CN116644270A - A method and system for evaluating myotonia based on recursive graph and width learning - Google Patents
A method and system for evaluating myotonia based on recursive graph and width learning Download PDFInfo
- Publication number
- CN116644270A CN116644270A CN202310597573.6A CN202310597573A CN116644270A CN 116644270 A CN116644270 A CN 116644270A CN 202310597573 A CN202310597573 A CN 202310597573A CN 116644270 A CN116644270 A CN 116644270A
- Authority
- CN
- China
- Prior art keywords
- myotonia
- recursion
- time series
- width learning
- graph
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Classifications
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/22—Ergometry; Measuring muscular strength or the force of a muscular blow
- A61B5/224—Measuring muscular strength
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/45—For evaluating or diagnosing the musculoskeletal system or teeth
- A61B5/4519—Muscles
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
- A61B5/7235—Details of waveform analysis
- A61B5/7264—Classification of physiological signals or data, e.g. using neural networks, statistical classifiers, expert systems or fuzzy systems
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
- A61B5/7235—Details of waveform analysis
- A61B5/7264—Classification of physiological signals or data, e.g. using neural networks, statistical classifiers, expert systems or fuzzy systems
- A61B5/7267—Classification of physiological signals or data, e.g. using neural networks, statistical classifiers, expert systems or fuzzy systems involving training the classification device
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/10—Pre-processing; Data cleansing
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/213—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
- G06N3/0455—Auto-encoder networks; Encoder-decoder networks
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2123/00—Data types
- G06F2123/02—Data types in the time domain, e.g. time-series data
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/08—Feature extraction
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/12—Classification; Matching
Landscapes
- Engineering & Computer Science (AREA)
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Physics & Mathematics (AREA)
- Artificial Intelligence (AREA)
- Theoretical Computer Science (AREA)
- Evolutionary Computation (AREA)
- Biophysics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Biomedical Technology (AREA)
- Molecular Biology (AREA)
- Data Mining & Analysis (AREA)
- General Health & Medical Sciences (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Animal Behavior & Ethology (AREA)
- Veterinary Medicine (AREA)
- Public Health (AREA)
- Surgery (AREA)
- Medical Informatics (AREA)
- Heart & Thoracic Surgery (AREA)
- Mathematical Physics (AREA)
- Pathology (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Signal Processing (AREA)
- Psychiatry (AREA)
- Physiology (AREA)
- Fuzzy Systems (AREA)
- Computational Linguistics (AREA)
- Computing Systems (AREA)
- Software Systems (AREA)
- Rheumatology (AREA)
- Orthopedic Medicine & Surgery (AREA)
- Oral & Maxillofacial Surgery (AREA)
- Dentistry (AREA)
- Physical Education & Sports Medicine (AREA)
- Measurement Of The Respiration, Hearing Ability, Form, And Blood Characteristics Of Living Organisms (AREA)
Abstract
本发明公开了一种基于递归图和宽度学习的肌强直评估方法及系统，涉及机器学习技术领域，用以解决现有模型对于输入的肌张力信号分类不准的问题。本发明的技术要点包括：采集不同人的肌张力时间序列信号，并对肌张力时间序列信号进行预处理；利用预处理后的肌张力时间序列信号构建递归图；提取不同人的肌张力时间序列信号所对应的基于递归图的量化特征；将量化特征输入基于宽度学习的分类模型中进行训练，获取训练好的分类模型；提取待测的肌张力时间序列信号所对应的量化特征，并将其输入训练好的分类模型中获取肌强直预测结果。本发明在实现时间序列可视化同时可获取具有明显区分度的肌张力特征，可满足肌强直评估中对评估准确率和效率的要求。
The invention discloses a myotonia evaluation method and system based on recursive graph and width learning, relates to the technical field of machine learning, and is used to solve the problem of inaccurate classification of input muscle tension signals by existing models. The technical points of the present invention include: collecting muscle tension time series signals of different people, and preprocessing the muscle tension time series signals; using the preprocessed muscle tension time series signals to construct a recursive graph; extracting muscle tension time series signals of different people The quantitative features based on the recursive graph corresponding to the signal; the quantitative features are input into the classification model based on width learning for training, and the trained classification model is obtained; the quantitative features corresponding to the muscle tension time series signal to be measured are extracted, and Input the trained classification model to obtain the prediction result of myotonia. The present invention can obtain muscle tension features with obvious distinction while realizing time series visualization, and can meet the requirements for evaluation accuracy and efficiency in the evaluation of myotonia.
