Search

CN-121617539-B - Individualized training guidance system for cerebral apoplexy nerve rehabilitation

CN121617539BCN 121617539 BCN121617539 BCN 121617539BCN-121617539-B

Abstract

The invention relates to the field of intelligent medical treatment and rehabilitation engineering, in particular to a cerebral apoplexy nerve rehabilitation personalized training guidance system, which comprises a core framework construction step, a state analysis and strategy generation step, a closed loop feedback execution step and a control system, wherein the core framework construction step is used for establishing communication connection between a central processing platform and a data sensing interface, a personalized strategy engine and a feedback execution interface, the state analysis and strategy generation step is used for judging a nerve driving stage based on a multi-mode rehabilitation data set, generating a nerve driving enhancement strategy aiming at a low nerve driving stage and generating a motion learning optimization strategy aiming at a high nerve driving stage, the closed loop feedback execution step is used for converting the strategy into a multi-sense feedback control instruction and driving an external device to execute personalized induction on a patient.

Inventors

  • GUAN JIE
  • JIN YANCHUN

Assignees

  • 陕西省康复医院(陕西省残疾人康复中心)

Dates

Publication Date
20260508
Application Date
20260203

Claims (6)

  1. 1. The cerebral apoplexy nerve rehabilitation personalized training guidance system is characterized by comprising a central processing platform, wherein the central processing platform is in communication connection with a data sensing interface, a personalized strategy engine and a feedback execution interface; the data perception interface is used for acquiring a multi-mode rehabilitation data set reflecting the movement intention and the performance of a patient from external perception equipment; the personalized policy engine comprises: The state analysis unit is used for judging whether the current rehabilitation training stage of the patient is a low nerve driving stage or a high nerve driving stage according to a preset nerve-motion state model based on the multi-mode rehabilitation data set; a strategy generation unit for generating a neural drive enhancement strategy comprising stochastic resonance parameters in response to the low neural drive phase; The feedback execution interface is used for converting the neural drive enhancement strategy or the motion learning optimization strategy into a multi-sense feedback control instruction, and driving the external feedback equipment to execute personalized induction on a patient; the process of generating the neural drive enhancement strategy by the strategy generation unit comprises the following steps: Based on the characteristic data of the low nerve driving stage, the difficulty barrier parameter of the current rehabilitation task is calculated, and a calculation model is as follows: ; Wherein, the In units of Is a standard myoelectric amplitude reference value; in units of A current real-time electromyographic signal root mean square value; in units of Is a normalized coefficient of (a); with the aim of maximizing the cross-barrier activation probability, calculating the optimal stochastic resonance parameters in a preset parameter space by using an optimization algorithm; the stochastic resonance parameters are used to configure noise characteristics in the feedback signal to induce a synergistic response of the nervous system, the cross-barrier activation probabilities Defined as normalized probability within a single time window, the formula is as follows: ; Wherein, the Is a difficulty barrier parameter with non-negative constraint and dimensionless properties; normalized noise intensity to be optimized for dimensionless attribute; The process of calculating the optimal stochastic resonance parameters: Establishing a mapping model of positive correlation of the cross barrier activation probability and the difficulty barrier parameter and nonlinear function correlation of noise intensity; defining target activation probabilities for a system using saturation growth functions The model formula is as follows: ; Wherein, the A damping constant for controlling the saturation velocity; the system adjusts noise strength by performing the following discrete time iterative update law So that the actual activation probability approaches the target value: ; Wherein, the For the number of iteration steps, In order to adjust the step size of the step, Calculating probability for the current noise intensity; In the parameter optimization process, information entropy output by strategies is evaluated in real time, and the embodiment adopts the method based on activation probability The binary shannon entropy definition of (c) is as follows: ; embodying information entropy change rate as entropy convergence index reflecting system steady state approximation degree The calculation formula is as follows: ; Wherein, the The time differential absolute value of the information entropy is used for representing the change rate of unit time internal entropy and the unit is bit/s; A scale constant for adjusting sensitivity; When the index is And when the convergence threshold exceeds the preset convergence threshold, the system judges that the steady state is entered, and the current stochastic resonance parameter combination is locked.
  2. 2. The system of claim 1, wherein the process of determining the rehabilitation training phase by the state parsing unit comprises: Extracting signal intensity characteristics and motion trail characteristics from the multi-mode rehabilitation data set; Comparing the signal intensity characteristic with a preset nerve activation baseline range, and judging as a low nerve driving stage if the signal intensity characteristic is in the nerve activation baseline range and the motion track characteristic is in unordered fluctuation; and if the signal intensity characteristic is continuously higher than a preset activation threshold and the motion trail characteristic presents directional displacement, judging the high neural drive stage.
  3. 3. The system according to claim 1, wherein the process of generating the motion learning optimization strategy by the strategy generation unit comprises: Calculating a deviation vector of the motion trail features extracted from the multi-modal rehabilitation data set relative to a standard rehabilitation trail; judging whether the direction of the deviation vector deviates from the target direction of the standard rehabilitation track; if the direction is deviated, a first group of error amplification parameters are called to carry out nonlinear transformation on the deviation vector, and a reinforcement learning signal for feedback is generated; and if the direction tends, calling a second group of parameters to normalize the deviation vector.
  4. 4. A system according to claim 3, further comprising a training load monitoring module for performing the following operations: calculating a fatigue index reflecting physiological load of the patient based on the multi-modal rehabilitation data set; Comparing the fatigue index with a preset load threshold; if the fatigue index indicates that the load exceeds the load threshold, sending a regulation and control instruction to the strategy generation unit; the strategy generation unit responds to the regulation and control instruction, pauses the generation of the motion learning optimization strategy, and switches to the generation of a protective strategy which takes error reduction and assistance as guiding.
  5. 5. The system of claim 1, wherein the process of generating the multi-sensory feedback control instruction by the feedback execution interface comprises: Acquiring current strategy parameters; dynamic weights are distributed to the visual feedback channels and the tactile feedback channels based on a preset multi-channel fusion model; Mapping stochastic resonance parameters in the neural drive enhancement strategy to dynamic change frequency of visual feedback or intensity modulation frequency of tactile feedback; And mapping the error amplification parameters and the deviation vector in the motion learning optimization strategy into a space guiding vector of a virtual object pose or a haptic force field in the visual scene.
  6. 6. The system of claim 1, wherein the stochastic resonance parameters in the neural drive enhancement strategy are used to generate synthetic noise data with non-gaussian statistical properties and specific spectral attenuation properties to simulate an information carrier suitable for neural modulation.

