differences of patients comprises the following steps: Based on a standardized multi-source heterogeneous data feature set, extracting individual difference dimensions of a patient course and muscle strength level, constructing a rehabilitation model through a machine learning and reinforcement learning algorithm, and setting an evaluation dimension and decision logic of the rehabilitation model on a rehabilitation state; Analyzing the current rehabilitation foundation and the rehabilitation potential of a patient based on the constructed rehabilitation model, intelligently matching passive, boosting, active and anti-resistance training modes through a gradient lifting decision tree algorithm, setting switching conditions among the modes, and simultaneously setting corresponding training difficulty parameters by combining the tolerance of the patient and the rehabilitation target to form a personalized training scheme; and matching the adaptive immersive training scene according to the training mode, the difficulty parameter and the rehabilitation target which are output by the rehabilitation model.
5. The rehabilitation method based on big data and AI driving according to claim 4, wherein the method is characterized in that the constructed rehabilitation model is used for analyzing the current rehabilitation basis and rehabilitation potential of the patient, intelligently matching passive, boosting, active and anti-resistance training modes through a gradient boosting decision tree algorithm, setting the switching conditions among the modes, simultaneously combining the tolerance of the patient and the rehabilitation target, setting corresponding training difficulty parameters, and forming a personalized training scheme, and comprises the following steps: Based on the rehabilitation foundation, rehabilitation potential and individual difference data of the patient, extracting a disease course stage, a muscle strength grade, a tolerance threshold value and rehabilitation target data weight, and integrating to form a patient foundation data set; taking a gradient lifting decision tree algorithm as a training mode and difficulty parameter matching algorithm, constructing a decision tree model based on iteration of a patient basic data set, calculating fitness scores of different training modes and the patient basic data set, and screening out a training mode with highest fitness and an alternative training mode; The fitness calculation formula is as follows: ; Wherein the method comprises the steps of Is the first A comprehensive fitness score of the training patterns and the patient base data, Is the first The weight coefficient of the class input feature, Is the first Training mode of the seed Data of class input features; setting a switching threshold between the training mode and the alternative training mode based on the fitness screening result, and optimizing difficulty parameters of the selected training mode by combining the tolerance of the patient and the rehabilitation target; And integrating the determined core training mode, the alternative training mode, the training mode switching threshold value and the optimized difficulty parameter to form a personalized rehabilitation training scheme.
6. The rehabilitation method based on big data and AI driving according to claim 1, wherein the real-time monitoring of patient rehabilitation data promotes training immersion by interactive technology, identifies anomalies in the rehabilitation data and triggers an intervention mechanism, and the therapist dynamically adjusts the rehabilitation scheme according to anomalies in the patient rehabilitation data comprises the following steps: Constructing a training scene through an interaction technology to promote the immersion of patient training, and simultaneously carrying out full-time real-time monitoring on rehabilitation data in the patient training process through monitoring equipment to obtain real-time rehabilitation data of the patient; the interaction technology comprises visual guidance, force sense assistance and voice excitation; setting an abnormality judgment rule of rehabilitation data, dynamically analyzing and comparing real-time rehabilitation data of a patient based on the preset abnormality judgment rule, identifying abnormal fluctuation conditions in the data, and automatically triggering a corresponding intervention mechanism if an abnormal state occurs; And feeding back the rehabilitation data, the abnormal recognition result and the intervention execution condition which are monitored in real time to a therapist terminal, comprehensively evaluating by the therapist in combination with the information, and adjusting a training scheme according to the abnormal data in the rehabilitation data.
7. The rehabilitation method based on big data and AI driving according to claim 1, wherein the adapting logic for continuously optimizing the rehabilitation model and the immersive training scenario based on the full-flow patient rehabilitation data and the rehabilitation scheme adjustment record to generate a data-based rehabilitation evaluation report and combining the rehabilitation data shared across institutions comprises the following steps: Collecting the whole-flow patient rehabilitation data and rehabilitation scheme adjustment records, wherein the rehabilitation scheme adjustment records comprise real-time monitoring data, abnormal intervention records and training scheme adjustment details in the training process, and classifying, carding and association matching are carried out to form a rehabilitation evaluation data set; based on the rehabilitation evaluation data set, extracting a rehabilitation progress index, and generating a data rehabilitation evaluation report through data visualization and quantitative analysis; On the premise of meeting the privacy compliance requirement, accessing rehabilitation data shared by a cross-mechanism, performing fusion comparison with the local whole-flow rehabilitation data, optimizing rehabilitation model parameters based on the fusion data, and synchronously adjusting the adaptation logic of the immersive training scene and the training requirement.
