Training method for preventing cerebral apoplexy patient from falling down

Abstract

The invention belongs to the field of rehabilitation training, and particularly relates to a training method for preventing a cerebral apoplexy patient from falling down, which comprises the steps of dynamic risk assessment, graded training starting, double-task execution, self-adaptive adjustment and environment simulation training. The bottleneck of balance function training is broken through, and the problem of cognition-motion separation is solved; three-level cognitive interference quantitative control is adopted (vision: the light spot moving frequency is 0.1-3 Hz; auditory sense: N-back task difficulty is self-adaptive at 1-4 levels; tactile sense: sole asymmetric vibration), double-task load ratio dynamic adjustment, and radically cure insufficient individual adaptation; and the four-level risk grading model is matched with an adaptive adjustment mechanism. Through a multi-modal closed-loop training system, the problems of subjective evaluation, single training, protection lag and the like in the field of cerebral apoplexy falling prevention are solved, the falling risk reduction effect is good, and remarkable technical breakthrough and market irreplaceability are achieved.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of rehabilitation training, and particularly relates to a training method for preventing falls of stroke patients. BACKGROUND

[0002] During the recovery period of stroke patients, falls are prone to occur, and the current training and evaluation system has the following defects:

[0003] 1. One-sided evaluation system: static indicators cannot reflect dynamic risks

[0004] Quantitative dimension is missing: the current clinical mainstream uses subjective scales such as Berg Balance Scale (BBS) and Timed Up and Go Test (TUG), which can only evaluate static balance function. For example, research shows that the moving speed of the center of pressure (COP) in dynamic walking is highly correlated with the risk of falling, as high as 0.87, while existing technologies generally do not collect COP trajectory parameters.

[0005] Cognitive load is ignored: 43% of falls of stroke patients occur in dual-task scenarios (such as walking while talking), but the current virtual reality scheme generally only records motor performance and does not integrate the key indicator of cognitive error rate (DT_error), resulting in a high risk of misjudgment.

[0006] 2. Training mode is mechanized: unable to adapt to the needs of neural plasticity

[0007] Motor-cognitive separation: traditional training separates balance training (such as standing on a balance pad) from cognitive training (such as calculation exercises), but neuroimaging studies have confirmed that cerebellum-frontal lobe joint activation is the basis for improving functional walking. Existing devices generally lack dual-task synchronization control mechanisms, resulting in low efficiency of neural function remodeling.

[0008] Parameter rigidity: although commercial intelligent treadmills can adjust speed and slope, key variables such as platform stability coefficient and cognitive interference timing are fixed and cannot be changed. For example, in the training of patients with grade III risk, the platform inclination is uniformly set to 5°, without considering the real-time change of the patient's electromyographic asymmetry index, resulting in insufficient or excessive training intensity.

[0009] 3. Environment simulation distortion: deviates from real-life scenarios

[0010] Single type of interference: existing virtual reality schemes generally only provide visual scene simulation, but tactile interference (such as uneven ground) is rare in real-life fall events.

[0011] Response delay is fatal: the mechanical response time of sudden obstacle simulation devices is > 500ms, while the posture adjustment response window of stroke patients is only 200-400ms. This delay makes the protection mechanism useless, directly leading to a high secondary injury rate in clinical trials.

[0012] 4. Individualized adaptation deficiency: neglecting neural functional heterogeneity

[0013] Dynamic adjustment deficiency: although devices can detect real-time fall risk, the adjustment is delayed for 5-8 minutes (manual adjustment by therapists is required). During this period, the patient may experience multiple fall critical events.

[0014] The existing technology has four core defects of incomplete evaluation dimension, insufficient neural plasticity stimulation, distorted environment simulation, and real-time regulation deficiency, resulting in the industry dilemma of "high injury risk and low rehabilitation benefit" in stroke fall prevention training.

[0015] And the present application fundamentally solves these technical bottlenecks through a multi-modal dynamic coupling mechanism. SUMMARY

[0016] The present application provides a multi-modal system that integrates biomechanical regulation, neural plasticity training, and environmental adaptation simulation, by:

[0017] Establishing a four-level fall risk quantitative evaluation model;

[0018] Developing a cognitive-motor coupling training protocol;

[0019] Building an adaptive resistance adjustment mechanism;

[0020] Integrating a real-time biofeedback system.

