Disclosure of Invention
The embodiment of the application provides a respiratory neuromuscular stimulation control method and system, which aim to solve the technical problems.
The application provides a respiratory neuromuscular stimulation control method, which comprises the following steps:
determining the stimulation parameters of the user according to the body parameter information and parameters stored in a database;
Before the user body is electrically stimulated, an initial muscle contraction signal obtained from a treatment terminal is processed through a signal processing module to obtain an initial muscle contraction filtering signal, wherein the initial muscle contraction signal is perceived by the treatment terminal attached to the user body;
According to the stimulation parameters of the user, the initial muscle contraction filtering signals, the real-time muscle contraction filtering signals and the real-time breathing conditions, the current of the electric stimulation generated by the treatment terminal arranged on the body of the user is controlled in real time through a self-adaptive current adjustment model which is completed through training optimization;
Wherein the real-time breathing condition comprises the inspiration time and the breathing frequency of the user, and the electric current of the electric stimulus acts on the nerve or the muscle of the user, the muscle comprises a single muscle group or a multi-muscle group, and the nerve comprises the phrenic nerve.
Further, the adaptive current adjustment model is constructed and train optimized based on the following manner:
Acquiring historical user stimulation data, historical muscle contraction signals, historical respiration data and real-time electric stimulation current data, wherein the historical muscle contraction signals comprise initial muscle contraction signals and real-time stimulation muscle contraction signals, and the historical respiration data comprise initial respiration data and real-time stimulation respiration data;
determining a set of stimulus factors for the historical user stimulus data;
Determining weight vectors and membership vectors of all the stimulus factors in the stimulus factor set;
Performing fuzzy transformation on the weight vector and the membership vector of each stimulus factor to obtain a first fuzzy judgment matrix;
Performing fuzzy operation and normalization on the first fuzzy judgment matrix and the weight vector of each stimulus factor to obtain a second fuzzy judgment matrix;
The deep neural network is constructed according to the second fuzzy evaluation matrix and comprises an input layer, a basic network and an adaptive weight layer, wherein the basic network comprises six convolution layers, three full-connection layers and a pooling layer, the pooling layer is connected behind the convolution layers and is used for polymerizing adjacent area values, the three full-connection layers comprise a first full-connection layer, a second full-connection layer and a third full-connection layer, and the adaptive weight layer is used for completing adaptive adjustment of current parameters through back propagation;
Performing feature extraction on the initial respiratory data and the real-time stimulated respiratory data to obtain an initial respiratory feature vector and a real-time stimulated respiratory feature vector;
calculating a distance between the initial respiratory feature vector and the real-time stimulated respiratory feature vector;
And carrying out optimization training on the depth neural network obtained by construction according to the distance between the initial respiration characteristic vector and the real-time stimulation respiration characteristic vector, an initial muscle contraction signal, a real-time stimulation muscle contraction signal, the initial respiration characteristic vector, the real-time stimulation respiration characteristic vector and the real-time electrical stimulation current data until a loss function converges.
The method comprises the steps of constructing a depth neural network, wherein the depth neural network comprises a real-time electric stimulation current data, a distance between an initial respiration characteristic vector and a real-time stimulation respiration characteristic vector, an initial muscle contraction signal, a real-time stimulation muscle contraction signal, an initial respiration characteristic vector, a real-time stimulation respiration characteristic vector and a real-time electric stimulation current data, wherein the real-time electric stimulation current data comprises the stimulation intensity, the pulse width, the pulse frequency and the stimulation duration of the current, and the method comprises the steps of optimally training the constructed depth neural network until a loss function converges, wherein the method comprises the following steps:
constructing a time sequence sample vector according to the distance between the initial respiratory feature vector and the real-time stimulation respiratory feature vector, the initial respiratory feature vector and the real-time stimulation respiratory feature vector:
inputting the time sequence sample vector into an input layer of the deep neural network, and acquiring a respiration time sequence output data set output by the first full-connection layer;
Performing non-negative matrix factorization on the initial muscle contraction signal and the real-time stimulation muscle contraction signal based on the time sequence sample vector to obtain a signal non-negative matrix factorization set;
Performing window processing on the signal non-negative matrix factorization set to obtain a time domain sample;
Quantifying a first correlation quantity of the frequency domain sample and the real-time electric stimulation current data in an information entropy and mutual information mode, and determining a first self-adaptive loss function according to a preset first self-adaptive adjustment function in the self-adaptive weight layer;
quantizing a second correlation quantity between the frequency domain sample and the respiration time sequence output data set in an information entropy and mutual information mode, and determining a second self-adaptive loss function according to a second self-adaptive adjustment function preset in the self-adaptive weight layer;
And constructing the loss function according to the first adaptive loss function and the second adaptive loss function, and performing optimization training on the deep neural network until the loss function converges.
