Dysphagia rehabilitation training method based on VR equipment

Abstract

The invention belongs to the technical field of medical treatment, and provides a dysphagia rehabilitation training method based on VR (virtual reality) equipment aiming at the problem that the existing dysphagia treatment cannot accurately recognize the dysphagia action of a patient and cannot provide a personalized rehabilitation scheme. The method comprises the following steps: acquiring a first electromyographic signal swallowing by a healthy person and a second electromyographic signal swallowing by a patient, slicing the signals by using an LSTM model, and comparing the difference between the two signals through a specially designed sliding window to determine the obstacle action of the patient; simulating treatment effects of different swallowing actions based on a neural network model, and selecting the swallowing action with the highest learning rate; and finally, making the action into a VR training model, and guiding the patient to train by virtue of VR equipment. According to the invention, by accurately diagnosing the dysphagia action, formulating a personalized treatment scheme and utilizing the VR technology, the training effect and interestingness of patient training are enhanced, and the quality of dysphagia rehabilitation training is effectively improved.

Description

Technical Field

[0001] The present invention belongs to the field of medical technology and relates to a dysphagia rehabilitation training method based on VR equipment. Background Art

[0002] Swallowing disorder is a common clinical problem, which manifests as obstruction in the transportation of food or liquid from the mouth to the stomach, seriously affecting the patient's nutritional intake, quality of life and respiratory safety. In the field of neurosurgery, the incidence of postoperative dysphagia increases significantly due to the operation of cranial nerves and surrounding muscle tissues in skull base tumor surgery. According to statistics from the "Chinese Journal of Neurosurgery" in 2022, approximately 40% to 65% of patients after skull base tumor surgery will experience varying degrees of swallowing dysfunction. Such disorders not only increase the risk of aspiration pneumonia, but may also lead to complications such as malnutrition and psychological anxiety, seriously threatening the patient's recovery process.

[0003] Traditional dysphagia rehabilitation training often relies on repetitive oral motor exercises or video tutorials, resulting in limited and unfocused training. For example, patients are required to imitate standard swallowing movements over a long period of time. However, such methods cannot accurately identify individual differences in dysphagia and cannot dynamically adjust training intensity based on the patient's degree of neurological damage. Furthermore, existing training models are tedious, resulting in low patient engagement and poor compliance, which directly impacts rehabilitation outcomes. Summary of the Invention

[0004] In order to solve the problems existing in the background technology, the present invention proposes a dysphagia rehabilitation training method based on VR equipment.

[0005] In order to achieve the above object, the technical solution adopted by the present invention is as follows: A dysphagia rehabilitation training method based on VR equipment, comprising: Setting a first electromyographic signal, where the first electromyographic signal is an electromyographic signal generated when a healthy person swallows, and obtaining the electromyographic signal of the patient when swallowing as a second electromyographic signal; Slicing the first and second myoelectric signals according to the swallowing action, comparing the first and second myoelectric signals in each slice based on a sliding window, determining the difference between the first and second myoelectric signals, and obtaining the corresponding obstructive action during the patient's swallowing process; According to the corresponding obstructive movements during the patient's swallowing process, the corresponding swallowing movements are selected as the treatment plan. The patient's swallowing disorder treatment is simulated based on the neural network model. The swallowing movement with the best treatment effect is selected to generate a VR training model, and the patient is given swallowing disorder rehabilitation training guidance through VR equipment.

[0006] Furthermore, the method for slicing the first electromyographic signal and the second electromyographic signal according to the swallowing action is: The electromyographic signal of a single swallowing action during swallowing is used as the third electromyographic signal, and the third electromyographic signal is input into the LSTM model for reading. The LSTM model retrieves the first electromyographic signal and the second electromyographic signal, and segments the signal associated with the third electromyographic signal; By reading different swallowing movements during the swallowing process, the first electromyographic signal and the second electromyographic signal are sliced.

[0007] Furthermore, the specific method for comparing the first electromyographic signal and the second electromyographic signal in each slice based on the sliding window is: A Cartesian coordinate system is established, with the x-axis located at the transverse center axis of the electromyographic signal, and a sliding window symmetrical about the x-axis is established.

[0008] Furthermore, the sliding window is divided into three parts, where: The first part of the window is arranged horizontally and slides at a position symmetrical about the x-axis, and the height of the first part of the window is equal to the average amplitude of the second electromyographic signal; The second part of the window is the first part of the window, which is shifted upwards to the height of the first part of the window and then slides horizontally; The third window is the first window, which is shifted downward by the height of the first window and arranged horizontally for sliding.