Description
Technical Field
The invention relates to the technical field of machine learning, in particular to a myotonic assessment method and system based on a recursion chart and width learning.
Background
Myotonic symptoms are marked by involuntary continuous contraction of muscles caused by increased muscle tension in the exercise process, are often seen in diseases such as myotonic muscular dystrophy, congenital myotonic, parkinsonism and the like, and are clinically diagnosed according to muscle knocks, electromyography examination, BFMDRS scale and the like. However, the myotonia and spasm with different degrees, and the changes of the myotonia of different parts of the body and different muscle groups cannot be objectively evaluated and quantified, and the slight changes of the myotonia are difficult to be specifically evaluated. In clinic, the myotonia degree of a patient is mainly judged by a doctor according to clinical experience, the artificial operation has uncertainty, and the evaluation result is too subjective.
The muscular tension signals acquired by the wearable equipment are usually in a time sequence form, and the non-stationary time sequence signals can be processed by utilizing the recursion phenomenon of a two-dimensional image visualization dynamics system based on the concept of a recursion diagram. Recursion is a fundamental property of a dynamic system that can be used to characterize the behavior of the dynamic system in phase space, meaning that the state of the system is restorative under certain conditions, and after a sufficient period of time, a state of the system will return to near its initial state. The periodicity, chaos and non-stationarity of the time series can be analyzed by using a recursion method, the internal structure of the time series is revealed, and prior knowledge about similarity, information quantity and predictability is given.
The traditional neural network often accompanies the problems of complex model network structure, huge parameters, time consumption, calculation resources and the like in classification processing. The width learning system only comprises a mapping feature layer and an enhancement layer, the model is strong in expansibility, gradient descent is not needed for updating the weight, and calculation is simple. In addition, the breadth-learning model can promote classification effects by laterally expanding the number of enhancement nodes and feature nodes in the network.
Disclosure of Invention
Therefore, the invention provides a myotonic assessment method and a myotonic assessment system based on a recursion diagram and width learning, which are used for solving the problem that the input myotonic signals are not classified accurately by the existing model.
According to an aspect of the present invention, there is provided a myotonic assessment method based on a recursive graph and a width learning, the method comprising the steps of:
step one, collecting muscular tension time series signals of different people, and preprocessing the muscular tension time series signals; wherein the time series signals of the muscular tension of different people comprise time series signals of the muscular tension corresponding to the patients with the mild myotonia, the patients with the severe myotonia and the normal people without the symptoms of the myotonia;
step two, constructing a recursion chart by using the preprocessed muscle tension time sequence signals;
extracting quantized features based on a recursion map, which correspond to muscular tension time series signals of different people;
inputting the quantized features into a classification model based on width learning for training, and obtaining a trained classification model;
and fifthly, extracting quantized features based on a recursion graph and corresponding to the muscle tension time sequence signals to be detected, and inputting the quantized features into a trained classification model to obtain a muscle rigidity prediction result.
Further, the preprocessing in the first step includes rejecting outliers present in the signal.
Further, the specific steps of the second step include: according to a time delay reconstruction principle, carrying out phase space reconstruction on the muscle tension time sequence signal; calculating the distance between any two points in the reconstructed phase space, wherein the distance corresponds to a recursion value; a two-dimensional recursion map is drawn using the plurality of recursion values.
Further, the quantization characteristic in the third step includes a recursion rate, a diagonal length entropy and a capturing time; wherein the recurrence rate is a measure of the density of recurrence points in the recurrence plot; the diagonal length entropy is shannon entropy of the diagonal distribution probability with the length of l in the recursion diagram; the capture time is a weighted average of the lengths of the vertical segments in the recursion chart.
Further, the recursive rate calculation formula is as follows:
wherein N represents the track length in the phase space; r is R i,j The recursive value is represented as a square matrix consisting of 0 and 1.
Further, the diagonal length entropy calculation formula is as follows:
wherein l represents the length of the diagonal line; p (l) represents the distribution probability of a diagonal of length l.
Further, the capturing time calculation formula is as follows:
wherein v represents a line segment perpendicular to the main diagonal, P (v) represents a probability of distribution of a perpendicular line segment of length v, v min Representing the minimum line segment length.
Further, the loss function of the classification model based on the width learning in the step four in the training process includes: the loss function is expressed as:
in the formula ,Zn Representing a mapping feature sequence obtained by nonlinear mapping; h m Representing a mapped feature layer sequence; w (W) m Representing an output weight matrix; λ represents a regularized term coefficient;representing the true value;
the corresponding output weight matrix when the loss function obtains the minimum value is:
wherein I represents an identity matrix.