[0021] A training method for preventing stroke patients from falling, comprising the following steps:

[0022] Step 1, dynamic risk assessment

[0023] Collecting gait parameters of patients through inertial sensors (IMU) and calculating fall risk index FRI:

[0024] FRI = 0.6 × COP velocity + 0.25 × EMG asym + 0.15 × DT error

[0025] Wherein, COP velocity is the center of pressure movement speed, EMG is the bilateral electromyographic asymmetry index, and DT error is the dual-task error rate;

[0026] Step 2, graded training initiation

[0027] According to the output value of the fall risk index FRI, the patients are divided into four risk levels:

[0028] Grade Ⅰ: FRI <0.3; Grade Ⅱ: 0.3-0.5; Grade Ⅲ: 0.5-0.7; Grade Ⅳ: >0.7; Matching corresponding intensity of cognitive-motor dual task protocol;

[0029] Step 3, dual task execution

[0030] Primary task: maintain dynamic balance on an adjustable balance platform, the tilt angle of the adjustable balance platform is adjusted in the range of 0°-15° with a step of 0.5°;

[0031] Secondary task: synchronously execute at least one of visual interference task, auditory N-back task or tactile interference task;

[0032] Step 4, adaptive adjustment

[0033] When the success rate of continuous 3 actions is >85%, increase the difficulty of the cognitive task by 1 level and increase the platform instability index by 0.1;

[0034] When the fall warning signal is detected, automatically activate the safety suspension system and reduce the training resistance by 30%;

[0035] Step 5, environmental simulation training

[0036] Respond to sudden obstacles in a virtual reality scene, the time interval of the obstacles is a random value in the range of 500-3000ms.

[0037] Further improvement, the visual interference task in step 3 is to track a randomly moving light spot, the moving frequency of the light spot is adjustable in the range of 0.1-3Hz, and the moving track includes horizontal, vertical and diagonal directions.

[0038] Further improvement, in the tactile interference task in step 3, an asymmetric vibration stimulus is used, a vibration wave with a phase difference of 0°-180° is applied through a pressure sensing array arranged on the patient's foot, and the vibration frequency is 2-5Hz.

[0039] Further improvement, the sudden obstacle response in step 5 includes three modes:

[0040] Mode A: avoid moving shopping carts in a virtual supermarket scene, the cart speed is dynamically adjusted in the range of 0.2-1.2m / s;

[0041] Mode B: walk irregular stairs, the step height difference changes randomly in the range of 5-25cm;

[0042] Mode C: ejectable physical obstacles, ejected by a gas pressure driven device on the training path.

[0043] A training system for implementing the above training method, comprising:

[0044] Dynamic evaluation module: contains IMU sensor placed on the waist of the patient and foot pressure distribution detection plate;

[0045] Multi-modal training platform: composed of anti-interference balance table, intelligent weight reduction treadmill, visual interference generator, wherein the base support area of the balance table can be electrically adjusted within 100-600cm 2

[0046] Central controller: real-time processing of EMG, COP, EEG signals, built-in fall risk prediction model based on LSTM;

[0047] Safety protection unit: contains electromagnetic damping device (damping coefficient 5-200N·s / m adjustable) and trunk safety suspension system.

[0048] Further improvement, the anti-interference balance table is integrated with sudden load transfer device, which applies 0.5-3kg lateral load transfer within <100ms, for triggering expected posture adjustment training.

[0049] Further improvement, the visual interference generator contains laser projection device and programmable mirror group, generating dynamic light spot with diameter of 2-10cm, position positioning error ≤1cm.

[0050] Further improvement, the central controller follows the following rules when performing parameter adjustment:

[0051] If the real-time EMG asymmetry index > preset threshold value, then:

[0052] Reduce the training resistance of the affected side by 15%, and simultaneously increase the difficulty of the cognitive task of the healthy side;

[0053] If the FRI value decreases by <5% for 2 consecutive minutes, then:

[0054] Start mode C in environment simulation training.

[0055] Compared with the prior art, the present application has the following beneficial effects:

[0056] The present application breaks through the bottleneck of balance function training and solves the problem of cognitive-motor separation; three-level cognitive interference quantization control (visual: light spot moving frequency 0.1-3Hz; auditory: N-back task difficulty 1-4 level adaptive; tactile: asymmetric vibration on the soles of the feet) is adopted, the dual-task load ratio is dynamically adjusted, and the individual adaptation deficiency is radically solved; a four-level risk grading model matching adaptive adjustment mechanism is adopted. The present application solves the problems of subjective evaluation, single training and lagging protection in the field of stroke fall prevention through a multi-modal closed-loop training system, and has good effect in reducing fall risk, and has significant technical breakthrough and market irreplaceability. DETAILED DESCRIPTION ​

[0057] The application will be further described in detail below through specific examples.