[0009] Furthermore, the specific method for simulating the treatment of dysphagia for patients based on the neural network model is: Each swallowing action has multiple training models for patients to learn; A first neural network is established based on any training model of a swallowing action, wherein the first neural network executes the swallowing action and outputs an electromyographic signal of the swallowing action to perform a teaching simulation; A second neural network is established based on the patient's current state. The second neural network learns the execution electromyographic signal of the swallowing action output by the first neural network, simulates the patient's learning, and obtains the learning rate of the training model.

[0010] Furthermore, the second neural network can modify the learning process according to the patient's actual recovery situation.

[0011] Compared with the prior art, the present invention has the following beneficial effects: The present invention achieves accurate diagnosis of obstructive movements by comparing the differences in electromyographic signals between healthy people and patients, slicing the signals using an LSTM model, and performing detailed comparisons using sliding window technology.

[0012] The present invention simulates the effects of different swallowing movements based on a neural network model and selects the swallowing movement with the highest learning rate as the treatment plan, ensuring the personalization and effectiveness of the treatment plan.

[0013] The present invention utilizes VR equipment to simulate and generate intuitive and highly interactive rehabilitation training scenes, thereby enhancing patients' sense of interest and training effects. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Figure 1 It is a flow chart of the method of the present invention; Figure 2 It is a workflow diagram of the neural network model of the present invention; Figure 3 This is a schematic diagram of the electromyographic signal slicing principle of the present invention. DETAILED DESCRIPTION

[0015] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0016] like Figure 1-Figure 3 As shown, the technical solution adopted by the present invention is as follows: a dysphagia rehabilitation training method based on VR equipment, comprising: A first myoelectric signal is set, where the first myoelectric signal is the myoelectric signal generated when a healthy person swallows, and the myoelectric signal of the patient when swallowing is obtained as the second myoelectric signal.

[0017] The first electromyographic signal and the second electromyographic signal are sliced according to the swallowing action, and the first electromyographic signal and the second electromyographic signal in each slice are compared based on the sliding window to determine the difference between the first electromyographic signal and the second electromyographic signal to obtain the corresponding obstacle action during the patient's swallowing process.

[0018] According to the corresponding obstructive movements during the patient's swallowing process, the corresponding swallowing movements are selected as the treatment plan. The patient's swallowing disorder treatment is simulated based on the neural network model. The swallowing movement with the best treatment effect is selected to generate a VR training model, and the patient is given swallowing disorder rehabilitation training guidance through VR equipment.

[0019] Swallowing is a complex reflex action that allows the food bolus to pass from the mouth through the pharynx and esophagus into the stomach. It generally includes three stages: (1) Food is moved to the back of the tongue due to the action of the buccal muscles and tongue. Then the front of the back of the tongue is pressed against the hard palate, and the food bolus is pushed to the back of the soft palate and into the pharynx. This process is random; (2) When the food bolus enters the pharynx through the soft palate, it stimulates the receptors in the soft palate, causing a series of muscle reflex contractions. As a result, the nasopharyngeal passage and the passage between the pharynx and the trachea are blocked, breathing is stopped, the upper esophageal opening opens, and the food bolus is squeezed from the pharynx into the esophagus. This process is very fast, usually only 0.1 seconds. (3) After the food bolus enters the esophagus, it causes esophageal peristalsis, pushing the food bolus into the stomach through the cardia. Peristalsis is a forward-moving wave-like movement formed by the sequential relaxation and contraction of the esophageal muscles. There is a contraction wave at the upper end of the food bolus and a relaxation wave at the lower end. The relaxation and contraction waves continue to move downward, and the food bolus is gradually pushed into the stomach.

[0020] The EMG signals generated by a healthy person during swallowing are recorded as the first EMG signal. The EMG signals generated by the patient during swallowing are extracted as the second EMG signal. Obtaining the EMG signals from the healthy person and the patient provides basic data for subsequent comparative analysis. By comparing the EMG signals, we can objectively identify any problems with the patient's swallowing process, providing a basis for developing targeted rehabilitation training plans.

[0021] The EMG signal from a single swallowing movement during swallowing is used as the third EMG signal. This signal is then fed into the LSTM model for reading. The LSTM model retrieves the first and second EMG signals and segments the signal associated with the third EMG signal. By analyzing different swallowing movements during the swallowing process, the first and second EMG signals are sliced. Using the LSTM model to slice the EMG signals accurately distinguishes the EMG signals corresponding to different swallowing movements. This facilitates subsequent detailed analysis of the EMG signals for each individual swallowing movement, more accurately identifying the patient's difficulty with each swallowing movement.