Further, in the fourth step, the classification model based on the width learning improves the model performance by adding additional enhancement nodes in the training process, the enhancement nodes are generated according to the feature layer calculation, and the new enhancement layer is expressed as:
wherein ,generating a weight matrix and a deviation matrix of the new enhancement node by mapping the feature nodes; ζ represents a nonlinear function.
According to another aspect of the present invention, there is provided a myotonic assessment system based on a recursive graph and a width learning, the system comprising:
the data acquisition module is configured to acquire muscular tension time series signals of different people and preprocess the muscular tension time series signals; wherein the time series signals of the muscular tension of different people comprise time series signals of the muscular tension corresponding to the patients with the mild myotonia, the patients with the severe myotonia and the normal people without the symptoms of the myotonia;
a recurrence map construction module configured to construct a recurrence map using the preprocessed muscle tone time-series signal;
the characteristic extraction module is configured to extract quantized characteristics based on a recursion chart, which correspond to the muscular tension time series signals of different people;
the model training module is configured to input the quantized features into a classification model based on width learning for training, and obtain a trained classification model;
the prediction module is configured to extract quantized features based on a recursion graph corresponding to the muscle tension time sequence signals to be detected, and input the quantized features into a trained classification model to obtain a muscle rigidity prediction result.
The beneficial technical effects of the invention are as follows:
clinically, the degree of myotonia is usually given by doctors according to clinical experience, slight change of the muscle tension is easy to ignore, and the magnitude of the muscle tension is difficult to be quantified, and the invention designs the muscle tension detection device, and acquires muscle tension signals in the process of driving a certain part of a body to perform specific movement by external force so as to realize objective analysis of the muscle tension; the invention carries out quantitative analysis on the muscle tension signal based on a recursion diagram, and obtains the muscle tension characteristic parameter with obvious distinction degree while realizing time sequence visualization; the neural network with independent depth of width learning has the advantages of simple and flexible structure, high calculation speed and stronger nonlinearity, and can meet the requirements of evaluation accuracy and efficiency in the myotonic evaluation process.
Drawings
The invention may be better understood by reference to the following description taken in conjunction with the accompanying drawings, which are included to provide a further illustration of the preferred embodiments of the invention and to explain the principles and advantages of the invention, together with the detailed description below.
Fig. 1 is a flowchart of a myotonic assessment method based on a recursion diagram and width learning according to an embodiment of the present invention.
Fig. 2 is a hardware connection diagram of a muscle tension testing device used in an embodiment of the present invention.
FIG. 3 is a schematic diagram of a method for selecting phase space reconstruction parameters for constructing a recursive graph according to an embodiment of the present invention; wherein, the graph (a) corresponds to a delay time determining method; the graph (b) corresponds to the embedding dimension determination method.
FIG. 4 is a diagram of a recursive graphical representation of the construction of the acquired muscular tension signals of three types of tested personnel in an embodiment of the present invention; wherein, figure (a) corresponds to a healthy person; figure (b) corresponds to a mild patient; panel (c) corresponds to a severe patient.
FIG. 5 is an exemplary graph of extracted feature results when performing quantization analysis on a recursive graph according to an embodiment of the present invention; wherein, the graph (a) corresponds to a recursion rate; graph (b) corresponds to diagonal length entropy; graph (c) corresponds to the capture time.
FIG. 6 is a diagram of a breadth-learning model with additional reinforcing nodes employed in an embodiment of the present invention.
FIG. 7 is a diagram illustrating an example confusion matrix of classification results according to an embodiment of the present invention.
Detailed Description
In order that those skilled in the art will better understand the present invention, exemplary embodiments or examples of the present invention will be described below with reference to the accompanying drawings. It is apparent that the described embodiments or examples are only implementations or examples of a part of the invention, not all. All other embodiments or examples, which may be made by one of ordinary skill in the art without undue burden, are intended to be within the scope of the present invention based on the embodiments or examples herein.
The invention provides a myotonia assessment method and a myotonia assessment system based on a recursion chart and width learning, which solve the problem that the change of the myotonia is difficult to objectively assess through quantitative analysis of a myotonia signal.
