Method and system for monitoring and evaluating swallowing function of stroke patient

Abstract

The invention relates to the technical field of physical disorder assessment, and discloses a method and a system for monitoring and assessing the swallowing function of a stroke patient, and the method comprises the steps: collecting the electromyographic signals of the sEMG of the sEMG of the stroke patient through arranging a multi-channel sEMG electrode; and based on the acquired multi-channel electromyographic signals, performing analysis processing through Fourier transform and a machine learning algorithm combining independent component analysis and wavelet transform to obtain an analysis result of deep muscle electrical activity and an independent activity result of different muscle groups, and based on the result, evaluating the swallowing function of the stroke patient. Therefore, the functional conditions of specific muscles and deep muscles are mastered, the accuracy of screening and evaluating dysphagia of a stroke patient is improved, and effective support is provided for subsequent clinical diagnosis.

Description

Technical Field

[0001] The present invention relates to the technical field of physical impairment assessment, and in particular to a method and system for monitoring and assessing swallowing function of stroke patients. Background Art

[0002] Stroke is the second leading cause of death in the world and the first in China. In my country, the annual incidence of stroke is about 150 / 100,000, and it is increasing at a rate of about 8.7% per year, becoming the leading cause of death among Chinese people.

[0003] According to research statistics, about 45% of stroke patients have dysphagia when they are admitted to the hospital. Although most of them will improve within 2 to 4 weeks, they will still face long-term swallowing problems. Among them, 43% to 54% of patients will experience aspiration, 37% of whom will develop pneumonia, and 4% will die from pneumonia. Malnutrition is also common. During hospitalization, patients with dysphagia will experience malnutrition and the risk of malnutrition will increase significantly (from 19% to 65%). Malnutrition will increase the incidence of lung infection, patient mortality and recurrence rate. In addition, dysphagia will also affect the quality of life. The detection rate of depression in patients with dysphagia after stroke is significantly higher than that in those without dysphagia (42.86% and 14.43%, respectively), and it directly affects the rehabilitation effect and quality of life. Screening for dysphagia can assess the possibility of aspiration before complications such as pneumonia, airway obstruction, and suffocation occur. In the field of stroke treatment, experts and scholars generally believe that all stroke patients should undergo rapid and timely swallowing function screening.

[0004] Surface electromyography (sEMG) is a non-invasive method for screening and evaluating swallowing disorders. It monitors the electromyographic activity of swallowing-related muscles during exercise, and qualitatively analyzes their functional status based on the collected electromyographic signals. The treatment sequence and focus of the swallowing muscles can be adjusted according to the sEMG analysis results, which can significantly improve the efficacy. However, it is impossible to grasp the functional status of specific muscles and deep muscles, and it is difficult to obtain data on structural displacement and smaller muscles. Summary of the invention

[0005] In view of this, the purpose of the present invention is to provide a method and system for monitoring and evaluating swallowing function in stroke patients, so as to solve the problem that when performing swallowing disorder screening and evaluation on stroke patients through surface electromyography, the functional status of specific muscles and deep muscles cannot be grasped, which in turn affects the results of swallowing disorder screening and evaluation.

[0006] The first aspect of the present invention discloses a method for monitoring and evaluating swallowing function of stroke patients, the method comprising the following steps:

[0007] Step S1, placing a multi-channel sEMG electrode on the surface of the target swallowing muscle of the stroke patient, collecting electromyographic signals through the electrodes, and obtaining a first electromyographic signal;

[0008] Step S2, storing the first electromyographic signal in a multi-channel sEMG device, processing and analyzing the first electromyographic signal through the multi-channel sEMG device, and obtaining an analysis result of deep muscle electrical activity;

[0009] Step S3, performing separation and extraction operations on the first electromyographic signal through a machine learning algorithm to obtain independent activity results of different muscle groups;

[0010] Step S4: performing an evaluation operation on the swallowing function of the stroke patient based on the analysis results of the deep muscle electrical activity and the independent activity results of different muscle groups.

[0011] Furthermore, the electrode placement locations include surface muscles of the larynx, oral cavity, and upper esophagus; and the number of electrode channels of the multi-channel sEMG device is ≥16.

[0012] Furthermore, in step S2, before the first electromyographic signal is processed and analyzed by the multi-channel sEMG device, the method further includes performing noise reduction processing on the first electromyographic signal by using an adaptive filtering and artifact elimination algorithm.