Description

Individualized training guidance system for cerebral apoplexy nerve rehabilitation Technical Field The invention relates to the field of intelligent medical treatment and rehabilitation engineering, in particular to a cerebral apoplexy nerve rehabilitation personalized training guidance system. Background In the medical informatization application scene of cerebral apoplexy nerve rehabilitation, the rehabilitation guidance system is taken as a typical human-computer interaction medical information platform, and is characterized in that multi-dimensional health data of a patient are collected and converted into an executable rehabilitation training strategy by utilizing an algorithm model. The existing rehabilitation information processing system generally adopts signal discrimination logic based on a fixed threshold value and combines a central control unit to carry out closed-loop regulation and control on auxiliary equipment. The existing rehabilitation informatization technology has remarkable limitations in the data perception and decision generation stage. First, in terms of data processing algorithms for patient intent recognition, prior art schemes commonly employ trigger mechanisms based on linear magnitude comparison. In the face of severe paralysis patients, bioelectric signals generated by nerve driving are extremely weak and tend to be submerged in system background noise. The traditional filtering and judging algorithm is lack of a nonlinear enhancement mechanism, so that weak intention data with rehabilitation value are easy to be removed as invalid noise, a sensing dead zone appears in the system, subsequent training response cannot be triggered, and the coverage capability of a medical system on severe patients is seriously affected. The existing training strategy generation logic is single, and passive guidance is performed based on an error minimization principle. The information processing mode can not fully consider the requirement for motion error perception in the nerve remodeling process, particularly when a patient enters a high nerve driving stage, if the system continuously provides accurate assistance and lacks a differential error processing mechanism, medical monitoring indexes can enter a platform stage, and the patient is difficult to correct an incorrect muscle cooperative mode through system feedback, so that the pertinence of the information rehabilitation guidance is reduced. The existing system architecture has the defects in the aspects of multi-mode data fusion depth and physiological load real-time monitoring, and is difficult to automatically prevent secondary damage caused by fatigue through an informatization means while guaranteeing training intensity. Therefore, how to build a closed-loop rehabilitation information processing system which can accurately extract weak characteristic signals, has the hierarchical strategy regulation and control capability and gives consideration to physiological safety is a technical problem which needs to be solved in the field of intelligent medical treatment and rehabilitation engineering at present. Disclosure of Invention The invention aims to provide a cerebral apoplexy nerve rehabilitation personalized training guidance system, which aims to solve the problems in the background technology, and specifically, the technical scheme of the invention comprises the following steps: the central processing platform is in communication connection with a data perception interface, a personalized strategy engine and a feedback execution interface; the data perception interface is used for acquiring a multi-mode rehabilitation data set reflecting the movement intention and the performance of a patient from external perception equipment; the personalized policy engine comprises: The state analysis unit is used for judging whether the current rehabilitation training stage of the patient is a low nerve driving stage or a high nerve driving stage according to a preset nerve-motion state model based on the multi-mode rehabilitation data set; a strategy generation unit for generating a neural drive enhancement strategy comprising stochastic resonance parameters in response to the low neural drive phase; The feedback execution interface is used for converting the neural drive enhancement strategy or the motion learning optimization strategy into a multi-sense feedback control instruction and driving the external feedback equipment to execute personalized induction on the patient. Preferably, the process of judging the rehabilitation training phase by the state analysis unit includes: Extracting signal intensity characteristics and motion trail characteristics from the multi-mode rehabilitation data set; Comparing the signal intensity characteristic with a preset nerve activation baseline range, and judging as a low nerve driving stage if the signal intensity characteristic is in the nerve activation baseline range and the motion track characteris