Further, the physical parameter information comprises age, joint mobility, muscle atrophy, spasticity parameters and lung function parameters, wherein the lung function parameters comprise lung capacity, forced lung capacity and forced breathing capacity for one second.
Further, the treatment terminal comprises a plurality of treatment electrodes, wherein the plurality of treatment electrodes comprise electrodes attached to the rear edges of left and right sternocleidomastoid muscles, left and right subclavian pectoral major muscles, oblique abdominal muscles, rectus abdominus muscles, biceps femoris muscles, anterior tibial muscle groups, total fibular nerves, anterior tibial muscles, long fibular muscles, quadriceps femoris, hip abductor muscle groups, and extensor carpi and extensor digitorum longus groups.
The application provides a respiratory nerve muscle stimulation control system, which comprises a user stimulation parameter determining unit, an initial muscle contraction filtering signal determining unit and an electric stimulation current real-time control unit,
The user stimulation parameter determining unit is used for acquiring physical parameter information of a user, and determining the stimulation parameter of the user according to the physical parameter information and parameters stored in a database;
The initial muscle contraction filtering signal determining unit is used for processing an initial muscle contraction signal acquired from a treatment terminal through the signal processing module before the user body is electrically stimulated to obtain an initial muscle contraction filtering signal, wherein the initial muscle contraction signal is perceived by the treatment terminal attached to the user body;
The electric stimulation current real-time control unit is used for controlling the electric stimulation current generated by the treatment terminal arranged on the body of the user in real time through a self-adaptive current adjustment model which is completed through training and optimization according to the stimulation parameters of the user, the initial muscle contraction filtering signal, the real-time muscle contraction filtering signal and the real-time respiration condition, wherein the real-time respiration condition comprises the inspiration time and the respiration frequency of the user, the electric stimulation current acts on the nerve or the muscle of the user, the muscle comprises a single muscle group or a multiple muscle group, and the nerve comprises a diaphragmatic nerve.
According to the embodiment of the application, body parameter information of a user is acquired, stimulation parameters of the user are determined according to the body parameter information and parameters stored in a database, an initial muscle contraction signal acquired from a treatment terminal is processed through a signal processing module before the body of the user is electrically stimulated to obtain an initial muscle contraction filtering signal, the initial muscle contraction signal is sensed by the treatment terminal attached to the body of the user, and an adaptive current adjustment model which is completed through training optimization is used for controlling current of electrical stimulation generated by the treatment terminal arranged on the body of the user in real time according to the stimulation parameters of the user, the initial muscle contraction filtering signal, the real-time muscle contraction filtering signal and real-time breathing conditions, wherein the real-time breathing conditions comprise inspiration time and breathing frequency of the user, the current of the electrical stimulation acts on nerves or muscles of the user, the muscles comprise single muscle groups or multiple muscle groups, and the nerves comprise diaphragm nerves. Therefore, the current data of the electric stimulation equipment are timely adjusted according to the body condition of the user and the response to the electric stimulation, and the action time of the electric stimulation equipment is not adjusted before the treatment is started singly according to the experience of medical staff, so that the efficiency and the accuracy of the electric stimulation are improved.
