Treatment method and system based on multifunctional cervical and lumbar vertebra comprehensive treatment machine

Abstract

The invention relates to the technical field of rehabilitation training, in particular to a treatment method and system based on a multifunctional cervical and lumbar vertebra comprehensive treatment machine, and the method comprises the steps: obtaining multi-modal sensing data in a rehabilitation training process; based on the multi-mode sensing data, a rehabilitation score representing the current rehabilitation state of the patient is obtained through calculation, and a training mode is determined according to the rehabilitation score; the fuzzy controller is used for taking a myoelectricity fatigue index, a joint angle deviation and a pressure risk coefficient which are monitored in real time as input variables of the fuzzy controller in the training process of the training mode; and performing fuzzy reasoning through a preset fuzzy rule base. According to the method, multi-modal sensor data such as myoelectricity, angle and pressure are fused, and accurate and objective quantitative scoring is carried out on the state of the patient by using the convolutional neural network and the random forest model. And on the basis of a scoring result, a personalized training mode can be automatically matched.

Description

Technical Field

[0001] This invention relates to the field of rehabilitation training technology. More specifically, this invention relates to a treatment method and system based on a multifunctional cervical and lumbar spine comprehensive treatment machine. Background Technology

[0002] Existing rehabilitation training equipment typically collects patient training data using a single type of sensor (such as an angle sensor or a pressure sensor) and combines it with fixed training programs to assist rehabilitation. While these devices can help patients maintain or restore joint range of motion to some extent, they often suffer from limitations in data dimensionality and comprehensive assessment results in practice. Furthermore, traditional rehabilitation scoring methods rely primarily on manual observation and simple quantitative indicators, lacking a comprehensive analysis of the patient's movement patterns, muscle function status, and fatigue levels. In addition, training programs are mostly pre-programmed, making it difficult to personalize them according to the patient's actual condition in a timely manner.

[0003] However, the shortcomings of existing technologies are: their assessment methods lack multimodal data fusion, and cannot simultaneously consider multi-dimensional information such as electromyographic signals, joint angles, pressure distribution, and movement trajectories, resulting in inaccurate evaluation results of rehabilitation effects; the training control methods also lack flexibility and cannot cope with situations such as fatigue, pain, or compensatory movements that occur during the training process, thus resulting in low rehabilitation efficiency and high risk of secondary injury. Summary of the Invention

[0004] This invention provides a treatment method and system based on a multifunctional cervical and lumbar spine comprehensive treatment machine, aiming to solve the problems of low rehabilitation efficiency and high risk of secondary injury caused by the lack of flexibility in training control methods in related technologies, which cannot cope with fatigue, pain or compensatory movements that occur in patients during training.

[0005] In a first aspect, the present invention provides a treatment method based on a multifunctional cervical and lumbar spine comprehensive treatment machine, comprising: acquiring multimodal sensing data during rehabilitation training; calculating a rehabilitation score characterizing the patient's current rehabilitation status based on the multimodal sensing data, and determining a training mode according to the magnitude of the rehabilitation score; using real-time monitored electromyographic fatigue index, joint angle deviation, and pressure risk coefficient as input variables of a fuzzy controller during the training process of the training mode; and generating control instructions for adjusting the training parameters of the training mode in real time through fuzzy inference using a preset fuzzy rule base, so as to achieve closed-loop adaptive rehabilitation training.

[0006] Furthermore, acquiring multimodal sensing data during rehabilitation training includes: the multimodal sensing data includes at least one of the following: electromyographic signal sequences acquired by an electromyographic sensor, joint angle sequences acquired by an angle sensor, and pressure distribution acquired by a pressure sensing matrix.

[0007] Furthermore, a convolutional neural network model is used to extract features from the multimodal sensing data to generate a high-dimensional feature vector containing joint angle features, motion trajectory features, and muscle force symmetry features; the high-dimensional feature vector is input into a pre-trained random forest model, and the random forest model outputs the rehabilitation score.

[0008] Furthermore, determining the training mode based on the rehabilitation score includes: if the rehabilitation score is greater than a first preset threshold, then the training mode is determined to be an active training mode; if the rehabilitation score is less than or equal to the first preset threshold but greater than a second preset threshold, then the training mode is determined to be a mixed training mode; if the rehabilitation score is less than or equal to the second preset threshold, then the training mode is determined to be a passive training mode.

[0009] Furthermore, the electromyographic fatigue index is obtained by calculating the rate of decrease of the average power frequency (MPF) of the electromyographic signal over time; the joint angle deviation is the difference between the current real-time joint angle and the preset target angle.