[0022] A Cartesian coordinate system is established with the x-axis positioned at the transverse midline of the EMG signal. A sliding window symmetrical about the x-axis is created. The sliding window is divided into three sections: the first section slides horizontally, symmetrically about the x-axis, with the height of the first section equal to the average amplitude of the second EMG signal; the second section slides horizontally, shifted upward by the first section's height; and the third section slides horizontally, shifted downward by the first section's height. In this way, the first and second EMG signals in each slice are compared to determine the difference between them, identifying the corresponding obstructive movement during the patient's swallowing process. Using the sliding window method to compare EMG signals allows for a comprehensive and detailed analysis of the differences between EMG signals in healthy individuals and patients during each swallowing movement. This comparison method can more accurately identify deviations between the patient's EMG signals and those of healthy individuals, thereby determining the specific obstructive movement during the patient's swallowing process and providing precise information for developing personalized rehabilitation training programs.

[0023] Each swallowing movement has multiple training models for patients to learn. To adapt to different patients' adaptability to different training models, the training model with the highest patient adaptability is selected for patient instruction. The highest adaptability means the best learning outcome for the patient.

[0024] A training model refers to a virtual training program or action sequence pre-set in the database for a specific swallowing step. This includes training methods for the tongue and throat muscles, as well as training methods or action sequences for the interconnected processes of different swallowing processes. For example, when a food bolus enters the throat through the soft palate, it stimulates the receptors in the soft palate, triggering a series of muscle reflex contractions, which results in a virtual training program or action sequence that closes the nasopharyngeal passage and the passage between the pharynx and the trachea.

[0025] According to the corresponding obstructive movements during the patient's swallowing process, the corresponding swallowing movements are selected as the treatment plan, and the patient's swallowing disorder treatment is simulated based on the neural network model.

[0026] A first neural network is established based on any training model of a swallowing action. The first neural network executes the swallowing action and outputs an execution electromyographic signal of the swallowing action to perform teaching simulation.

[0027] When a food bolus is swallowed through the soft palate, it stimulates the receptors in the soft palate, causing a series of muscle reflex contractions, which is set as a teaching action. Based on this teaching action, multiple training models corresponding to the current teaching action are extracted from the database. The first neural network learns one of the training models in the current teaching action and outputs the execution electromyographic signal corresponding to the training model to perform teaching simulation. The other training models of the current teaching action are also simulated by establishing the first neural network.

[0028] A second neural network is established based on the patient's current state. The second neural network learns the execution electromyographic signal of the swallowing action output by the first neural network, performs patient learning simulation, and obtains the learning rate of the training model.

[0029] A one-to-one corresponding first neural network is established for different training models of the current swallowing action. Each first neural network outputs the execution electromyographic signal of the corresponding training model. All execution electromyographic signals are input into the second neural network. The second neural network simulates the patient's learning and obtains the patient's learning rate for each training model. The electromyographic signal with the highest learning rate is selected from it, and the training model corresponding to the electromyographic signal is selected to generate a VR scene. The immersive rehabilitation training environment presented by the VR device is provided for patients to learn.

[0030] Patients use VR equipment to observe and imitate training movements, and the system simultaneously collects the patient's electromyographic signals and evaluates the completion of the movements in real time. Specific training scenarios are customized based on the patient's impaired movements (such as abnormalities of the tongue and throat muscles), such as simulating different food textures or swallowing stages; or specific rehabilitation scenarios, such as a swallowing reflex training scenario, which involves rubbing the thyroid cartilage up and down to the skin below the mandible with your fingers, causing the mandible to move up and down and the tongue to move forward and backward, thereby inducing swallowing. VR scenarios improve patient training compliance through intuitive and interesting interactive forms, while combining electromyographic signal feedback to achieve precise rehabilitation guidance.

[0031] The training model corresponding to the electromyographic signal with the highest learning rate has the best learning effect and the best rehabilitation effect for a certain patient.

[0032] By simulating the therapeutic effects of different swallowing maneuvers using a neural network model, we can predict the rehabilitation effect of each swallowing maneuver on the patient before actual training. By selecting the swallowing maneuver with the highest learning rate (i.e., the one with the best rehabilitation effect), we can develop a more scientific and effective treatment plan. This can improve the relevance and effectiveness of rehabilitation training, avoid ineffective training for patients, and save rehabilitation time and resources.