The method has the advantages of being beneficial to personalized treatment, being capable of providing customized treatment plans for each user by collecting detailed body data of the individual, improving treatment effects, being beneficial to more accurately evaluating treatment requirements and adjusting treatment schemes by accurate body parameter information, improving safety, being capable of preventing potential treatment risks by knowing body conditions of the user, ensuring treatment safety, being beneficial to improving treatment effects by determining stimulation parameters of the user according to body parameter information and parameters stored in a database, being capable of optimizing treatment decisions, combining historical data and real-time parameters, being capable of more scientifically determining stimulation parameters, being capable of adjusting stimulation parameters dynamically according to changes of the user conditions, being capable of adapting to changes of the treatment effects and the user comfort, being capable of processing initial muscle contraction signals by a signal processing module, being capable of removing noise and interference by means of removing more accurate muscle contraction signals, being capable of accurately controlling electric stimulation currents, being beneficial to improving treatment effects by accurate control, being capable of controlling initial and muscle contraction signals, being used for carrying out control, being capable of adjusting the electric stimulation parameters, being capable of being adjusted in real time, being matched with the breathing control parameters, being capable of achieving real-time, being adjusted according to changes of the user conditions, being capable of adjusting the breathing control parameters, being capable of being adjusted to be capable of achieving breathing effects, and being adjusted in real-time, and being adjusted to the breathing control parameters, being adjusted to be capable of breathing parameters are adjusted to be appropriate to the user conditions, the method can customize more personalized schemes for different users, the electric stimulation current acts on nerves or muscles of the users, targeted treatment is carried out, the electric stimulation can directly act on target nerves or muscles to provide targeted treatment, muscle functions are enhanced, muscle strength and functions can be enhanced by stimulating specific muscle groups, muscle atrophy or dysfunction is improved, the electric stimulation current generated by a treatment terminal is controlled in real time, accurate dosage is controlled in real time, the user can be ensured to receive treatment with proper intensity, excessive or insufficient stimulation is avoided, the curative effect is improved, the accurate current control is beneficial to improving the treatment effect, side effects are reduced, and the overall treatment experience of the users is improved.
Detailed Description
In order to enable those skilled in the art to better understand the present application, the following description will make clear and complete descriptions of the technical solutions according to the embodiments of the present application with reference to the accompanying drawings.
Optionally, as shown in fig. 1, the present application provides a respiratory neuromuscular stimulation control method, including:
S101, acquiring physical parameter information of a user, and determining stimulation parameters of the user according to the physical parameter information and parameters stored in a database;
S102, before the user body is electrically stimulated, processing an initial muscle contraction signal acquired from a treatment terminal through a signal processing module to obtain an initial muscle contraction filtering signal, wherein the initial muscle contraction signal is perceived by the treatment terminal attached to the user body;
And S103, performing real-time control on electric stimulation current generated by a treatment terminal arranged on the body of the user through a self-adaptive current adjustment model which is completed through training according to the stimulation parameters, the initial muscle contraction filtering signals, the real-time muscle contraction filtering signals and the real-time respiration conditions of the user, wherein the real-time respiration conditions comprise the inspiration time and the respiration frequency of the user, the electric stimulation current acts on the nerve or the muscle of the user, the muscle comprises a single muscle group or a multiple muscle group, and the nerve comprises the phrenic nerve.
According to the embodiment of the application, body parameter information of a user is acquired, stimulation parameters of the user are determined according to the body parameter information and parameters stored in a database, an initial muscle contraction signal acquired from a treatment terminal is processed through a signal processing module before the body of the user is electrically stimulated to obtain an initial muscle contraction filtering signal, the initial muscle contraction signal is sensed by the treatment terminal attached to the body of the user, and an adaptive current adjustment model which is completed through training optimization is used for controlling current of electrical stimulation generated by the treatment terminal arranged on the body of the user in real time according to the stimulation parameters of the user, the initial muscle contraction filtering signal, the real-time muscle contraction filtering signal and real-time breathing conditions, wherein the real-time breathing conditions comprise inspiration time and breathing frequency of the user, the current of the electrical stimulation acts on nerves or muscles of the user, the muscles comprise single muscle groups or multiple muscle groups, and the nerves comprise diaphragm nerves. Therefore, the electric stimulation equipment is timely adjusted according to the body condition of the user and the response to the electric stimulation, and the efficiency and the accuracy of the electric stimulation are improved.
