CN112370340B

CN112370340B - Implementation method of foot stimulation training device based on machine learning

Info

Publication number: CN112370340B
Application number: CN202011247500.7A
Authority: CN
Inventors: 屈云
Original assignee: West China Hospital of Sichuan University
Current assignee: West China Hospital of Sichuan University
Priority date: 2020-11-10
Filing date: 2020-11-10
Publication date: 2023-03-14
Anticipated expiration: 2040-11-10
Also published as: CN112370340A

Abstract

The invention discloses a method for realizing a foot stimulation training device based on machine learning, which comprises a foot sleeving device, a plurality of stimulation components, a control device, a wireless module and a remote terminal, wherein the foot sleeving device is used for being fixed on a foot; the implementation method comprises the following steps: establishing a database; (2) preprocessing the original data; (3) missing data processing; (4) developing a neural network model; and (5) generating a prediction guide for foot stimulation training. The invention adopts a machine learning mode to establish a corresponding neural network model, and utilizes the model to predict the stimulation training recovery degree under different foot nerve injury conditions, thereby providing guidance for a doctor to formulate an optimal training plan. Therefore, the invention is suitable for popularization and application.

Description

Implementation method of foot stimulation training device based on machine learning

Technical Field

The invention relates to the technical field of medical instruments, in particular to a method for realizing a foot stimulation training device based on machine learning.

Background

Foot acupuncture therapy (foot acupunture), also known as foot acupuncture. The foot acupuncture therapy method is formed on the basis of Chinese traditional acupuncture and combined with the cattle holographic theory rule, and stimulates specific areas and acupuncture points of feet by acupuncture and other methods to treat diseases. Through years of practice, the electric needle is applied to foot needle stimulation treatment of diseases in recent times, more experience is accumulated on the positioning, the adaptation range and the stimulation method of foot needle points, the foot needle becomes one of the common treatment methods of some clinicians, and the electric needle achieves obvious curative effect on treating a large number of children patients. The foot needle system has wide treatment range, simple and convenient syndrome differentiation and acupoint selection and is scientific. The foot acupuncture treatment can obtain the curative effect, is easy to operate and is easily accepted by patients and extremely family members.

In addition, direct Current Stimulation (DCS) is a non-invasive technique for regulating neuronal activity using constant, low-intensity dc electrical potentials. Attempts to treat diseases by electricity have been made as early as the 11 th century, and the technology of direct current stimulation has matured gradually with the development of knowledge. The DCS has two different electrodes and power supply battery equipment thereof, and is additionally provided with a control software for setting the output of the stimulation type. The stimulation modes include 3 types, namely anodal stimulation, cathodal stimulation and pseudo stimulation. Anodal stimulation generally enhances the excitability of the nerve at the stimulation site, while cathodal stimulation reduces the excitability of the nerve at the stimulation site. DCS does not cause nerve endings to discharge by suprathreshold stimulation, but rather acts by modulating the activity of the neural network.

Therefore, the existing foot stimulation training scheme is often added with direct current stimulation so as to achieve the purpose of good auxiliary treatment. However, the existing foot stimulation training scheme is often poor in training effect, and finally is difficult to achieve the expected auxiliary treatment purpose, mainly because of the lack of reference data, the training result is difficult to predict or the deviation between the prediction and the actual result is large, which brings great difficulty for a doctor to make a training plan. If the corresponding stimulation training recovery degree can be predicted according to different foot nerve injury conditions, so that guidance is provided for a doctor to make an optimal training plan, the effect of foot stimulation training is greatly promoted.

Disclosure of Invention

The invention aims to provide a method for realizing a foot stimulation training device based on machine learning, which can better predict the recovery degree of corresponding stimulation training according to different foot nerve injury conditions and provide guidance for a doctor to formulate an optimal training plan.

In order to realize the purpose, the technical scheme adopted by the invention is as follows:

the foot stimulation training device comprises a foot sleeving device for fixing on the foot, a plurality of stimulation components which are arranged in the foot sleeving device and distributed according to foot acupuncture points, a control device connected with all magnetic pole components, a wireless module connected with the control device, and a remote terminal connected with the wireless module;

the implementation method comprises the following steps:

(1) Collecting previous foot stimulation training reports, obtaining related original data, and transmitting the data to a database of a remote terminal;

(2) Assigning initial values to missing data in the database based on existing biological interpretations;

(3) Setting the initial value assigned to the missing data as the average value of the values existing in the database, then recursively applying the following equations until convergence, and finally returning the result of convergence to obtain the predicted value of the missing data:

a ^(s+1) ＝(1-η)f(a ^s )+ηa ^s

wherein, a ^(s+1) Representing a predicted value, and s represents the iteration number; f (a) ^s ) Representing a prediction about a obtained from a neural network; eta is less than 1; the function f remains fixed in each loop iteration;

(4) Establishing a foot neural network model according to the predicted values of the original data and the missing data;

(5) Using the generated model to train a neural network while recording the model and evaluating the performance of the prediction, wherein the prediction with less uncertainty is used to generate a prediction guideline for foot stimulation training, including an expected best training result; the generated prediction guide is used as a reference for making a training plan so as to control the stimulation component to perform stimulation training on the foot according to the training plan by using the control device.

Preferably, the foot cover device has the same shape as any one of a foot cover, a sock, a shoe and a foot clip.

Furthermore, an air cushion is arranged below the stimulation component in the foot sleeve device.

Preferably, the stimulation component is an electrode or an infrared lamp.

Preferably, in the step (3), η =0.4.

Compared with the prior art, the invention has the following beneficial effects:

(1) The invention sets a set of foot stimulation training device, establishes a corresponding neural network model by adopting a machine learning mode, predicts the stimulation training recovery degree under different foot nerve injury conditions by using the model, thereby providing guidance for a doctor to formulate an optimal training plan, and then can safely, long-time and foolproof perform stimulation training on the foot of a patient by combining the application of the foot stimulation training device according to the formulation of the optimal training plan and obtain a predictable training effect.

(2) An air cushion is arranged below the stimulating component in the foot sleeving device, so that the comfort degree of the device is improved, and the stimulating effect is better guaranteed by matching with the stimulating component.

(3) Based on the effective formulation of the optimal training plan, the designed training device has the advantages of simpler structure, lower cost and portability compared with the traditional training device, after the optimal training plan is formulated, the training plan is transmitted to the control device to be stored, and then part or all of the stimulation components can be automatically controlled by the control device, so that stimulation of different electrode (or infrared) stimulation sequences, different acupoint combinations and different time lengths is given to corresponding acupuncture points on the foot, ordered stimulation training is realized, and auxiliary treatment is completed.

(4) Moreover, for some patients who are not suitable for foot stimulation training, after the training effect is predicted by the method, the stimulation training scheme can be definitely eliminated, so that the condition of an illness is prevented from being delayed, and meanwhile, the unnecessary medical expenditure of the patients is saved.

(5) When missing data processing is carried out, the invention finally returns a convergence result instead of f (a) ^s ) On one hand, the convergence result simultaneously contains data internal information and a data external machine learning feedback result, which is equivalent to jointly predicting by using the correlation inside the data and the result of external neural network learning; on the other hand, the parameter η being less than 1 can prevent the prediction from fluctuating. Therefore, the invention can utilize all information in the database, and provides guarantee for obtaining more reliable models and improving the prediction quality.

Drawings

Fig. 1 is a schematic structural view of the foot stimulation training device of the invention.

Fig. 2 is a schematic flow chart of the implementation of the present invention.

Wherein, the part names corresponding to the reference numbers are:

the device comprises a foot sleeving device, a stimulation component, a control device, a wireless module and a remote terminal, wherein the foot sleeving device comprises 1, the stimulation component 2, the control device 3, the wireless module 4 and the remote terminal 5.

Detailed Description

The present invention will be further described with reference to the following description and examples, which include but are not limited to the following examples.

Examples

The invention provides a foot stimulation training device based on machine learning, which comprises a foot sleeving device 1 fixed on a foot, a plurality of stimulation components 2 arranged in the foot sleeving device 1 and distributed according to foot acupuncture points, a control device 3 connected with all the stimulation components 2, a wireless module 4 connected with the control device 3, and a remote terminal 5 connected with the wireless module 4, as shown in figure 1. In this embodiment, the shape of the foot covering device is the same as that of any one of the foot cover, the sock, the shoe and the foot clip, so that the foot covering device can be conveniently fixed on the foot, and the stimulation component preferably adopts an electrode or an infrared lamp, and the stimulation to the acupuncture points of the foot is realized through direct current stimulation or infrared irradiation. And an air cushion is arranged below the stimulating component in the foot sleeve device 1.

The training device is used as a hardware part of foot stimulation training, and a software control part of the invention is introduced below, which continuously collects original data in a machine learning mode, then completes the original data by processing missing data in the original data, further generates a training model and continuously performs deep learning, and finally predicts stimulation training recovery degrees under different foot nerve injury conditions by using the model, thereby providing guidance for a doctor to formulate an optimal training plan.

As shown in fig. 2, the implementation process of the present invention is as follows:

1. database establishment

Published reports on the condition and the result of the foot stimulation training are collected, and raw data are obtained and classified into input and output of the data. The inputs for the raw data include condition variables such as age, gender, BMI, degree of foot nerve damage, and the like. The output of the data will be the training results, measured by the foot sensation score.

2. Preprocessing of raw data

Not all reports, the raw data are complete and this will lead to "loss" of the input data, the present invention makes preliminary guesses of these values based on the underlying assumptions, based on the existing biological interpretation: some initial values are assigned to missing data in the database as preliminary predicted missing values. These initial values will serve as starting points for iterative machine learning to take place later.

3. Missing data processing

It is pointed out that due to the problem of differences in the data acquisition method and the training scheme, the database may contain entries with incomplete input information, and if the results of studies with acceptable differences in design and purpose are collected to form the database, the probability of missing data is higher.

The idea behind the invention is, however, to first set the initial value to the average of the values present in the data set for any unknown property. The predicted value of the missing data is obtained by estimating all values of the neural network, then recursively applying the following equations until convergence, and finally returning the result of convergence:

a ^(s+1) ＝(1-η)f(a ^s )+ηa ^s

wherein, a ^(s+1) Representing a predicted value, and s represents the iteration times; f (a) ^s ) Representing a prediction obtained from a neural network for a; eta is 0.4; the function f remains fixed in each loop iteration.

4. Model development

After the lost data is preliminarily predicted and processed, a corresponding neural network model is established based on the basis.

5. Generating predictive guidelines for foot stimulation training

The generated model is used for training a neural network, and meanwhile, the model is recorded and the predicted performance is evaluated; then, the overall performance of the prediction is obtained from the performance of all the different models, wherein the prediction with greater uncertainty will guide how to supplement the database with additional data of a particular input type; the smaller uncertainty predictions will be used to generate a series of quantitative guidelines for foot stimulation training, the guidelines including: expected best training results for a particular patient, a training program that achieves the best results.

6. Making an optimal training plan and performing stimulation training by a training device

The generated prediction guide is used as a reference for a doctor to formulate a training plan, the formulated training plan is transmitted to the control device by the remote terminal through the wireless module to be stored, then the patient fixes the foot sleeving device to the foot by himself, the control device is started, and the control device automatically controls part or all of the stimulation components according to the training plan, so that stimulation in different stimulation sequences, different stimulation combinations and different duration is given to corresponding acupuncture points on the foot, and finally, ordered stimulation training is realized.

The invention adopts a machine learning mode to establish a corresponding neural network model, and utilizes the model to predict the stimulation training recovery degree under different foot nerve injuries, thereby providing guidance for a doctor to formulate an optimal training plan, and then according to the formulation of the optimal training plan, the invention can safely, long-time and foolproof use the foot sheathing device to perform stimulation training on the foot of a patient and better obtain the expected training effect. The invention can not only greatly improve the effect of foot stimulation training and provide reliable auxiliary training for treating foot nerve injury, but also definitely eliminate the stimulation training scheme after the training effect is predicted by the invention for some patients who are not suitable for foot stimulation training, thereby not only avoiding the delay of the state of an illness, but also helping the patients to save unnecessary medical expenses.

The above-mentioned embodiment is only one of the preferred embodiments of the present invention, and should not be used to limit the scope of the present invention, but all the insubstantial modifications or changes made within the spirit and scope of the main design of the present invention, which still solve the technical problems consistent with the present invention, should be included in the scope of the present invention.

Claims

1. The foot stimulation training device based on machine learning is characterized by comprising a foot sleeving device (1) for fixing on a foot, a plurality of stimulation components (2) which are arranged in the foot sleeving device (1) and distributed according to foot acupuncture points, a control device (3) connected with all the stimulation components, a wireless module (4) connected with the control device (3), and a remote terminal (5) connected with the wireless module (4);

the implementation method of the foot stimulation training device based on machine learning comprises the following steps:

a. collecting previous foot stimulation training reports, obtaining original data, classifying the original data into input and output of the data, and then establishing a database;

b. assigning initial values to missing data in the database based on existing biological interpretations;

c. setting the initial value assigned to the missing data as the average of the values present in the database, then recursively applying the following equations until convergence, and finally returning the result of convergence to obtain the predicted value of the missing data:

a ^(s+1 )＝(1-η)f(a ^s )+ηa ^s

wherein, a ^(s+1 ) Representing a predicted value, and s represents the iteration number; f (a) ^s ) Representing a prediction about a obtained from a neural network; eta is a constant and the value is less than 1; the function f remains fixed in each loop iteration;

d. establishing a foot neural network model according to the predicted values of the original data and the missing data;

e. using the established model to train a neural network, and simultaneously recording the model and evaluating the performance of prediction, wherein the prediction with less uncertainty is used for generating a prediction guide for foot stimulation training, including an expected optimal training result; the generated prediction guide is used as a reference for making a training plan;

the wireless module (4) is used for sending the formulated training plan to the control device (3);

the control device (3) is used for controlling the stimulation component to work according to the training plan.

2. The machine learning-based foot stimulation training device according to claim 1, wherein the foot casing device has the same shape as any one of a foot casing, a sock, a shoe and a foot clip.

3. The machine learning-based foot stimulation training device according to claim 2, characterized in that an air cushion is further arranged in the foot cover device (1) below the stimulation component.

4. The machine learning-based foot stimulation training device according to any one of claims 1 to 3, wherein the stimulation means is an electrode or an infrared lamp.

5. The machine-learning based foot stimulation training device of claim 1, wherein η =0.4.