[0013] Furthermore, in step S2, the first electromyographic signal is processed and analyzed by a multi-channel sEMG device to obtain the analysis result of the deep muscle electrical activity, which specifically includes:

[0014] The preprocessed multi-channel electromyographic signals are constructed into a matrix, wherein each column represents the signal of one channel;

[0015] The signal is converted from the time domain to the time-frequency domain by short-time Fourier transform (STFT), and the source signal is separated in the time-frequency domain by non-negative matrix decomposition algorithm, and the time-frequency matrix is ​​decomposed into two non-negative matrices, namely the basis matrix and the coefficient matrix; wherein the basis matrix represents the characteristic spectrum and the coefficient matrix represents the time activation;

[0016] Convert the separated time-frequency representation back to the time domain signal, perform inverse STFT on each independent component, and obtain the separated time domain signal;

[0017] Evaluate and validate the isolated components to determine the corresponding deep muscle activity;

[0018] The muscle activity intensity is obtained based on the electrical activity intensity of the deep muscles at different time points. The intensity distribution of different frequency components is displayed based on the spectrum analysis results of the electromyographic signal, and the muscle fiber type and fatigue status are obtained. The timing of muscle activity is obtained based on the activity time series of the deep muscles during swallowing. The coordination and synchronization of swallowing movements are evaluated based on the coordinated activities between different deep muscles, and abnormal muscle activity patterns are identified.

[0019] Furthermore, the machine learning algorithm in step S3 includes independent component analysis and wavelet transform.

[0020] Furthermore, before step S4, the method further includes:

[0021] Collecting electrocardiogram, blood oxygen saturation and clinical imaging data of the stroke patient, performing multimodal fusion analysis operation on the electrocardiogram, blood oxygen saturation, clinical imaging data and first electromyographic signal of the stroke patient, and obtaining fusion analysis results;

[0022] After obtaining the fusion analysis result, step S4 includes:

[0023] The swallowing function of stroke patients was evaluated based on the fusion analysis results, the analysis results of the deep muscle electrical activity of stroke patients, and the independent activity results of different muscle groups.

[0024] Furthermore, the method also includes obtaining the electromyographic signals of the stroke patient within a preset time period and the corresponding evaluation results and expert diagnosis results, and establishing an abnormal feedback mechanism based on the electromyographic signals of the stroke patient within a preset time period and the corresponding evaluation results and expert diagnosis results.

[0025] The second aspect of the present invention discloses a swallowing function monitoring and evaluation system for stroke patients, which is implemented based on the method disclosed in the first aspect, and includes a multi-channel sEMG device, a second data processing unit, and an evaluation unit; wherein the multi-channel sEMG device also includes a first acquisition unit, a first storage unit, and a first data processing unit;

[0026] The multi-channel sEMG device is used to provide multi-channel sEMG electrodes, and allows the user to place the multi-channel sEMG electrodes on the surface of the target swallowing muscles of the stroke patient;

[0027] The first acquisition unit is used to acquire the electromyographic signal through the multi-channel sEMG electrode to obtain the first electromyographic signal;

[0028] The first storage unit is used to store a first electromyographic signal;

[0029] The first data processing unit is used to process and analyze the first electromyographic signal to obtain an analysis result of deep muscle electrical activity;

[0030] The second data processing unit is used to perform separation and extraction operations on the first electromyographic signal through a machine learning algorithm to obtain independent activity results of different muscle groups;

[0031] The evaluation unit is used to perform an evaluation operation on the swallowing function of a stroke patient based on the analysis results of deep muscle electrical activity and the independent activity results of different muscle groups.

[0032] Furthermore, the system further comprises a second acquisition unit, which is used to acquire electrocardiogram, blood oxygen saturation and clinical imaging data of the stroke patient;

[0033] The second data processing unit is further used to perform a multimodal fusion analysis operation on the electrocardiogram, blood oxygen saturation, clinical imaging data and the first electromyographic signal of the stroke patient to obtain a fusion analysis result;

[0034] After obtaining the fusion analysis results, the swallowing function of the stroke patient is evaluated by the evaluation unit based on the fusion analysis results, the analysis results of the deep muscle electrical activity of the stroke patient, and the independent activity results of different muscle groups.