[0010] Furthermore, the fuzzification process of the electromyographic fatigue index includes: if the electromyographic signal value is less than 100 μV, its membership degree is in a low fuzzy set; if the electromyographic signal value is in the range of 100 μV to 250 μV, its membership degree is in a medium fuzzy set; if the electromyographic signal value is greater than 200 μV, its membership degree is in a high fuzzy set.

[0011] Furthermore, the training parameters include at least one of the following: the training resistance value applied to the patient, the angular velocity of the patient's limb movement driven by the treatment machine, and the ratio of active training duration to passive training duration in the mixed training mode.

[0012] Furthermore, the pre-defined fuzzy rule base includes: Rule 1: IF EMG fatigue index is high AND joint angle deviation is severe THEN training resistance is significantly reduced AND active training ratio is significantly reduced; Rule 2: IF EMG fatigue index is low AND joint angle deviation is minor THEN training resistance is appropriately increased; Rule 3: IF stress risk coefficient is dangerous THEN movement speed is stopped abruptly.

[0013] Furthermore, the pressure risk coefficient is the reciprocal of the absolute value of the difference between the patient's real-time pressure value and the target pressure value.

[0014] In a second aspect, the present invention also provides a treatment system based on a multifunctional cervical and lumbar spine comprehensive treatment machine, comprising a processor and a memory, wherein the memory stores a computer program, and the processor executes the computer program to implement the treatment method based on the multifunctional cervical and lumbar spine comprehensive treatment machine as described above.

[0015] Beneficial Effects: By integrating multimodal sensor data such as electromyography, angle, and pressure, and utilizing convolutional neural networks and random forest models, the system accurately and objectively quantifies and scores the patient's condition. Based on the scoring results, personalized training modes can be automatically matched. More importantly, it introduces fuzzy logic control to monitor indicators such as patient fatigue and movement deviations in real time, dynamically adjusting training parameters to achieve personalized, intelligent, and safe rehabilitation training, significantly improving treatment effectiveness and reducing the risk of secondary injury. Attached Figure Description

[0016] Figure 1 This is a flowchart illustrating the adjustment of training parameters according to an embodiment of the present invention. Detailed Implementation

[0017] The specific embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0018] like Figure 1 As shown, S101: Collect patient data.

[0019] Specifically, when a patient begins using the device, the system first enters the initialization and evaluation phase. Specifically, after logging into the system via card swiping, facial recognition, or manual ID input, the system first performs a hardware self-test and sensor calibration. The hardware self-test includes confirming that the servo motor, resistance adjustment device, various sensors, and control terminal are all in normal working order. Sensor calibration is a crucial step to ensure the accuracy of subsequent data. For example, the system guides the patient to a completely relaxed resting state, collects the electromyographic signals at this time as a baseline value, and defines the current joint position as the zero-point angle.

[0020] It should be noted that data acquisition requires the use of the following sensors, including: Electromyography (EMG) sensors, exemplarily surface EMG sensors, attached to the core muscle groups of the patient's neck or lower back (such as the erector spinae and multifidus muscles), to collect real-time electrical activity signals of the muscles during training, i.e., EMG signals. The collected raw EMG signals are bandpass filtered (e.g., a filtering range of 20-450Hz) to remove power frequency interference and motion artifacts, and root mean square (RMS) values ​​are calculated to quantify muscle activation and fatigue levels. Angle sensors: installed at joint movement points on the treatment bed (such as the flexion-extension axis of a lumbar support frame) to accurately monitor the range of motion of relevant joints during patient movements, such as the angle of lumbar flexion or extension. Pressure sensors: embedded in the contact surface between the treatment bed and the patient's body, forming a pressure sensing matrix to analyze the pressure distribution of the patient's position and monitor for the risk of excessive local compression. Motion capture cameras: deployed on both sides of the device, based on skeletal keypoint tracking technology, to record and reconstruct the patient's movement trajectory in real time in three dimensions, especially the bending angle and movement posture of the spine. Understandably, all the data collected by the sensors is synchronously acquired and timestamped through the embedded system in the control terminal, providing high-quality, multi-dimensional data input for the subsequent comprehensive evaluation algorithm.

[0021] The system will then guide the patient through a multimodal assessment, which consists of two parts: a static assessment and a dynamic assessment.

[0022] For static assessment, the following steps are taken: the patient lies supine and at rest on the treatment bed. Electromyography (EMG) sensors deployed in key muscle groups such as the neck and lower back collect EMG signals at rest; angle sensors installed in the joints of the treatment bed (such as lumbar support frames) record the initial joint angles; and pressure sensors embedded in the contact surface of the treatment bed analyze the patient's positional pressure distribution. This step is to obtain the patient's baseline physiological parameters under no-load conditions.