The method has the advantages of being beneficial to personalized treatment, being capable of providing customized treatment plans for each user by collecting detailed body data of the individual, improving treatment effects, being beneficial to more accurately evaluating treatment requirements and adjusting treatment schemes by accurate body parameter information, improving safety, being capable of preventing potential treatment risks by knowing body conditions of the user, ensuring treatment safety, being beneficial to improving treatment effects by determining stimulation parameters of the user according to body parameter information and parameters stored in a database, being capable of optimizing treatment decisions, combining historical data and real-time parameters, being capable of more scientifically determining stimulation parameters, being capable of adjusting stimulation parameters dynamically according to changes of the user conditions, being capable of adapting to changes of the treatment effects and the user comfort, being capable of processing initial muscle contraction signals by a signal processing module, being capable of removing noise and interference by means of removing more accurate muscle contraction signals, being capable of accurately controlling electric stimulation currents, being beneficial to improving treatment effects by accurate control, being capable of controlling initial and muscle contraction signals, being used for carrying out control, being capable of adjusting the electric stimulation parameters, being capable of being adjusted in real time, being matched with the breathing control parameters, being capable of achieving real-time, being adjusted according to changes of the user conditions, being capable of adjusting the breathing control parameters, being capable of being adjusted to be capable of achieving breathing effects, and being adjusted in real-time, and being adjusted to the breathing control parameters, being adjusted to be capable of breathing parameters are adjusted to be appropriate to the user conditions, the method can customize more personalized schemes for different users, the electric stimulation current acts on nerves or muscles of the users, targeted treatment is carried out, the electric stimulation can directly act on target nerves or muscles to provide targeted treatment, muscle functions are enhanced, muscle strength and functions can be enhanced by stimulating specific muscle groups, muscle atrophy or dysfunction is improved, the electric stimulation current generated by a treatment terminal is controlled in real time, accurate dosage is controlled in real time, the user can be ensured to receive treatment with proper intensity, excessive or insufficient stimulation is avoided, the curative effect is improved, the accurate current control is beneficial to improving the treatment effect, side effects are reduced, and the overall treatment experience of the users is improved.
Further, the adaptive current adjustment model is built and optimized based on the following manner:
Acquiring historical user stimulation data, historical muscle contraction signals, historical respiration data and real-time electric stimulation current data, wherein the historical muscle contraction signals comprise initial muscle contraction signals and real-time stimulation muscle contraction signals;
determining a set of stimulus factors for the historical user stimulus data;
determining weight vectors and membership vectors of all the stimulus factors in the stimulus factor set;
performing fuzzy transformation on the weight vector and the membership vector of each stimulus factor to obtain a first fuzzy judgment matrix;
Performing fuzzy operation and normalization on the first fuzzy judgment matrix and the weight vector of each stimulus factor to obtain a second fuzzy judgment matrix;
The method comprises the steps of constructing a deep neural network according to a second fuzzy evaluation matrix, wherein the deep neural network comprises an input layer, a basic network and an adaptive weight layer, the basic network comprises six convolution layers, three full-connection layers and a pooling layer, the pooling layer is connected with the convolution layers and is used for aggregating adjacent area values, the three full-connection layers comprise a first full-connection layer, a second full-connection layer and a third full-connection layer, and the adaptive weight layer is used for completing adaptive adjustment of current parameters through back propagation;
Extracting features of the initial respiration data and the real-time stimulated respiration data to obtain an initial respiration feature vector and a real-time stimulated respiration feature vector;
Calculating a distance between the initial respiration feature vector and the real-time stimulation respiration feature vector;
And carrying out optimization training on the depth neural network obtained by construction according to the distance between the initial respiration characteristic vector and the real-time stimulation respiration characteristic vector, the initial muscle contraction signal, the real-time stimulation muscle contraction signal, the initial respiration characteristic vector, the real-time stimulation respiration characteristic vector and the real-time electric stimulation current data until the loss function converges.
The embodiment provided by the application has the beneficial effects that historical and real-time data are acquired, a self-adaptive current adjustment model can be constructed and optimized more accurately by analyzing the historical and real-time data, the generalization capability of the model is enhanced, the model can adapt to the changes of different users and different time points by combining the historical and real-time data, a stimulation factor set is determined, the comprehensiveness of the model is ensured, the prediction accuracy of treatment effect is improved by considering all relevant stimulation factors, customized treatment is performed, key stimulation factors are identified, customized treatment parameters are provided for each user, weight vectors and membership vectors are determined, the quantization influence is determined, the influence of different stimulation factors is quantized by the weight and membership, the model is more accurate, priority ordering is clear, the factors are more important, the higher priority is given to the model when the model is optimized, fuzzy conversion is performed, the fuzzy logic can process the uncertainty and the ambiguity in the data, the robustness of the model is improved, the model is allowed to make effective decisions when the model is still inaccurate or partial information is met, the self-adaptive weight vectors are identified, the key vectors can be better understood, the accuracy is better, the key characteristics can be better tracked, the user can be better calculated, and the accuracy is better calculated, and the user can track the characteristics are better.
Further, the real-time electrical stimulation current data comprises stimulation intensity, pulse width, pulse frequency and stimulation duration of the current, and the method comprises the steps of optimally training the constructed deep neural network until a loss function converges according to the distance between the initial respiration characteristic vector and the real-time stimulation respiration characteristic vector, the initial muscle contraction signal, the real-time stimulation muscle contraction signal, the initial respiration characteristic vector, the real-time stimulation respiration characteristic vector and the real-time electrical stimulation current data, and comprises the following steps:
Constructing a time sequence sample vector according to the distance between the initial respiration feature vector and the real-time stimulation respiration feature vector, the initial respiration feature vector and the real-time stimulation respiration feature vector:
Inputting the time sequence sample vector into an input layer of the deep neural network, and acquiring a respiration time sequence output data set output by a first full-connection layer;
performing non-negative matrix factorization on the initial muscle contraction signal and the real-time stimulation muscle contraction signal based on the time sequence sample vector to obtain a signal non-negative matrix factorization set;
Performing window processing on the non-negative matrix factorization set of the signal to obtain a time domain sample;
The method comprises the steps of quantizing a first correlation quantity of a frequency domain sample and real-time electric stimulation current data in an information entropy and mutual information mode, and determining a first self-adaptive loss function according to a preset first self-adaptive adjustment function in a self-adaptive weight layer;
Wherein,The method comprises the steps of obtaining a first adaptive loss function, obtaining the number of layers of an adaptive weight layer by U, obtaining the number of the adaptive weight layer by U, obtaining a first strategy of a U-th adaptive weight layer by Strat 1.1u, obtaining a first adaptive adjustment function by Ajust1 (), obtaining a regularization parameter by alpha 1, obtaining a natural constant by e, and obtaining a first correlation value by relev1;
quantizing a second correlation quantity of the frequency domain sample and the respiration time sequence output data set in an information entropy and mutual information mode, and determining a second self-adaptive loss function according to a second self-adaptive adjustment function preset in the self-adaptive weight layer;
and constructing a loss function according to the first self-adaptive loss function and the second self-adaptive loss function, and performing optimization training on the deep neural network until the loss function converges.
Based on the embodiment provided by the application, the method has the beneficial effects that the time sequence sample vector is constructed, and a comprehensive time sequence sample vector can be created by combining initial and real-time respiration and muscle contraction signals, so that the deep neural network is helped to capture the dynamic change of the physiological state of the user; the deep neural network is input, a time sequence sample vector is input into the network, so that the network can receive abundant physiological state information from an input layer, a foundation is laid for subsequent feature learning and state prediction, a first full-connection layer output data set can provide initial breathing mode features for the network, the initial and real-time stimulation muscle contraction signals are subjected to non-negative matrix decomposition, key features of muscle activity can be extracted, the internal structure of the muscle activity is facilitated to be understood, window processing and short-time Fourier transformation, time domain samples can be converted into frequency domain samples through window processing and short-time Fourier transformation, the frequency characteristics of the muscle activity can be analyzed, basis is provided for frequency adjustment of electric stimulation, information entropy and mutual information quantization are utilized, the relation between different physiological signals and electric stimulation effects can be evaluated by utilizing the information entropy and mutual information quantization frequency domain samples, guidance is provided for optimizing electric stimulation parameters, self-adaptive loss function determination is performed, the self-adaptive loss function is determined according to a first correlation quantity, the self-adaptive loss is enabled to be self-adaptive to a second adaptive loss model, the self-adaptive loss is enabled to be adjusted to a constant training process, the method comprises the steps of carrying out network optimization training by combining a first self-adaptive loss function and a second self-adaptive loss function, improving the robustness of a model, carrying out network optimization training by combining the first self-adaptive loss function and the second self-adaptive loss function, ensuring that the network is continuously self-adjusted in the training process so as to adapt to the change of the physiological state of a user until the loss function converges, carrying out real-time current control, namely carrying out real-time current control on the electric stimulation, so that the electric stimulation can be dynamically adjusted according to the real-time physiological state of the user, and improving the treatment effect and the comfort level of the user.
The steps together form a closed-loop self-adaptive current adjustment system which can be intelligently adjusted according to the real-time physiological data of the user so as to realize personalized and optimized electrical stimulation treatment. By this method, the effectiveness and safety of the electrical stimulation therapy can be significantly improved.
Further, the method comprises the steps of,
Wherein,The method is characterized by comprising the steps of obtaining a second adaptive loss function, obtaining the number of layers of an adaptive weight layer by U, obtaining the number of the adaptive weight layer by U, obtaining a second strategy of the adaptive weight layer of a U-th layer by Strat & lt 2 & gtu & gtand obtaining a second adaptive adjustment function by Ajust & lt 2 & gtand a regularization parameter by alpha & lt 2 & gt, obtaining a natural constant by e and obtaining a second correlation value by relev2.
Further, the sample point distance is calculated based on the following formula:
Wherein distan (X, Y) represents the distance between the initial respiratory feature vector X and the real-time stimulation respiratory feature vector Y, m is the dimension of the initial respiratory feature vector, n is the dimension of the real-time stimulation respiratory feature vector, Xi represents the coordinate value of the initial respiratory feature vector X in the dimension with the number i, Yj is the coordinate value of the real-time stimulation respiratory feature vector Y in the dimension with the number j, and t is the controllable variable parameter.
Further, the body parameter information includes age, joint mobility, muscle atrophy, spasticity parameters, and lung function parameters, wherein the lung function parameters include lung capacity, forced lung capacity, and forced breathing capacity for one second.
Further, the treatment terminal comprises a plurality of treatment electrodes, wherein the plurality of treatment electrodes comprise electrodes which are attached to the rear edges of left and right sternocleidomastoid muscles, left and right subclavian pectoral large muscles, oblique abdominal muscles, rectus abdominus muscles, quadriceps biceps, tibialis anterior muscle groups, common fibular nerve, tibialis anterior muscle, long fibular longus, quadriceps femoris, hip abductor muscle groups, and extensor carpi and extensor digitorum.
Alternatively, as shown in fig. 2, the present application provides a respiratory neuromuscular stimulation control system comprising a user stimulation parameter determining unit 201, an initial muscle contraction filtering signal determining unit 202 and an electrical stimulation current real-time control unit 203, wherein,
A user stimulation parameter determining unit 201 for acquiring physical parameter information of the user, determining stimulation parameters of the user according to the physical parameter information and parameters stored in a database;
An initial muscle contraction filtering signal determining unit 202, configured to process, by using a signal processing module, an initial muscle contraction signal obtained from a treatment terminal before electrically stimulating a user's body, to obtain an initial muscle contraction filtering signal, where the initial muscle contraction signal is perceived by the treatment terminal attached to the user's body;
The electric stimulation current real-time control unit 203 is configured to perform real-time control on an electric stimulation current generated by a treatment terminal disposed on a user's body through an adaptive current adjustment model that is completed through training and optimization according to a user's stimulation parameter, an initial muscle contraction filtering signal, a real-time muscle contraction filtering signal, and a real-time respiration condition, where the real-time respiration condition includes an inhalation time and a respiration frequency of the user, and the electric stimulation current acts on a nerve or a muscle of the user, where the muscle includes a single muscle group or a multiple muscle group, and the nerve includes a phrenic nerve.
Further, the body parameter information includes age, joint mobility, muscle atrophy, spasticity parameters, and lung function parameters, wherein the lung function parameters include lung capacity, forced lung capacity, and forced breathing capacity for one second.
Further, the treatment terminal comprises a plurality of treatment electrodes, wherein the plurality of treatment electrodes comprise electrodes which are attached to the rear edges of left and right sternocleidomastoid muscles, left and right subclavian pectoral large muscles, oblique abdominal muscles, rectus abdominus muscles, quadriceps biceps, tibialis anterior muscle groups, common fibular nerve, tibialis anterior muscle, long fibular longus, quadriceps femoris, hip abductor muscle groups, and extensor carpi and extensor digitorum.
The foregoing description is only of the preferred embodiments of the present invention, and is not intended to limit the scope of the invention, but rather is intended to cover any equivalents of the structures or equivalent processes disclosed herein or in the alternative, which may be employed directly or indirectly in other related arts.