The embodiment of the invention provides a myotonic assessment method based on a recursion diagram and width learning, as shown in fig. 1, comprising the following steps:
step one, collecting muscular tension time series signals of different people, and preprocessing the muscular tension time series signals; wherein the time series signals of the muscular tension of different people comprise time series signals of the muscular tension corresponding to the patients with the mild myotonia, the patients with the severe myotonia and the normal people without the symptoms of the myotonia;
step two, constructing a recursion chart by using the preprocessed muscle tension time sequence signals;
extracting quantized features based on a recursion map, which correspond to muscular tension time series signals of different people;
inputting the quantized features into a classification model based on width learning for training, and obtaining a trained classification model;
and fifthly, extracting quantized features based on a recursion graph and corresponding to the muscle tension time sequence signals to be detected, and inputting the quantized features into a trained classification model to obtain a muscle rigidity prediction result.
In the first step, a muscle tension signal is collected, and an abnormal value existing in the data is preprocessed and removed.
According to an embodiment of the invention, for data acquisition: and acquiring muscle tension data of a tested person in the movement process of a specific part through a clinical experiment, and recording the blocked force and the position information of the blocked force of the device in the process of driving the body part to perform specific reciprocating movement by external force. The arm is straightened to be used as an initial position, and the motor is used for driving the upper arm to perform specific movements such as uniform low-speed movement, uniform high-speed movement and the like. The encoder is used for collecting the rotation angle of the arm, the output force of the motor when the motor drives the arm to move is equal to the resistance of the arm, and the hardware connection of the detection device is shown in fig. 2. Myotonia data was collected for right arms of healthy persons, mild myotonia, and severe patients.
For data preprocessing: because the data acquisition process inevitably has the reasons of noise, equipment failure, manual misoperation and the like, the acquired data usually has various problems to influence the data analysis result, so the data preprocessing is performed first, and the data quality is improved. Based on an abnormal value detection mode of statistical characteristics, motor output torque data of each tested person in the test process are checked, the absolute value average value of torque under each motion instruction is calculated, and if the average value of a certain period is larger than that of other periods, the motion data of the period is most likely to be influenced by the force actively applied by the tested person, and the motion data are removed.
In the second step, a recursive graph is constructed to analyze the muscle tension time series signal, and the specific steps include:
(1) Reconstructing a phase space: phase space reconstruction is carried out on the time sequence through a time delay reconstruction technology;
(2) And (3) calculating the distance: calculating the distance between any two points in the reconstructed phase space;
(3) And (3) calculating a recursive value: the distance between any two points in the phase space corresponds to a recursion value;
(4) And (3) drawing a recursion chart: the recursion values are plotted as a two-dimensional recursion map.
According to the embodiment of the invention, the time series signal of the muscle tension is made to be u i I=1, 2, n, constructing a recursion graph to analyze the muscle tension time series signal, the method comprises the following specific steps:
(1) Reconstructing a phase space: let x= { X 1 ,X 2 ,X 3 ,...,X N }，X i ＝[u i ,u i+τ ,u i+2τ ,...,u i+(m-1)τ ]Where i=1, 2,..n, n=n- (m-1) τ, N is the track length of X in the reconstructed phase space. m is the embedding dimension, determined using a false nearest neighbor method, as shown in fig. 3 (a), m=3. τ is a delay time, and is a value corresponding to when the value of the mutual information function reaches the first minimum value in the average mutual information method, as shown in fig. 3 (b), τ=5.
(2) And (3) calculating the distance: calculating the distance r between any two points in the reconstructed phase space ij ＝||X i -X j I, where i, j=1, 2,..n, |·| represents a norm.
(3) And (3) calculating a recursive value: the distance between any two points in the phase space corresponds to a recursion value, R i,j ＝Θ(ε-||X i -X j ||)＝Θ(ε-r ij ) Where i, j=1, 2,..where N, epsilon is a preset threshold distance, Θ (x) is a Heaviside function,
(4) And (3) drawing a recursion chart: r is R i,j Is an n×n square matrix composed of 0 and 1, and on a two-dimensional coordinate axis, i is an abscissa, and j is an ordinate. R is R i,j When 1, the pixel at the two-dimensional coordinate (i, j) is 0, R i,j When 0, the pixel at (i, j) is 255. The recursion chart of the muscular tension signal conversion of the three tested persons is shown in fig. 4.
From an analysis of fig. 4, it can be seen that the three categories of recursion have distinct diagonals parallel to the main diagonal, which corresponds to the period of a fixed instruction in the instruction set. The existence of vertical lines and horizontal lines in the recursion diagrams of healthy people and mild patients indicates that in a certain process, the system phase space track is in an approximately unchanged state, the change of moment values is small, and the moment values correspond to the pause time after the execution of each instruction is finished; more arcs appear in the recurrence plot for a severe patient, indicating that two close trajectories in the phase space have changed at some time.