[0035] Furthermore, the system also includes a feedback unit, which is used to obtain the electromyographic signals of the stroke patient within a preset time period and the corresponding evaluation results and expert diagnosis results, and establish an abnormal feedback mechanism based on the electromyographic signals of the stroke patient within a preset time period and the corresponding evaluation results and expert diagnosis results.

[0036] Compared with the prior art, the present invention has the following beneficial effects:

[0037] The present invention collects the swallowing muscle electromyographic signals of stroke patients by setting up multi-channel sEMG electrodes, and analyzes and processes the collected multi-channel electromyographic signals through Fourier transform operation and a machine learning algorithm combining independent component analysis and wavelet transform, so as to obtain the analysis results of deep muscle electrical activity and the independent activity results of different muscle groups. Based on the results, the swallowing function of stroke patients is evaluated, so as to grasp the functional status of specific muscles and deep muscles, improve the accuracy of screening and evaluation of swallowing disorders in stroke patients, and provide effective support for subsequent clinical diagnosis. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] The drawings described herein are used to provide a further understanding of the embodiments of the present invention, constitute a part of the economic application, and do not constitute a limitation of the embodiments of the present invention. In the drawings:

[0039] Figure 1 A schematic diagram of a flow chart of a method for monitoring and evaluating swallowing function of stroke patients disclosed in an embodiment of the present invention;

[0040] Figure 2 The present invention is a schematic structural diagram of a swallowing function monitoring and evaluation system for stroke patients disclosed in yet another embodiment of the present invention. DETAILED DESCRIPTION

[0041] In order to enable those skilled in the art to better understand the solutions of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all of the embodiments.

[0042] Embodiment 1

[0043] The first aspect of the present invention discloses a method for monitoring and evaluating swallowing function in stroke patients. Figure 1 , Figure 1 The present invention discloses a method for monitoring and evaluating swallowing function of stroke patients, which comprises the following steps:

[0044] Step S1: placing multi-channel sEMG electrodes on the surface of target swallowing muscles of a stroke patient, collecting electromyographic signals through the electrodes, and obtaining a first electromyographic signal.

[0045] Specifically, in an embodiment of the present invention, a sEMG device with multi-channel input is used, and each electrode is connected to an independent acquisition channel to simultaneously acquire electrical signals of multiple muscle groups.

[0046] The electrodes are preferably placed on the surface muscles of the larynx, oral cavity and upper esophagus. Placing electrodes on the surface of the larynx of stroke patients can be used to monitor the electromyographic signals of the larynx-related muscles (such as the pharyngeal muscles and cricothyroid muscles). Placing electrodes on the surface muscles on both sides of the oral cavity (such as the orbicularis oris and masseter muscles) can capture the muscle activity during swallowing. Placing electrodes on the surface muscles of the upper esophagus can be used to monitor the electrical activity of the upper esophageal-related muscles.

[0047] When arranging electrodes, they can be fixed to the muscle surface using medical tape or electrode gel to ensure close contact between the electrodes and the skin, reduce motion artifacts and signal interference, and reasonably determine the distance between electrodes based on the anatomical structure of the target muscle to ensure effective signal acquisition and spatial resolution.

[0048] Step S2: storing the first electromyographic signal in a multi-channel sEMG device, processing and analyzing the first electromyographic signal through the multi-channel sEMG device, and obtaining an analysis result of deep muscle electrical activity.

[0049] Specifically, the preprocessed multi-channel electromyographic signal is constructed into a matrix, in which each column represents the signal of a channel. The signal is converted from the time domain to the time-frequency domain by short-time Fourier transform (STFT), and the source signal is separated in the time-frequency domain by a non-negative matrix decomposition algorithm. The time-frequency matrix is ​​decomposed into two non-negative matrices, namely the basis matrix and the coefficient matrix, where the basis matrix represents the characteristic spectrum and the coefficient matrix represents the time activation. The separated time-frequency representation is converted back to the time domain signal, and each independent component is inversely STFTed to obtain the separated time domain signal.

[0050] Evaluate and verify the separated independent components to ensure that they correspond to actual deep muscle activity. Specifically, check the spectrum of the separated signal to verify that it matches the frequency characteristics of the electromyographic signal. Analyze the time series of the separated signal to verify that it is consistent with known muscle activity patterns. When reference data of deep muscle activity is available (such as ultrasound or MRI imaging), verify the separation effect by comparing the separated signal with the reference data.