8. A system for realizing the rehabilitation method based on big data and AI drive of claim 1, which is characterized by comprising a data acquisition and storage module, a data processing and feature extraction module, a rehabilitation training execution module and an evaluation and model optimization module; the data acquisition and storage module is used for constructing a data access channel based on a multi-service scene, designing an interface to acquire multi-source heterogeneous data, accessing the multi-source heterogeneous data into a virtual data lake, and classifying, storing and managing the multi-source heterogeneous data; The data processing and feature extraction module is used for reading data through Spark, processing the data through a sliding window method and a Z-score normalization method after regular screening, and obtaining a multi-source heterogeneous data feature set through verification and integration; the rehabilitation training execution module is used for extracting the disease course of a patient and the difference dimension of muscle strength levels to construct a rehabilitation model, constructing an immersive training scene, monitoring rehabilitation data in real time, identifying abnormality and triggering an intervention mechanism, and feeding back information to a therapist to adjust a training scheme; the evaluation and model optimization module is used for collecting rehabilitation data to form an evaluation set, generating a data report, and optimizing rehabilitation model parameters and adjusting scene adaptation logic by combining shared rehabilitation data.

Description

Rehabilitation system and method based on big data and AI drive Technical Field The invention relates to the technical field of rehabilitation medical treatment, in particular to a rehabilitation system and a rehabilitation method based on big data and AI driving. Background Cerebral apoplexy is the primary cause of disability of middle-aged and elderly people in China, about 50% -70% of survivors can leave upper limb movement dysfunction, the daily life capacity is seriously influenced, upper limb rehabilitation becomes the key requirement of clinical rehabilitation, and along with the promotion of 'health China' strategy, the requirements of rehabilitation medical treatment on 'accurate, intelligent and personalized' are increasingly urgent. In the existing rehabilitation mode, partial training equipment is introduced and data integration is tried, but the key bottleneck still exists that multisource heterogeneous rehabilitation data are scattered to form a 'data island', the scheme is adjusted with lag caused by difficult collaborative analysis, individual differences such as patient disease course, muscle strength grade and the like are ignored, the scheme is seriously homogenized, the training process is boring, the patient compliance is low, large data, AI and immersion technology are not fused sufficiently, and the technical value is not fully exerted, so that a set of full-flow and personalized intelligent rehabilitation solution is needed, and the industry is promoted to change from 'experience driving' to 'data driving'. Disclosure of Invention The invention aims to provide a rehabilitation system and a rehabilitation method based on big data and AI drive, which solve the problems in the background technology. In order to solve the technical problems, the invention adopts the following technical scheme that the invention provides a rehabilitation method based on big data and AI drive, which comprises the following steps: S1, acquiring multi-source heterogeneous data based on different service scenes, and summarizing the acquired multi-source heterogeneous data and accessing the collected multi-source heterogeneous data into a virtual data lake for unified storage and treatment; the multi-source heterogeneous data comprises structured data, unstructured data, time sequence data and text records; s2, based on the stored multi-source heterogeneous data, performing data screening, feature extraction and quality control processing through a Spark tool to obtain a standardized multi-source heterogeneous data feature set; S3, constructing a rehabilitation model based on a standardized multi-source heterogeneous data feature set through a machine learning and reinforcement learning algorithm, and intelligently matching a training mode, a difficulty parameter and a training scene by combining individual differences of patients; S4, monitoring patient rehabilitation data in real time, improving training immersion through an interactive technology, identifying abnormality in the rehabilitation data and triggering an intervention mechanism, and dynamically adjusting a rehabilitation scheme by a therapist according to the abnormality data in the patient rehabilitation data; S5, generating a data recovery evaluation report based on the full-flow patient recovery data and the recovery scheme adjustment record, and continuously optimizing the adaptation logic of the recovery model and the immersive training scene by combining the recovery data shared by the cross-institutions. Preferably, the collecting the multi-source heterogeneous data based on different service scenes, and collecting the collected multi-source heterogeneous data and accessing the collected multi-source heterogeneous data into a virtual data lake for unified storage and management comprises the following steps: Constructing a multi-source heterogeneous data access channel based on data generated by a clinical evaluation system, various rehabilitation training devices and different business scenes recorded by a patient daily, and designing an adaptive acquisition interface to acquire multi-source heterogeneous data; setting a classification rule, accessing the collected multi-source heterogeneous data into a virtual data lake according to the classification rule, and carrying out classified storage and treatment through a virtual data lake storage architecture. Preferably, the data screening, feature extraction and quality control processing are performed through a Spark tool based on the stored multi-source heterogeneous data, and the standardized multi-source heterogeneous data feature set is obtained, which comprises the following steps: S21, reading the multi-source heterogeneous data stored in the virtual data lake through a Spark tool, carrying out format normalization on the multi-source heterogeneous data of different types, setting screening rules, screening according to the set screening rules, and reserving the multi-source heterogen