In the third step, the quantized features based on the recursive graph corresponding to the muscular tension time series signals of different people are extracted, and the graphic features of the recursive graph are analyzed by using the recursive quantization. Some properties of the recursive graph are quantized based on the density of the recursive points of the recursive graph and structural features such as diagonal lines, vertical lines, etc. in the recursive graph. And reserving the recursion characteristics with higher discrimination.
According to the embodiment of the invention, some features of the recursion chart are subjected to recursion quantization analysis, and the result is shown in fig. 5, and the specific description of the features is as follows:
(1) The recurrence rate is a simple measure of the density of recurrence points in a recurrence plot, expressed asN is the track length in phase space.
(2) The diagonal length entropy is shannon entropy of the diagonal distribution probability of length l in the recursion diagram, expressed asP (l) represents the distribution probability of a diagonal of length l. Reflecting the complexity of the recursive diagram with respect to the diagonal, the entropy of the uncorrelated temporal sequences, such as white noise, is very low, indicating that the signal is very low in complexity.
(3) The capture time refers to a weighted average of the lengths of the vertical segments in the recursion map, usingRepresentation, where v denotes a line segment perpendicular to the main diagonal, P (v) is the probability of distribution of a vertical line segment of length v, v min Is the smallest segment length, typically taken as 2. The value reflects that the system is staying at a certain positionThe duration of a particular state can be used to evaluate the stability of the system.
As can be seen from fig. 5, the quantization characteristic of the muscle tension signal based on the recurrence map is well differentiated for three types of tested persons.
In the fourth step, the quantized features are input into a classification model based on width learning for training, and a trained classification model is obtained.
According to an embodiment of the present invention, a width learning model is first constructed. Inputting data X, obtaining a characteristic layer Zn through nonlinear mapping, obtaining a reinforcing layer Hm through the characteristic layer, and obtaining a predicted value output by a width learning model through parallel connection of the reinforcing layer and the characteristic layer by an output weight matrix Wm, namely, obtaining an evaluation result of myotonia. The network structure is shown in fig. 6.
(1) Let the input data be X, through nonlinear mappingObtaining a mapping feature, wherein->Nonlinear activation function, let-> and />The method is characterized in that an input weight matrix and a bias vector are randomly generated, and in order to enable the generated mapping characteristics to be more simplified, a sparse self-coding idea is introduced to optimize the input weight matrix +.>Let Z n ＝[Z 1 ,Z 2 ,...,Z n ]The random features are generated as a more sparse and compact feature layer.
(2) Mapping feature layer passesGenerating an increaseStrong node, wherein-> and />Is randomly generated, nonlinear function ζ (x) =1/(1+e) -x ) Let the mapping layer be H m ＝[H 1 ,H 2 ,...,H m ]。
(3) The enhancement layer and the mapping layer are connected in parallel and pass through an output weight matrix W m Obtain output byTo represent the true value, the predicted value y= [ Z ] of the width learning network n |H m ]W m 。
Then, dividing the data set, and training a width learning model; in the training process, a loss function is defined as the square of the difference between the predicted value and the true valueWeight matrix W m The solution is performed by means of a least squares method,to avoid [ Z ] n |H m ]For the case of large errors caused by non-rank full order matrix, the loss function is added with a regularization term based on the ridge regression theory, and isλ represents a regularized term coefficient; the corresponding output weight matrix when the loss function obtains the minimum value is +.>I represents an identity matrix.
In the training process, if the performance of the width learning does not meet the expectations, the model performance is improved by adding additional enhancement nodes, and the newly added nodes are generated according to the feature layer calculation. The new enhancement layer is expressed in the following way wherein />The weight matrix and the deviation matrix of the new enhancement node are generated by mapping the feature nodes.
The muscle tension signal recursion chart quantitative feature set is divided into a training set and a testing set by a five-fold cross validation method. Training the constructed width learning through the training set, and storing the number of nodes and the related numerical value of the output weight matrix when the network performance reaches the expected standard, and verifying the performance of the method by utilizing the data of the testing set. And counting the evaluation results of each test set, wherein the accuracy rate is 89.06%, the accuracy rate is 83.57%, the recall rate is 82.26%, and the confusion matrix is drawn as shown in fig. 7.