[0051] The electrical activity of the deep muscles is then analyzed and evaluated. Specifically, the muscle activity intensity is obtained based on the electrical activity intensity of the deep muscles at different time points; the intensity distribution of different frequency components is displayed based on the spectrum analysis results of the electromyographic signal to obtain the muscle fiber type and fatigue status; the muscle activity timing is obtained based on the activity time series of the deep muscles during swallowing, including the start time, duration and activity pattern; the coordination and synchronization of the swallowing movement is evaluated based on the coordinated activity between different deep muscles; and abnormal muscle activity patterns are identified to provide support for subsequent clinical diagnosis.

[0052] Step S3: performing separation and extraction operations on the first electromyographic signal through a machine learning algorithm to obtain independent activity results of different muscle groups.

[0053] Further, as a preferred implementation of Example 1 of the present invention, the machine learning algorithm includes independent component analysis and wavelet transform.

[0054] Specifically, the preprocessed multi-channel electromyographic signals are constructed into a matrix X, in which each column represents a signal of a channel. There are n channels, and each channel has m sampling points. Then the dimension of the input matrix X is m×n.

[0055] Perform time domain feature extraction, including but not limited to the average value of each channel signal, the variance of each channel signal, the root mean square value of each channel signal, the peak value of each channel signal, and the number of times the signal crosses the zero point, namely the zero crossing rate.

[0056] Frequency domain feature extraction is performed, including but not limited to the power distribution of the signal at different frequencies, namely the power spectral density (PSD), the center frequency of the signal, namely the frequency center, and the energy in a specific frequency band, namely the frequency band energy.

[0057] Center the input matrix X and remove the mean of each channel signal to make its mean zero:

[0058] X centered =X-μ X

[0059] Among them, X is the input electromyographic signal matrix with a dimension of m×n; μ X is the mean vector of the input matrix, with a dimension of 1×n; X centered is the centralized signal matrix.

[0060] The centered signal is whitened so that its components are independent of each other and have unit variance:

[0061] X whitened =V∧ -1 / 2 V T X centered

[0062] Among them, V is the eigenvector matrix with dimension n×n; ∧ is the eigenvalue diagonal matrix with dimension n×n; X whitened is the signal matrix after whitening;

[0063] Based on the lCA algorithm, independent components are extracted from the whitened signal, including:

[0064] Initialize the weight vector w;

[0065] Iteratively update the weight vector:

[0066]

[0067] Among them, w is the weight vector with dimension n×1; η is the learning rate, which controls the update step size; σ is the nonlinear activation function; σ′ is the derivative of the activation function σ; d is the adaptive scaling factor with dimension n×1; w new is the updated weight vector;

[0068] Orthogonalize and normalize the weight vector until convergence:

[0069]

[0070] Repeat the above steps for all independent components until the weight vectors of all independent components are found, and the weight vectors of all independent components constitute the separation matrix W.

[0071] Assume that the independent component matrix is ​​S, then:

[0072] S=WX whitened

[0073] In the embodiment of the present invention, the separation matrix W is a linear transformation matrix, which is used to transform the whitened signal matrix X whitened Converted into an independent component matrix S, the dimension is n×n, that is, the number of channels or independent components of the signal. The independent component matrix S contains the independent source signals separated from the electromyographic signal, each column represents an independent component, and each row represents a time point, that is, the dimension is m×n.

[0074] Furthermore, the features in the electromyographic signal are extracted by wavelet transform for further signal separation and extraction, specifically including:

[0075] Determine the wavelet function and the number of decomposition levels;

[0076] Perform wavelet decomposition on the signal of each channel and decompose the signal into sub-signals and approximation signals in different frequency bands:

[0077]

[0078] Among them, d j is the detail coefficient of the jth layer, with a dimension of m j ×n,a J is the approximation coefficient of the Jth layer, with dimension m J ×n;X wavelet is the signal matrix after wavelet decomposition.

[0079] Adaptive threshold processing is performed on the detail coefficients after wavelet decomposition to remove noise and enhance features:

[0080] d j，thresholded =soft(d j ,λ j )

[0081] Among them, soft is the soft threshold function, λ j is the adaptive threshold of the jth layer.