[0058] Example 1

[0059] A training method for preventing falls of stroke patients, comprising the following steps:

[0060] Step 1, dynamic risk assessment

[0061] Collect gait parameters of the patient through an inertial sensor (IMU), and calculate a fall risk index FRI:

[0062] FRI = 0.6 x COP + 0.25 x EMG + 0.15 x DT velocity asym error

[0063] Wherein, COP is the center of pressure movement speed, EMG is a bilateral electromyographic asymmetry index, and DT is a dual-task error rate. velocity error

[0064] Step 2, graded training initiation

[0065] According to the FRI output value, the patient is divided into four risk levels:

[0066] Level I: FRI < 0.3; Level II: 0.3-0.5; Level III: 0.5-0.7; and Level IV: > 0.7; and a cognitive-motor dual-task protocol with a corresponding intensity is matched.

[0067] Step 3, dual-task execution

[0068] Primary task: maintaining dynamic balance on an adjustable balance platform, the tilt angle of the adjustable balance platform is adjusted in a range of 0°-15° at a step of 0.5°.

[0069] Secondary task: synchronously performing at least one of a visual interference task, an auditory N-back task, or a tactile interference task.

[0070] The visual interference task is to track a randomly moving light spot, the moving frequency of the light spot is adjustable in a range of 0.1-3 Hz, and the moving track includes horizontal, vertical, and diagonal directions.

[0071] The tactile interference task adopts asymmetric vibration stimulation, and a pressure sensing array arranged on the sole of the patient is used to apply vibration waves with a phase difference of 0°-180°, and the vibration frequency is 2-5 Hz.

[0072] Step 4, adaptive adjustment

[0073] ​​​​When the success rate of 3 consecutive actions is > 85%, increase the cognitive task difficulty by 1 level and increase the platform instability index by 0.1;

[0074] When a fall warning signal is detected, automatically activate the safety suspension system and reduce the training resistance by 30%;

[0075] Step 5, environmental simulation training

[0076] Respond to sudden obstacles in the virtual reality scene, with an obstacle appearance interval of 500-3000ms random value.

[0077] The sudden obstacle response includes three modes:

[0078] Mode A: Avoid moving shopping carts in a virtual supermarket, with a dynamic adjustment of cart speed between 0.2-1.2m / s;

[0079] Mode B: Irregular stair walking, with a random step height difference of 5-25cm;

[0080] Mode C: Projectile physical obstacles, ejected by a pneumatic drive device on the training path.

[0081] Example 2

[0082] A training system, comprising:

[0083] Dynamic assessment module: containing IMU sensors placed on the patient's waist and plantar pressure distribution detection plates;

[0084] Multi-modal training platform: composed of anti-interference balance table, intelligent weight-reducing treadmill, and visual disturbance generator, wherein the base support area of the balance table can be electrically adjusted within 100-600cm 2

[0085] The anti-interference balance table is integrated with a sudden load transfer device, which applies a 0.5-3kg lateral load transfer within <100ms, used to trigger anticipatory posture adjustment training.

[0086] The visual disturbance generator includes a laser projection device and a programmable mirror group, generating dynamic light spots with a diameter of 2-10cm, with a position positioning error ≤1cm.

[0087] Central controller: real-time processing of EMG, COP, EEG signals, with a built-in LSTM-based fall risk prediction model;

[0088] The central controller follows the following rules when adjusting parameters:

[0089] If the real-time EMG asymmetry index > preset threshold:

[0090] ​Reduce the training resistance of the affected side by 15% and simultaneously increase the difficulty of the cognitive task on the unaffected side.

[0091] If the FRI value decreases by <5% for 2 consecutive minutes, then:

[0092] Start mode C in the environmental simulation training.

[0093] Safety protection unit: contains electromagnetic damping device (damping coefficient 5-200 N·s / m adjustable) and trunk safety suspension system.

[0094] Example 3

[0095] Training protocol for patients in the intermediate rehabilitation stage

[0096] Subject: right hemiplegic patient (Berg score 32)

[0097] Training cycle:

[0098] Week 1: dual task load ratio 70:30;

[0099] Platform stability coefficient 0.8;

[0100] Week 3: load ratio adjusted to 50:50;

[0101] Introduce sudden obstacle response (800ms warning);

[0102] Effect: after 6 weeks, the functional gait assessment (FGA) was improved by 47%.