Another embodiment of the present invention provides a myotonic assessment system based on a recursive graph and a width learning, the system comprising:
the data acquisition module is configured to acquire muscular tension time series signals of different people and preprocess the muscular tension time series signals; wherein the time series signals of the muscular tension of different people comprise time series signals of the muscular tension corresponding to the patients with the mild myotonia, the patients with the severe myotonia and the normal people without the symptoms of the myotonia;
a recurrence map construction module configured to construct a recurrence map using the preprocessed muscle tone time-series signal;
the characteristic extraction module is configured to extract quantized characteristics based on a recursion chart, which correspond to the muscular tension time series signals of different people;
the model training module is configured to input the quantized features into a classification model based on width learning for training, and obtain a trained classification model;
the prediction module is configured to extract quantized features based on a recursion graph corresponding to the muscle tension time sequence signals to be detected, and input the quantized features into a trained classification model to obtain a muscle rigidity prediction result.
The function of the myotonic assessment system based on the recursion graph and the width learning in the embodiment of the present invention may be illustrated by the aforementioned myotonic assessment method based on the recursion graph and the width learning, so that the system embodiment is not described in detail, and reference may be made to the above method embodiment, which is not described herein.
While the invention has been described with respect to a limited number of embodiments, those skilled in the art, having benefit of the above description, will appreciate that other embodiments are contemplated within the scope of the invention as described herein. The disclosure of the present invention is intended to be illustrative, but not limiting, of the scope of the invention, which is defined by the appended claims.
Claims (10)
1. The myotonic assessment method based on the recursion diagram and the width learning is characterized by comprising the following steps of:
step one, collecting muscular tension time series signals of different people, and preprocessing the muscular tension time series signals; wherein the time series signals of the muscular tension of different people comprise time series signals of the muscular tension corresponding to the patients with the mild myotonia, the patients with the severe myotonia and the normal people without the symptoms of the myotonia;
step two, constructing a recursion chart by using the preprocessed muscle tension time sequence signals;
extracting quantized features based on a recursion map, which correspond to muscular tension time series signals of different people;
inputting the quantized features into a classification model based on width learning for training, and obtaining a trained classification model;
and fifthly, extracting quantized features based on a recursion graph and corresponding to the muscle tension time sequence signals to be detected, and inputting the quantized features into a trained classification model to obtain a muscle rigidity prediction result.
2. The method according to claim 1, wherein the preprocessing in the first step includes eliminating outliers existing in the signal.
3. The method for assessing myotonia based on recursive graph and width learning according to claim 1, wherein the specific steps of the second step include: according to a time delay reconstruction principle, carrying out phase space reconstruction on the muscle tension time sequence signal; calculating the distance between any two points in the reconstructed phase space, wherein the distance corresponds to a recursion value; a two-dimensional recursion map is drawn using the plurality of recursion values.
4. A method of assessing myotonia based on recursive graph and width learning as claimed in claim 3, wherein the quantified features in step three include a rate of recursion, diagonal length entropy and capture time; wherein the recurrence rate is a measure of the density of recurrence points in the recurrence plot; the diagonal length entropy is shannon entropy of the diagonal distribution probability with the length of l in the recursion diagram; the capture time is a weighted average of the lengths of the vertical segments in the recursion chart.
5. The method for evaluating myotonia based on a recursion graph and width learning as claimed in claim 4, wherein the recursion rate calculation formula is as follows:
wherein N represents the track length in the phase space; r is R i,j The recursive value is represented as a square matrix consisting of 0 and 1.
6. The method for assessing myotonia based on recursive graph and width learning according to claim 4, wherein the diagonal length entropy calculation formula is as follows:
wherein l represents the length of the diagonal line; p (l) represents the distribution probability of a diagonal of length l.
7. The method for estimating myotonia based on recursive graph and width learning according to claim 4, wherein the capturing time calculation formula is as follows:
wherein v represents a line segment perpendicular to the main diagonal, P (v) represents a probability of distribution of a perpendicular line segment of length v, v min Representing the minimum line segment length.
8. The method for evaluating myotonia based on recursive graph and width learning according to claim 1, wherein the step four is characterized in that the loss function of the classification model based on width learning in the training process comprises: the loss function is expressed as:
in the formula ,Zn Representing a mapping feature sequence obtained by nonlinear mapping; h m Representing a mapped feature layer sequence; w (W) m Representing an output weight matrix; λ represents a regularized term coefficient;representing the true value;
the corresponding output weight matrix when the loss function obtains the minimum value is:
wherein I represents an identity matrix.