[0082] Perform multi-resolution analysis on the processed detail coefficients to extract features at different scales, including energy, mean value, and standard deviation:

[0083]

[0084] Where k is a discrete time point, N is the length of the detail coefficient, that is, the total number of discrete time points; Energy j is the energy of the detail coefficient of the jth layer, Mean j is the average value of the detail coefficient of the jth layer, Mean jis the standard deviation of the detail coefficients of the jth layer.

[0085] After the above operations, the features extracted by I CA and wavelet transform are fused to construct a feature matrix. The features are classified and identified using a machine learning model to obtain independent activity results of different muscle groups. The independent activity results of different muscle groups include but are not limited to muscle activity intensity, muscle activity spectrum, time series of muscle activity, coordinated activity between different muscle groups, and abnormal patterns in muscle activity.

[0086] In the embodiment of the present invention, the ICA algorithm and the wavelet transform are improved. Among them, the ICA algorithm includes introducing an adaptive scaling factor in the iterative process so that the algorithm can better adapt to different signal strengths and changes and improve the separation accuracy; using a nonlinear activation function (such as the ReLU function) to replace the traditional logarithmic cosh function to enhance the processing ability of complex signals, especially more effective when processing non-Gaussian distributed signals; by optimizing the update steps and introducing orthogonalization and normalization processing, the convergence speed and separation effect are improved. Adaptive threshold processing is introduced in the wavelet transform algorithm to dynamically adjust the threshold according to the signal characteristics to enhance the denoising effect while retaining the useful information of the signal; and by analyzing at different resolution levels, the characteristics of the signal in different frequency bands are captured to improve the separation and recognition capabilities of complex signals.

[0087] Based on the improved l CA algorithm and wavelet transform, the independent components of the electromyographic signal can be separated more effectively, making each component more independent and representative, and better removing noise, while retaining the key features of the signal, improving the quality of the signal and the accuracy of the analysis. Through the above operations, the independent activities of different muscle groups can be more accurately identified, so that each independent component corresponds to a specific muscle activity. Through better signal separation and feature extraction, the analysis accuracy of muscle activity is improved, providing more reliable data support for clinical diagnosis and rehabilitation.

[0088] Step S4: performing an evaluation operation on the swallowing function of the stroke patient based on the analysis results of the deep muscle electrical activity and the independent activity results of different muscle groups.

[0089] Specifically, a training data set including labeled normal swallowing and abnormal swallowing samples is constructed, and the evaluation model established based on the supervised learning algorithm is trained through the training data set. Feature extraction is performed based on the analysis results of deep muscle electrical activity and the independent activity results of different muscle groups, and the extracted features are used as the input of the evaluation model. The swallowing function status of stroke patients is output through the evaluation model.

[0090] Furthermore, the number of electrode channels of the multi-channel sEMG device is ≥16.

[0091] As a preferred embodiment of the present invention, before the first electromyographic signal is processed and analyzed, the first electromyographic signal is subjected to noise reduction processing by using an adaptive filtering and artifact elimination algorithm.

[0092] Specifically, the adaptive filtering process is implemented based on the least mean square LMS filtering, including:

[0093] Set the initial weights and learning rate of the filter;

[0094] The first electromyographic signal is divided into several small segments, each of which can be regarded as a stable signal;

[0095] Iteratively update the adaptive filter for each signal segment and adjust the filter weights to minimize the noise;

[0096] The filtered results of each signal segment are spliced ​​into a complete noise reduction signal.

[0097] Furthermore, the motion artifacts (such as signal fluctuations caused by electrode movement) and power supply interference (such as power frequency noise) in the electromyographic signal are removed through the artifact elimination algorithm. Specifically, it includes:

[0098] The signal is high-pass filtered through a high-pass filter to remove low-frequency artifacts, preferably, the cut-off frequency is set to 20 Hz;

[0099] The signal is band-pass filtered by a band-pass filter to remove power supply interference and high-frequency noise. Preferably, the band-pass range is 20-450 Hz;

[0100] Based on the Daubechies wavelet algorithm, the signal is decomposed into wavelets, detail coefficients at different scales are extracted, and artifacts are detected and removed from the detail coefficients;

[0101] The signal is decomposed by ICA to obtain multiple independent components, the components related to artifacts are identified and removed, and the purified signal is reconstructed.