[0023] For dynamic assessment, this includes guiding the patient through a set of standardized assessment movements via voice or screen prompts, such as lumbar flexion of 30° and neck rotation to the left and right. During this process, a multimodal sensor array will simultaneously collect dynamic data: electromyography (EMG) sensors record the activation level and recruitment sequence of relevant muscle groups; angle sensors monitor joint range of motion; pressure sensors analyze the shift of the center of pressure during the movement; and simultaneously, motion capture cameras deployed on both sides of the device record the patient's movement trajectory in real time using skeletal keypoint tracking technology, such as the spinal curvature angle.

[0024] At this point, all the collected data, including electromyographic signal sequences, joint angle sequences, pressure distribution matrices, and motion trajectory video streams, can be obtained for subsequent calculations and processing.

[0025] S102: Calculate the patient's rehabilitation score.

[0026] Specifically, the multi-dimensional data collected in the above steps are analyzed to generate a quantitative rehabilitation score and recommend personalized training programs. First, a CNN model is used to extract features from the raw sensor data and video stream. Because the data during the rehabilitation process has significant spatiotemporal correlations, CNN can effectively learn automatically from high-dimensional inputs and extract deep, abstract features.

[0027] Specifically, a convolutional neural network (CNN) is used to extract deep features from the input raw sensor data sequences (such as EMG time-series signals and angle change sequences) and motion capture video streams. The CNN model can automatically learn the spatiotemporal features in the data. Its construction is based on simulating the hierarchical structure of the biological visual cortex, effectively capturing local correlations and high-order abstract features. In this embodiment, the CNN model mainly extracts the following five types of key features: Joint angle features: such as maximum flexion and extension angles, average angular velocity, etc. Motion trajectory features: such as the smoothness of the movement, the deviation of the trajectory from the standard trajectory, etc. Symmetry features: analyzing the EMG signals of bilateral muscle groups to assess the symmetry of muscle force exertion.

[0028] Following the above, the CNN model can employ a structure containing three convolutional layers and two fully connected layers. The convolutional kernel size can be set to 3×3, the activation function is ReLU, and pooling layers are used for dimensionality reduction. These hyperparameter settings aim to balance the depth of feature extraction with computational efficiency. Then, the high-dimensional feature vector extracted by the CNN is input into a pre-trained random forest model for final rehabilitation scoring. The rehabilitation score ranges from 0 to 10; a higher score indicates a better patient condition. The random forest model is an ensemble learning algorithm that improves the model's accuracy and robustness by constructing multiple decision trees and aggregating their predictions. In this embodiment, the model is used to perform multi-task prediction, and the output includes: Muscle strength rating: grading the patient's muscle strength (e.g., referring to the Lovett strength rating scale from 0 to 5). Pain index: quantifying the patient's pain level by combining indirect indicators such as heart rate variability (HRV). Movement standardization: classifying the quality of movement completion, such as "normal," "compensatory," or "incorrect."

[0029] S103: Determine the training mode based on the patient's rehabilitation score.

[0030] Specifically, based on rehabilitation scores, three training modes are defined: patients with rehabilitation scores above a first threshold are assigned to the active mode; patients with rehabilitation scores between the first and second thresholds are assigned to the hybrid mode; and patients with rehabilitation scores below the second threshold are assigned to the passive mode. Further, the active mode involves the patient performing movements independently while the device provides adaptive resistance. This is suitable for patients with good muscle strength who require strength and endurance training. The passive mode involves a servo motor-driven device that gently stretches the patient's limbs to relax muscles and maintain joint mobility. This is suitable for patients in the early postoperative period or those with extremely weak muscles. The hybrid mode combines active and passive training, dynamically adjusting the ratio based on real-time assessments.

[0031] For patients with a mixed rehabilitation pattern, the initial ratio of passive to active training can be adaptively adjusted. For example, if a patient's rehabilitation score is determined to be mixed, the initial ratio could be set to 60% passive training and 40% active training.

[0032] S104: Adjust the training parameters of the training mode based on fuzzy control.

[0033] To address the fuzziness and uncertainty of human rehabilitation parameters (such as fatigue level and pain threshold) during training, this embodiment employs a dynamic training controller based on fuzzy logic to adjust training parameters in real time. This controller primarily comprises three core components: input variable fuzzification, fuzzy rule base construction, and defuzzification strategies.

[0034] Following the above, input variable fuzzification includes converting precise numerical values ​​collected by sensors into linguistic variables that conform to human fuzzy concepts. In this embodiment, three key input variables and their fuzzy sets are defined: Electromyography (EMG) fatigue index: calculated using the rate of decrease in the average power frequency (MPF) of the EMG signal. Specifically, an EMG sensor attached to the body collects electrical signals emitted by muscles. At the start of training, a reference value is recorded. As training progresses, if the muscles become increasingly fatigued, the frequency in the signal gradually decreases. The current frequency is then compared with the initial reference value, and the magnitude of the decrease is calculated. The magnitude of the decrease is the EMG fatigue index; a higher EMG fatigue index indicates greater muscle fatigue. Its fuzzy set is defined as {low, medium, high}. The membership function uses a trigonometric membership function. Specifically, low: corresponds to an EMG signal value less than 100μ; medium: corresponds to an EMG signal value between 100-250μ; high: corresponds to an EMG signal value greater than 200μ. Joint angle deviation: the difference between the current joint angle and the target angle. The fuzzy set is defined as {minor, moderate, severe}. The membership function uses a Gaussian membership function. Specifically, minor: corresponds to a deviation less than 5⁻¹. Moderate: corresponds to a deviation between 5⁻¹ and 10⁻¹. Severe: corresponds to a deviation greater than 10⁻¹. Pressure risk coefficient: local pressure value, specifically the reciprocal of the absolute value of the difference between the patient's real-time pressure value and the target pressure value. Its fuzzy set is defined as {safe, warning, dangerous}. Specifically, the membership function uses a trapezoidal membership function. Safe: corresponds to a pressure value less than 100N. Warning: corresponds to a pressure value between 100-180N. Dangerous: corresponds to a pressure value greater than 180N.

[0035] Next, a fuzzy rule base needs to be constructed: this rule base is a series of "IF-THEN" rules summarized from the clinical experience of numerous rehabilitation physicians. Exemplary key rules are as follows: Rule 1: IF (high electromyographic fatigue index) AND (severe joint angle deviation) THEN (resistance adjusted to be significantly reduced) AND (active proportion significantly reduced). Rule 2: IF (low electromyographic fatigue index) AND (minor joint angle deviation) THEN (resistance adjusted to be appropriately increased). Rule 3: IF (dangerous pressure risk factor) THEN (stretching speed stopped immediately).

[0036] The fuzzy output set obtained from fuzzy inference is converted into precise control commands. This embodiment uses the center-of-gravity method, widely used in industrial control, for defuzzification to obtain specific adjustment values ​​and output precise control commands for the adjustment parameters, thereby controlling the hardware of the active and passive training units. In active mode, if the fuzzy controller outputs a command to reduce resistance by 15%, the excitation current of the electromagnetic resistance adjustment device will decrease accordingly, thus reducing the resistance felt by the patient when performing actions. In passive mode, if the pressure sensor reports a pressure value exceeding a threshold (e.g., 200N), and the fuzzy controller outputs a command to reduce speed by 50%, the servo motor speed will immediately decrease to ensure patient safety. In hybrid mode, if a 20% decrease in the patient's heart rate variability is detected (indicating increased fatigue), the controller will dynamically reduce the proportion of active training from 40% to 0% and increase the proportion of passive relaxation.

[0037] In summary, by integrating multimodal sensing assessment, intelligent scoring, and fuzzy logic dynamic control, a closed-loop personalized rehabilitation training system was constructed, which significantly improved the accuracy, safety, and active participation of the treatment, and achieved the effects of shortening the rehabilitation cycle and reducing the risk of secondary injury.

[0038] The present invention also provides a treatment system based on a multifunctional cervical and lumbar spine comprehensive treatment machine. The system includes a processor and a memory, the memory storing computer program instructions. When the processor executes the computer program instructions, it implements the treatment method based on the multifunctional cervical and lumbar spine comprehensive treatment machine according to the first aspect of the present invention.

[0039] The system also includes other components well known to those skilled in the art, such as communication buses and communication interfaces, the settings and functions of which are known in the art and therefore will not be described in detail here.

[0040] In this invention, the aforementioned memory can be any tangible medium containing or storing a program that can be used or combined with an instruction execution system, apparatus, or device. For example, a computer-readable storage medium can be any suitable magnetic or magneto-optical storage medium, such as Resistive Random Access Memory (RRAM), Dynamic Random Access Memory (DRAM), Static Random Access Memory (SRAM), Enhanced Dynamic Random Access Memory (EDRAM), High-Bandwidth Memory (HBM), Hybrid Memory Cube (HMC), etc., or any other medium that can be used to store desired information and can be accessed by an application, module, or both. Any such computer storage medium can be part of a device or accessible to or connected to a device. Any application or module described in this invention can be implemented using computer-readable / executable instructions stored or otherwise maintained on such a computer-readable medium.

[0041] The embodiments described above are merely examples of several implementations of the present invention, and while the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these modifications and improvements all fall within the scope of protection of the present invention.

Claims

1. A treatment method based on a multifunctional cervical and lumbar spine comprehensive treatment machine, characterized in that, include: Acquire multimodal sensor data during rehabilitation training; Based on the multimodal sensing data, a rehabilitation score representing the patient's current rehabilitation status is calculated, and a training mode is determined according to the magnitude of the rehabilitation score. During the training process of the training mode, the real-time monitored electromyography fatigue index, joint angle deviation, and pressure risk coefficient are used as input variables for the fuzzy controller. Fuzzy inference is performed using a preset fuzzy rule base to generate control instructions for adjusting training parameters of the training mode in real time, thereby achieving closed-loop adaptive rehabilitation training.

2. The treatment method based on the multifunctional cervical and lumbar spine comprehensive treatment machine according to claim 1, characterized in that, Acquire multimodal sensor data during rehabilitation training, including: The multimodal sensing data includes at least one of the following: an electromyographic signal sequence acquired by an electromyographic sensor, a joint angle sequence acquired by an angle sensor, and a pressure distribution acquired by a pressure sensing matrix.

3. The treatment method based on the multifunctional cervical and lumbar spine comprehensive treatment machine according to claim 1, characterized in that, A convolutional neural network model is used to extract features from the multimodal sensing data to generate a high-dimensional feature vector containing joint angle features, motion trajectory features, and muscle force symmetry features. The high-dimensional feature vector is input into a pre-trained random forest model, and the random forest model outputs the rehabilitation score.

4. The treatment method based on the multifunctional cervical and lumbar spine comprehensive treatment machine according to claim 1, characterized in that, The training mode is determined based on the magnitude of the rehabilitation score, including: If the rehabilitation score is greater than the first preset threshold, then the training mode is determined to be an active training mode; If the rehabilitation score is less than or equal to the first preset threshold and greater than the second preset threshold, then the training mode is determined to be a hybrid training mode. If the rehabilitation score is less than or equal to the second preset threshold, then the training mode is determined to be a passive training mode.

5. The treatment method based on the multifunctional cervical and lumbar spine comprehensive treatment machine according to claim 1, characterized in that, The electromyographic fatigue index is obtained by calculating the rate of decrease of the average power frequency (MPF) of the electromyographic signal over time. The joint angle deviation is the difference between the current real-time joint angle and the preset target angle.

6. The treatment method based on the multifunctional cervical and lumbar spine comprehensive treatment machine according to claim 1, characterized in that, The fuzzification process of the electromyographic fatigue index includes: If the electromyographic signal value is less than 100 μV, its membership degree is in a low-fuzzy set; If the electromyographic signal value is in the range of 100μV to 250μV, then its membership degree is in the fuzzy set. If the electromyographic signal value is greater than 200 μV, then its membership degree is in a highly fuzzy set.

7. The treatment method based on the multifunctional cervical and lumbar spine comprehensive treatment machine according to claim 1, characterized in that, The training parameters include at least one of the following: the training resistance applied to the patient, the angular velocity of the patient's limb movement driven by the therapeutic machine, and the ratio of active training duration to passive training duration in the mixed training mode.

8. The treatment method based on the multifunctional cervical and lumbar spine comprehensive treatment machine according to claim 1, characterized in that, The preset fuzzy rule base includes: Rule 1: IF high electromyography fatigue index AND severe joint angle deviation THEN significantly reduced training resistance AND significantly reduced active training ratio; Rule 2: IF low electromyography fatigue index AND minimal joint angle deviation THEN appropriately increase training resistance; Rule 3: IF pressure risk factor is dangerous THEN movement speed emergency stop.

9. The treatment method based on the multifunctional cervical and lumbar spine comprehensive treatment machine according to claim 1, characterized in that, The pressure risk factor is the reciprocal of the absolute value of the difference between the patient's real-time pressure value and the target pressure value.

10. A treatment system based on a multifunctional cervical and lumbar spine comprehensive treatment machine, comprising a processor and a memory, characterized in that, The memory stores a computer program, and the processor executes the computer program to implement the treatment method based on the multifunctional cervical and lumbar spine comprehensive treatment machine as described in any one of claims 1-9.

Images