[0103] Example 4

[0104] Technical effect verification data

[0105] After 4 weeks of training for 30 patients:

[0106] Timed up and go test (TUG) improvement rate 39.7±6.2%;

[0107] Fall efficacy scale (FES) score improved by 58.4%;

[0108] Dual task walking speed decreased from 28.3% to 9.1%.

[0109] The above are only preferred embodiments of the present application and are not intended to limit the present application. For those skilled in the art, the present application can have various modifications and changes. Any modification, equivalent replacement, improvement, etc. within the spirit and principles of the present application shall be included in the protection scope of the present application.

Claims

1. A training method for preventing stroke patients from falling, characterized by: The following steps are involved: Step 1: Dynamic risk assessment The patient's gait parameters are collected through an inertial sensor (IMU) to calculate the fall risk index FRI: FRI=0.6×COP velocity +0.25×EMG asym +0.15×DT error Among them, COP velocity is the speed of the pressure center movement, is the bilateral electromyography asymmetry index, DT error is the dual-task error rate; Step 2: Start the graded training Patients are divided into four risk levels according to the fall risk index FRI output value: Level I: FRI < 0.3; Level II: 0.3-0.5; Level III: 0.5-0.7; Level IV: > 0.7; cognitive-motor dual-task protocol matching corresponding intensity; Step 3: Dual-task execution Primary Task: Maintain dynamic balance on an adjustable balancing platform with an inclination angle adjustable in 0.5° increments from 0° to 15°. Secondary task: Simultaneously perform at least one of the following tasks: visual interference task, auditory N-back task, or tactile interference task; Step 4: Adaptive Adjustment When the success rate of three consecutive actions is greater than 85%, the difficulty of the cognitive task is increased by 1 level and the platform instability index is increased by 0.1; When a fall warning signal is detected, the safety suspension system is automatically activated and the training resistance is reduced by 30%; Step 5: Environmental simulation training In response to sudden obstacles in the virtual reality scene, the obstacle appearance time interval is a random value of 500-3000ms.

2. A training method for preventing stroke patients from falling according to claim 1, characterized in that: The visual interference task in step 3 is to track a randomly moving light spot. The frequency of the light spot movement is adjustable in the range of 0.1-3 Hz, and the movement trajectory includes horizontal, vertical, and diagonal directions.

3. A training method for preventing stroke patients from falling according to claim 1, characterized in that: The tactile interference task in step 3 uses asymmetric vibration stimulation, and applies vibration waves with a phase difference of 0°-180° and a vibration frequency of 2-5 Hz through a pressure sensor array set on the sole of the patient's foot.

4. A training method for preventing stroke patients from falling according to claim 1, characterized in that: The sudden obstacle response in step 5 includes three modes: Mode A: Avoiding a moving shopping cart in a virtual supermarket scene, with the vehicle speed dynamically adjusted between 0.2-1.2 m / s; Mode B: walking on irregular stairs, with the step height difference varying randomly between 5 and 25 cm; Mode C: Catapult-type solid obstacles are ejected onto the training path by a pneumatic drive.

5. A training system for implementing a training method for preventing stroke patients from falling as claimed in any one of claims 1 to 4, characterized in that: include: Dynamic assessment module: Contains an IMU sensor placed on the patient's waist and a plantar pressure distribution detection board; Multimodal training platform: It consists of an anti-interference balance platform, an intelligent weight-reducing treadmill, and a visual interference generator. The base support area of ​​the balance platform can be between 100-600cm 2 Electric adjustment within the range; Central controller: processes EMG, COP, and EEG signals in real time, and has a built-in LSTM-based fall risk prediction model; Safety protection unit: includes electromagnetic damping device (damping coefficient 5-200N·s / m adjustable) and torso safety suspension system.

6. The training system according to claim 5, characterized in that: The anti-interference balance platform is integrated with a sudden load transfer device, which applies a lateral load transfer of 0.5-3kg in <100ms to trigger anticipatory posture adjustment training.

7. The training system according to claim 5, characterized in that: The visual interference generator includes a laser projection device and a programmable reflector group, which generates a dynamic light spot with a diameter of 2-10 cm and a position positioning error of ≤1 cm.

8. The training system according to claim 5, characterized in that: The central controller follows the following rules when performing parameter adjustment: If the real-time EMG asymmetry index is > the preset threshold, then: Reduce the training resistance on the affected side by 15% and simultaneously increase the difficulty of cognitive tasks on the unaffected side; If the FRI value decreases by less than 5% for 2 consecutive minutes, then: Start mode C in environmental simulation training.