[0102] In the embodiment of the present invention, low-frequency noise, motion artifacts and power supply interference in the electromyographic signal can be effectively removed by adaptive filtering and artifact elimination algorithms. The adaptive filter dynamically adjusts the filtering parameters according to the statistical characteristics of the input signal, thereby improving the filtering effect; the artifact elimination algorithm ensures that the obtained signal is cleaner and more reliable, providing a high-quality input signal for subsequent processing and analysis, and ensuring that the final analysis result is more accurate.

[0103] Furthermore, as a preferred embodiment of the present invention, a muscle activity model based on anatomy and physiology is established to distinguish the electrical activity of specific muscles through simulation and calculation. The sEMG signal and muscle model are combined to more accurately evaluate the function of specific muscles.

[0104] Specifically, the steps include:

[0105] Collect human anatomical data through anatomy textbooks, 3D anatomical models, or medical imaging (such as MRI or CT), and collect muscle physiological properties such as muscle fiber type, maximum contraction force, contraction speed, and fatigue characteristics through experimental measurements;

[0106] Use anatomical data 3D modeling software (such as SolidWorks, Blender or Matlab) to build a 3D geometric anatomical model of the muscle; according to the physiological characteristics of the muscle, the mechanical model of the muscle is established through the Hill model and the Zajac model, including the force-length relationship, force-velocity relationship and fatigue characteristics of the muscle.

[0107] The anatomical model describes the anatomical structure of the muscle, such as its position, shape, volume, and tendon attachment points in the body, and provides the geometric shape and topological structure of the muscle, which is used to determine the starting and ending points, force points, and direction of action of the muscle in biomechanical simulations, and to define the movement path of the muscle. The mechanical model of the muscle is used to simulate the mechanical response of the muscle under different conditions and calculate the force and movement generated by the muscle.

[0108] Furthermore, an electrophysiological model of muscle is established based on the anatomical model, mechanical model and electrophysiological characteristics of muscle fibers. The electrophysiological model is used to simulate the propagation process of electrical activity of muscle fibers, calculate the distribution of electrical activity of muscles under different stimulation conditions; and superimpose the electrical activity of individual muscle fibers to synthesize the sEMG signal of the entire muscle.

[0109] The collected multi-channel sEMG electromyographic signals are extracted in terms of amplitude, frequency components, activity patterns, etc. The collected sEMG signals are compared and calibrated with the simulation results of the muscle model, and the model parameters are adjusted to more accurately reflect the actual muscle activity. Based on the comparison results, the maximum contraction force and average activity intensity of the muscles, the fatigue characteristics of the muscles during repeated contractions, and the coordination and synchronization between different muscle groups are evaluated, and the results are used as auxiliary data to evaluate the function of specific muscles.

[0110] Furthermore, before step S4, the method further includes:

[0111] Collecting electrocardiogram, blood oxygen saturation and clinical imaging data of the stroke patient, performing multimodal fusion analysis operation on the electrocardiogram, blood oxygen saturation, clinical imaging data and first electromyographic signal of the stroke patient, and obtaining fusion analysis results;

[0112] After obtaining the fusion analysis result, step S4 includes:

[0113] The swallowing function of stroke patients was evaluated based on the fusion analysis results, the analysis results of the deep muscle electrical activity of stroke patients, and the independent activity results of different muscle groups.

[0114] Furthermore, the method also includes obtaining the electromyographic signals of the stroke patient within a preset time period and the corresponding evaluation results and expert diagnosis results, and establishing an abnormal feedback mechanism based on the electromyographic signals of the stroke patient within a preset time period and the corresponding evaluation results and expert diagnosis results.

[0115] Embodiment 2

[0116] The second aspect of the present invention discloses a swallowing function monitoring and evaluation system for stroke patients, please refer to Figure 2 , Figure 2 It is a structural schematic diagram of a swallowing function monitoring and evaluation system for stroke patients disclosed in another embodiment of the present invention, the system includes a multi-channel sEMG device, a second data processing unit, and an evaluation unit; wherein the multi-channel sEMG device also includes a first acquisition unit, a first storage unit, and a first data processing unit.