9. The method for evaluating myotonia based on recursive graph and width learning according to claim 8, wherein in the fourth step, the classification model based on width learning improves model performance by adding additional enhancement nodes in the training process, the enhancement nodes are generated according to feature layer calculation, and the new enhancement layer is expressed as:
wherein ,generating a weight matrix and a deviation matrix of the new enhancement node by mapping the feature nodes; ζ represents a nonlinear function.
10. A system for assessing myotonia based on a recursive graph and a breadth-based study, comprising:
the data acquisition module is configured to acquire muscular tension time series signals of different people and preprocess the muscular tension time series signals; wherein the time series signals of the muscular tension of different people comprise time series signals of the muscular tension corresponding to the patients with the mild myotonia, the patients with the severe myotonia and the normal people without the symptoms of the myotonia;
a recurrence map construction module configured to construct a recurrence map using the preprocessed muscle tone time-series signal;
the characteristic extraction module is configured to extract quantized characteristics based on a recursion chart, which correspond to the muscular tension time series signals of different people;
the model training module is configured to input the quantized features into a classification model based on width learning for training, and obtain a trained classification model;
the prediction module is configured to extract quantized features based on a recursion graph corresponding to the muscle tension time sequence signals to be detected, and input the quantized features into a trained classification model to obtain a muscle rigidity prediction result.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310597573.6A CN116644270B (en) | 2023-05-25 | 2023-05-25 | Myotonic assessment method and system based on recursion diagram and width learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310597573.6A CN116644270B (en) | 2023-05-25 | 2023-05-25 | Myotonic assessment method and system based on recursion diagram and width learning |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN116644270A true CN116644270A (en) | 2023-08-25 |
| CN116644270B CN116644270B (en) | 2025-11-11 |
Family
ID=87618210
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310597573.6A Active CN116644270B (en) | 2023-05-25 | 2023-05-25 | Myotonic assessment method and system based on recursion diagram and width learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116644270B (en) |
Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111898526A (en) * | 2020-07-29 | 2020-11-06 | 南京邮电大学 | EMG gesture recognition method based on multi-stream convolutional neural network |
| CN112396109A (en) * | 2020-11-19 | 2021-02-23 | 天津大学 | Motor bearing fault diagnosis method based on recursion graph and multilayer convolution neural network |
| CN112807632A (en) * | 2019-11-18 | 2021-05-18 | 深圳市理邦精密仪器股份有限公司 | Parameter optimization method and system of training instrument, muscle training equipment and device |
| US20210236044A1 (en) * | 2020-02-03 | 2021-08-05 | nQ Medical, Inc. | Methods and Apparatus for Assessment of Health Condition or Functional State from Keystroke Data |
| CN113509151A (en) * | 2021-08-30 | 2021-10-19 | 北京理工大学 | Method and device for assessing muscle tone levels |
| CN113855053A (en) * | 2021-06-07 | 2021-12-31 | 钧晟(天津)科技发展有限公司 | Wearable muscle threshold monitoring system based on myoelectricity |
| CN113967025A (en) * | 2021-10-28 | 2022-01-25 | 燕山大学 | A multi-level motor function assessment method based on electromyography |
| CN114587385A (en) * | 2022-02-25 | 2022-06-07 | 之江实验室 | A deep learning model construction method for post-stroke rehabilitation assessment based on graph theory features of brain muscle network |
| US20220369999A1 (en) * | 2019-09-30 | 2022-11-24 | Massachusetts Eye And Ear Infirmary | Objective evaluation of neurological movement disorders from medical imaging |
| CN115579130A (en) * | 2022-11-10 | 2023-01-06 | 中国中医科学院望京医院(中国中医科学院骨伤科研究所) | Method, device, equipment and medium for evaluating limb function of patient |
| US20230039900A1 (en) * | 2021-08-07 | 2023-02-09 | Fuzhou University | Method for realizing a multi-channel convolutional recurrent neural network eeg emotion recognition model using transfer learning |
-
2023
- 2023-05-25 CN CN202310597573.6A patent/CN116644270B/en active Active
Patent Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20220369999A1 (en) * | 2019-09-30 | 2022-11-24 | Massachusetts Eye And Ear Infirmary | Objective evaluation of neurological movement disorders from medical imaging |