[0033] After teaching the patient about dysphagia, the accuracy of the second neural network's simulation of the patient's learning is determined based on the patient's recovery. Based on this accuracy, the second neural network is modified, including by deleting some fully connected layers and modifying the weights within each fully connected layer.

[0034] The most effective swallowing movements are selected and generated into VR scenes. VR equipment is then used to provide patients with swallowing disorder rehabilitation training guidance. VR technology provides patients with an immersive rehabilitation training experience. By showcasing optimal swallowing movements in VR scenes, patients can more intuitively learn and imitate these movements, making training more engaging and engaging. Furthermore, the VR environment can simulate a variety of scenarios, making training more realistic and diverse, and helping to improve patient rehabilitation outcomes.

[0035] This invention aims to develop personalized rehabilitation training plans for patients with dysphagia by utilizing electromyographic signal analysis, neural networks, and VR technology. By comparing the electromyographic signals of healthy individuals with those of patients, the patient's dysphagia movements are identified. The neural network is then used to simulate the therapeutic effects of different swallowing movements. Finally, a VR training scenario is generated to guide the patient's rehabilitation training.

[0036] Although the present invention has been described in detail with reference to the aforementioned embodiments, it is still possible for those skilled in the art to modify the technical solutions described in the aforementioned embodiments, or to make equivalent substitutions for some of the technical features therein. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A dysphagia rehabilitation training method based on VR equipment, characterized in that: Includes: Setting a first electromyographic signal, where the first electromyographic signal is an electromyographic signal generated when a healthy person swallows, and obtaining the electromyographic signal of the patient when swallowing as a second electromyographic signal; Slicing the first and second myoelectric signals according to the swallowing action, comparing the first and second myoelectric signals in each slice based on a sliding window, determining the difference between the first and second myoelectric signals, and obtaining the corresponding obstructive action during the patient's swallowing process; According to the corresponding obstructive movements during the patient's swallowing process, the corresponding swallowing movements are selected as the treatment plan. The patient's swallowing disorder treatment is simulated based on the neural network model. The swallowing movement with the best treatment effect is selected to generate a VR training model, and the patient is given swallowing disorder rehabilitation training guidance through VR equipment.

2. The dysphagia rehabilitation training method based on VR equipment according to claim 1, characterized in that: The method for slicing the first electromyographic signal and the second electromyographic signal according to the swallowing action is: The electromyographic signal of a single swallowing action during swallowing is used as the third electromyographic signal, and the third electromyographic signal is input into the LSTM model for reading. The LSTM model retrieves the first electromyographic signal and the second electromyographic signal, and segments the signal associated with the third electromyographic signal; By reading different swallowing movements during the swallowing process, the first electromyographic signal and the second electromyographic signal are sliced.

3. The dysphagia rehabilitation training method based on VR equipment according to claim 1, characterized in that: The specific method for comparing the first electromyographic signal and the second electromyographic signal in each slice based on the sliding window is: A Cartesian coordinate system is established, with the x-axis located at the transverse center axis of the electromyographic signal, and a sliding window symmetrical about the x-axis is established.

4. The dysphagia rehabilitation training method based on VR equipment according to claim 3, characterized in that: The sliding window is divided into three parts: The first part of the window is arranged horizontally and slides at a position symmetrical to the x-axis, and the height of the first part of the window is equal to the average amplitude of the second electromyographic signal; The second part of the window is the first part of the window, which is shifted upwards to the height of the first part of the window and then slides horizontally; The third window is the first window, which is shifted downward by the height of the first window and arranged horizontally for sliding.

5. The dysphagia rehabilitation training method based on VR equipment according to claim 1, characterized in that: The specific method for simulating the patient's dysphagia treatment based on the neural network model is: Each swallowing action has multiple training models for patients to learn; A first neural network is established based on any training model of a swallowing action, wherein the first neural network executes the swallowing action and outputs an electromyographic signal of the swallowing action to perform a teaching simulation; A second neural network is established based on the patient's current state. The second neural network learns the execution electromyographic signal of the swallowing action output by the first neural network, simulates the patient's learning, and obtains the learning rate of the training model.

6. The dysphagia rehabilitation training method based on VR equipment according to claim 5, characterized in that: The second neural network can modify the learning process according to the patient's actual recovery situation.

Images