[0117] The multi-channel sEMG device is used to provide multi-channel sEMG electrodes, and allows a user to place the multi-channel sEMG electrodes on the surface of target swallowing muscles of a stroke patient;

[0118] The first acquisition unit is used to acquire the electromyographic signal through the multi-channel sEMG electrode to obtain the first electromyographic signal;

[0119] The first storage unit is used to store the first electromyographic signal;

[0120] The first data processing unit is used to process and analyze the first electromyographic signal to obtain an analysis result of deep muscle electrical activity;

[0121] The second data processing unit is used to perform separation and extraction operations on the first electromyographic signal through a machine learning algorithm to obtain independent activity results of different muscle groups;

[0122] The evaluation unit is used to perform evaluation operations on the swallowing function of stroke patients based on the analysis results of deep muscle electrical activity and the independent activity results of different muscle groups.

[0123] Further, electrode placement sites include the surface muscles of the larynx, oral cavity, and upper esophagus.

[0124] Furthermore, the number of electrode channels of the multi-channel sEMG device is ≥16.

[0125] Furthermore, the second data processing unit is also used to perform noise reduction processing on the first electromyographic signal through adaptive filtering and artifact elimination algorithm.

[0126] Further, the machine learning algorithms include independent component analysis and wavelet transform.

[0127] Furthermore, the system also includes a second acquisition unit, which is used to acquire electrocardiogram, blood oxygen saturation and clinical imaging data of the stroke patient;

[0128] The second data processing unit is also used to perform a multi-modal fusion analysis operation on the electrocardiogram, blood oxygen saturation, clinical imaging data and the first electromyographic signal of the stroke patient to obtain a fusion analysis result;

[0129] After obtaining the fusion analysis results, the swallowing function of the stroke patient is evaluated by the evaluation unit based on the fusion analysis results, the analysis results of the deep muscle electrical activity of the stroke patient, and the independent activity results of different muscle groups.

[0130] Furthermore, the system also includes a feedback unit, which is used to obtain the electromyographic signals of the stroke patient within a preset time period and the corresponding evaluation results and expert diagnosis results, and establish an abnormal feedback mechanism based on the electromyographic signals of the stroke patient within a preset time period and the corresponding evaluation results and expert diagnosis results.

[0131] It should be noted that the specific implementation process of the second embodiment is similar to that of the first embodiment and will not be repeated in the second embodiment.

[0132] Finally, it should be noted that the swallowing function monitoring and evaluation method and system for stroke patients disclosed in the embodiment of the present invention only discloses the preferred embodiment of the present invention, which is only used to illustrate the technical solution of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, it should be understood by those skilled in the art that the technical solutions described in the aforementioned embodiments can still be modified, or some of the technical features therein can be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the various embodiments of the present invention.

Claims

1. A method for monitoring and evaluating swallowing function in stroke patients, characterized in that: The method comprises the following steps: Step S1, placing a multi-channel sEMG electrode on the surface of the target swallowing muscle of the stroke patient, collecting electromyographic signals through the electrodes, and obtaining a first electromyographic signal; Step S2, storing the first electromyographic signal in a multi-channel sEMG device, processing and analyzing the first electromyographic signal through the multi-channel sEMG device, and obtaining an analysis result of deep muscle electrical activity; Step S3, performing separation and extraction operations on the first electromyographic signal through a machine learning algorithm to obtain independent activity results of different muscle groups; Step S4: performing an evaluation operation on the swallowing function of the stroke patient based on the analysis results of the deep muscle electrical activity and the independent activity results of different muscle groups.

2. The method for monitoring and evaluating swallowing function of stroke patients according to claim 1, characterized in that: The electrodes are placed at locations including the surface muscles of the larynx, oral cavity, and upper esophagus; The number of electrode channels of the multi-channel sEMG device is ≥16.

3. The method for monitoring and evaluating swallowing function of stroke patients according to claim 1, characterized in that: In the step S2, before the first electromyographic signal is processed and analyzed by the multi-channel sEMG device, the method further includes performing noise reduction processing on the first electromyographic signal by using an adaptive filtering and artifact elimination algorithm.

4. The method for monitoring and evaluating swallowing function of stroke patients according to claim 3, characterized in that: In step S2, the first electromyographic signal is processed and analyzed by a multi-channel sEMG device to obtain the analysis result of the deep muscle electrical activity, which specifically includes: The preprocessed multi-channel electromyographic signals are constructed into a matrix, wherein each column represents the signal of one channel; The signal is converted from the time domain to the time-frequency domain by short-time Fourier transform (STFT), and the source signal is separated in the time-frequency domain by non-negative matrix decomposition algorithm, and the time-frequency matrix is ​​decomposed into two non-negative matrices, namely the basis matrix and the coefficient matrix; wherein the basis matrix represents the characteristic spectrum and the coefficient matrix represents the time activation; Convert the separated time-frequency representation back to the time domain signal, perform inverse STFT on each independent component, and obtain the separated time domain signal; Evaluate and validate the isolated components to determine the corresponding deep muscle activity; The muscle activity intensity is obtained based on the electrical activity intensity of the deep muscles at different time points. The intensity distribution of different frequency components is displayed based on the spectrum analysis results of the electromyographic signal, and the muscle fiber type and fatigue status are obtained. The timing of muscle activity is obtained based on the activity time series of the deep muscles during swallowing. The coordination and synchronization of swallowing movements are evaluated based on the coordinated activities between different deep muscles, and abnormal muscle activity patterns are identified.

5. The method for monitoring and evaluating swallowing function of stroke patients according to claim 1, characterized in that: The machine learning algorithm in step S3 includes independent component analysis and wavelet transform.

6. The method for monitoring and evaluating swallowing function of stroke patients according to claim 1, characterized in that: Before step S4, the method further includes: Collecting electrocardiogram, blood oxygen saturation and clinical imaging data of the stroke patient, performing multimodal fusion analysis operation on the electrocardiogram, blood oxygen saturation, clinical imaging data and first electromyographic signal of the stroke patient, and obtaining fusion analysis results; After obtaining the fusion analysis result, step S4 includes: The swallowing function of stroke patients was evaluated based on the fusion analysis results, the analysis results of the deep muscle electrical activity of stroke patients, and the independent activity results of different muscle groups.

7. The method for monitoring and evaluating swallowing function of stroke patients according to any one of claims 1 to 6, characterized in that: The method also includes obtaining the electromyographic signals of the stroke patient within a preset time period and the corresponding evaluation results and expert diagnosis results, and establishing an abnormal feedback mechanism based on the electromyographic signals of the stroke patient within the preset time period and the corresponding evaluation results and expert diagnosis results.

8. A swallowing function monitoring and evaluation system for stroke patients, the system is implemented based on the method according to any one of claims 1 to 7, characterized in that: The system comprises a multi-channel sEMG device, a second data processing unit, and an evaluation unit; wherein the multi-channel sEMG device further comprises a first acquisition unit, a first storage unit, and a first data processing unit; The multi-channel sEMG device is used to provide multi-channel sEMG electrodes, and allows the user to place the multi-channel sEMG electrodes on the surface of the target swallowing muscles of the stroke patient; The first acquisition unit is used to acquire the electromyographic signal through the multi-channel sEMG electrode to obtain the first electromyographic signal; The first storage unit is used to store a first electromyographic signal; The first data processing unit is used to process and analyze the first electromyographic signal to obtain an analysis result of deep muscle electrical activity; The second data processing unit is used to perform separation and extraction operations on the first electromyographic signal through a machine learning algorithm to obtain independent activity results of different muscle groups; The evaluation unit is used to perform an evaluation operation on the swallowing function of a stroke patient based on the analysis results of deep muscle electrical activity and the independent activity results of different muscle groups.

9. The swallowing function monitoring and evaluation system for stroke patients according to claim 8, characterized in that: The system further comprises a second acquisition unit, which is used to acquire electrocardiogram, blood oxygen saturation and clinical imaging data of the stroke patient; The second data processing unit is further used to perform a multimodal fusion analysis operation on the electrocardiogram, blood oxygen saturation, clinical imaging data and the first electromyographic signal of the stroke patient to obtain a fusion analysis result; After obtaining the fusion analysis results, the swallowing function of the stroke patient is evaluated by the evaluation unit based on the fusion analysis results, the analysis results of the deep muscle electrical activity of the stroke patient, and the independent activity results of different muscle groups.

10. The swallowing function monitoring and evaluation system for stroke patients according to claim 9, characterized in that: The system also includes a feedback unit, which is used to obtain the electromyographic signals of the stroke patient within a preset time period and the corresponding evaluation results and expert diagnosis results, and establish an abnormal feedback mechanism based on the electromyographic signals of the stroke patient within the preset time period and the corresponding evaluation results and expert diagnosis results.

Images