| CN112807632A (en) * | 2019-11-18 | 2021-05-18 | 深圳市理邦精密仪器股份有限公司 | Parameter optimization method and system of training instrument, muscle training equipment and device |
| US20210236044A1 (en) * | 2020-02-03 | 2021-08-05 | nQ Medical, Inc. | Methods and Apparatus for Assessment of Health Condition or Functional State from Keystroke Data |
| CN111898526A (en) * | 2020-07-29 | 2020-11-06 | 南京邮电大学 | EMG gesture recognition method based on multi-stream convolutional neural network |
| CN112396109A (en) * | 2020-11-19 | 2021-02-23 | 天津大学 | Motor bearing fault diagnosis method based on recursion graph and multilayer convolution neural network |
| CN113855053A (en) * | 2021-06-07 | 2021-12-31 | 钧晟(天津)科技发展有限公司 | Wearable muscle threshold monitoring system based on myoelectricity |
| US20230039900A1 (en) * | 2021-08-07 | 2023-02-09 | Fuzhou University | Method for realizing a multi-channel convolutional recurrent neural network eeg emotion recognition model using transfer learning |
| CN113509151A (en) * | 2021-08-30 | 2021-10-19 | 北京理工大学 | Method and device for assessing muscle tone levels |
| CN113967025A (en) * | 2021-10-28 | 2022-01-25 | 燕山大学 | A multi-level motor function assessment method based on electromyography |
| CN114587385A (en) * | 2022-02-25 | 2022-06-07 | 之江实验室 | A deep learning model construction method for post-stroke rehabilitation assessment based on graph theory features of brain muscle network |
| CN115579130A (en) * | 2022-11-10 | 2023-01-06 | 中国中医科学院望京医院(中国中医科学院骨伤科研究所) | Method, device, equipment and medium for evaluating limb function of patient |
Non-Patent Citations (2)
| Title |
|---|
| PARK B K 等: "Analysis of viscoelastic properties of wrist joint for quantification of parkinsonian rigidity", IEEE TRANSACTIONS ON NEURAL SYSTEMS AND REHABILITATION ENGINEERING, vol. 19, no. 2, 31 December 2010 (2010-12-31), pages 2196 - 2203 * |
| 李新杰;胡铁松;董秀明;: "递归图法在径流时间序列非线性分析中的应用", 武汉大学学报(工学版), no. 01, 1 February 2013 (2013-02-01) * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN116644270B (en) | 2025-11-11 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN116602642B (en) | Heart rate monitoring method, device and equipment | |
| CN109993230B (en) | TSK fuzzy system modeling method for brain function magnetic resonance image classification | |
| CN114549589B (en) | Rotator vibration displacement measurement method and system based on lightweight neural network | |
| CN111626182A (en) | A method and system for accurate detection of human heart rate and facial blood volume based on video | |
| CN119439876A (en) | State monitoring method and system for multi-axis linkage CNC machining | |
| CN116541771A (en) | A Fault Diagnosis Method of Unbalanced Sample Bearing Based on Multi-scale Feature Fusion | |
| CN111127467A (en) | Image quantification method, computer equipment and storage medium | |
| CN109949280B (en) | Image processing method, image processing apparatus, device storage medium, and growth evaluation system | |
| CN110414306B (en) | Baby abnormal behavior detection method based on meanshift algorithm and SVM | |
| CN108399434B (en) | Analysis and prediction method of high-dimensional time series data based on feature extraction | |
| CN112069621B (en) | Method for predicting residual service life of rolling bearing based on linear reliability index | |
| CN115630278A (en) | Vibration Damage Detection Method Based on Channel-Space-Temporal Attention Mechanism Network | |
| CN110308713A (en) | A Method for Identification of Industrial Process Fault Variables Based on k-Nearest Neighbor Reconstruction | |
| CN119197362A (en) | A method and device for monitoring deformation of a structure based on a laser measuring device | |
| CN116740426A (en) | A classification and prediction system for functional magnetic resonance images | |
| CN114680904B (en) | Electroencephalogram signal quality assessment method based on characteristic wave detection and stage algorithm | |
| CN116644270B (en) | Myotonic assessment method and system based on recursion diagram and width learning | |
| CN111598966B (en) | Magnetic resonance imaging method and device based on generation countermeasure network | |
| CN118464429B (en) | Gearbox fault detection method, device, electronic equipment and storage medium | |
| CN116910544B (en) | A method and system for enhancing target signals based on signal similarity calculation | |
| CN114898241B (en) | Video repetitive motion counting system based on computer vision | |
| CN111796173B (en) | Partial discharge pattern recognition method, computer device, and storage medium | |
| CN120071973B (en) | Method for identifying heart sound abnormality of early-stage acute heart failure patient | |
| CN120431101B (en) | Skin aging gridding scoring system and method thereof | |
| CN120747872B (en) | A modal identification method for beam structures based on DIC, wavelet and PLSCF technologies |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |