CN111134664B - Epileptic discharge identification method and system based on capsule network and storage medium - Google Patents
Epileptic discharge identification method and system based on capsule network and storage medium Download PDFInfo
- Publication number
- CN111134664B CN111134664B CN201911309016.XA CN201911309016A CN111134664B CN 111134664 B CN111134664 B CN 111134664B CN 201911309016 A CN201911309016 A CN 201911309016A CN 111134664 B CN111134664 B CN 111134664B
- Authority
- CN
- China
- Prior art keywords
- data
- model
- training
- accuracy
- brain wave
- 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.)
- Active
Links
Images
Classifications
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/24—Detecting, measuring or recording bioelectric or biomagnetic signals of the body or parts thereof
- A61B5/316—Modalities, i.e. specific diagnostic methods
- A61B5/369—Electroencephalography [EEG]
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/24—Detecting, measuring or recording bioelectric or biomagnetic signals of the body or parts thereof
- A61B5/316—Modalities, i.e. specific diagnostic methods
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/40—Detecting, measuring or recording for evaluating the nervous system
- A61B5/4076—Diagnosing or monitoring particular conditions of the nervous system
- A61B5/4094—Diagnosing or monitoring seizure diseases, e.g. epilepsy
-
- 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
-
- 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; CALCULATING OR 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
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
Landscapes
- Health & Medical Sciences (AREA)
- Engineering & Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Physics & Mathematics (AREA)
- Artificial Intelligence (AREA)
- Veterinary Medicine (AREA)
- Medical Informatics (AREA)
- General Health & Medical Sciences (AREA)
- Public Health (AREA)
- Biomedical Technology (AREA)
- Animal Behavior & Ethology (AREA)
- Surgery (AREA)
- Pathology (AREA)
- Biophysics (AREA)
- Molecular Biology (AREA)
- Heart & Thoracic Surgery (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Data Mining & Analysis (AREA)
- Theoretical Computer Science (AREA)
- Evolutionary Computation (AREA)
- Neurology (AREA)
- Psychiatry (AREA)
- Physiology (AREA)
- Bioinformatics & Computational Biology (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Biology (AREA)
- Signal Processing (AREA)
- Bioinformatics & Cheminformatics (AREA)
- General Physics & Mathematics (AREA)
- Neurosurgery (AREA)
- Fuzzy Systems (AREA)
- Mathematical Physics (AREA)
- Psychology (AREA)
- Measurement And Recording Of Electrical Phenomena And Electrical Characteristics Of The Living Body (AREA)
Abstract
The invention relates to an epileptic discharge identification method, an epileptic discharge identification system and a storage medium based on a capsule network, wherein the method comprises the steps of obtaining a plurality of original brain wave data, preprocessing all the original brain wave data to obtain a plurality of target brain wave data; making all target electroencephalogram data into a data set, and obtaining a classification model according to the data set based on a capsule network learning method; sequentially carrying out parameter tuning and verification on the classification model according to the data set to obtain a target optimization model; and identifying the electroencephalogram data to be identified according to the target optimization model to obtain an identification result. The capsule network learning method based on the electroencephalogram data detection can replace manual visual inspection and reading, automatically carry out intelligent detection and identification on a large amount of electroencephalogram data, detect the electroencephalogram data of epileptic-like discharge in real time, intervene in clinic timely, and is high in efficiency, strong in stability and high in accuracy.
Description
Technical Field
The invention relates to the field of artificial intelligence and bioscience, in particular to an epileptic discharge identification method and system based on a capsule network and a storage medium.
Background
The brain, the most important organ of the human body, is very complex in structure and function, and with the continuous development of the neuroelectrophysiological technology, the study of cranial nerves is one of the most important study directions at present. In clinical application, electroencephalogram is the spontaneous and rhythmic electrical activity of brain cell populations recorded by electrodes, is the most sensitive method for detecting brain functions, is an important means for assisting in diagnosing and treating neurological diseases, and has irreplaceable effects particularly on solving the qualitative and positioning problems of paroxysmal brain dysfunction such as epilepsy and the like.
Epilepsy is a common chronic syndrome with seizures as a clinical characteristic. In patients with a typical seizure in the clinic, epileptiform discharges are found in brain wave data examination in about 80%. Therefore, it is important to identify whether epileptiform discharge occurs in the electroencephalogram data. At present, an expert usually analyzes a large amount of brain wave data by visual inspection, and from the brain wave data of a patient suspected of suffering from epilepsy or epilepsy, identifies an irregularly occurring transient characteristic waveform related to epilepsy. Because of the characteristics of complexity and uncertainty of brain wave data, automatic identification and classification of the brain wave data by means of instruments are difficult to realize at present, the brain wave data are continuously monitored for a long time, and only the reading by manual visual inspection of professionals can be used, so that the workload is high, the identification efficiency is low, the professionals read by manual visual inspection for a long time easily generate fatigue, errors are easy to occur, the high identification accuracy rate cannot be ensured, and meanwhile, the brain wave data are difficult to judge in real time and feed back to clinics for timely intervention.
Therefore, an artificial intelligent, efficient and high-accuracy epileptic discharge identification method is urgently needed, which can replace artificial visual inspection reading, carry out intelligent detection and identification on a large amount of brain wave data, detect the brain wave data of epileptic-like discharge in real time and intervene in clinic timely.
Disclosure of Invention
The invention aims to solve the technical problems of the prior art, provides an epileptic discharge identification method, system and storage medium based on a capsule network, overcomes the defects of large workload, low efficiency and insufficient accuracy rate of traditional manual visual inspection reading for epileptic discharge identification, and can replace manual visual inspection reading to intelligently detect and identify a large amount of electroencephalogram data.
The technical scheme for solving the technical problems is as follows:
an epileptic discharge identification method based on a capsule network comprises the following steps:
step 1: acquiring a plurality of original brain wave data, and preprocessing all the original brain wave data to obtain a plurality of target brain wave data;
and 2, step: making all target electroencephalogram data into a data set, and obtaining a classification model according to the data set based on a capsule network learning method;
and step 3: sequentially carrying out parameter tuning and verification on the classification model according to the data set to obtain a target optimization model;
and 4, step 4: and identifying the electroencephalogram data to be identified according to the target optimization model to obtain an identification result.
The invention has the beneficial effects that: the acquired original brain wave data is preprocessed, so that data containing missing values can be screened out, more useful information data and information with larger influence on identification and classification can be analyzed, a data set can be conveniently manufactured according to the acquired target brain wave data, and a classification model with higher identification and classification accuracy can be conveniently acquired; the Capsule Network (capsuleNet, called capsuleNet for short) is a Capsule structure composed of neurons, and the input and output of the Capsule Network are vectors, so that the probability of a certain feature appearing can be represented through the modular length of the vectors, and the spatial information of the feature, including position, direction, size, deformation and the like, can be represented through the vectors, therefore, compared with the traditional Convolutional Network (called CNN for short), the Capsule Network learning method can learn most of the spatial information of an input image, and extract various different variants of the feature, which can also be obtained by changing the vectorization feature of a digital Capsule layer (DigiCaps), so that for the input of each category, the Capsule Network can learn a more robust representation than the traditional Convolutional Network; therefore, the classification model obtained based on the capsule network learning method can effectively improve the reliability and stability of classifying the epileptic discharge, and the obtained target optimization model can further improve the accuracy of epileptic discharge identification through parameter optimization and verification of the classification model, overcomes the defects of large workload, low efficiency and insufficient accuracy of traditional manual visual inspection reading for epileptic discharge identification, can replace manual visual inspection reading, automatically and intelligently detect and identify a large amount of electroencephalogram data, and detect the electroencephalogram data of epileptic discharge in real time, so that the clinical intervention can be timely performed, and the efficiency, the stability and the accuracy are high.
On the basis of the technical scheme, the invention can be further improved as follows:
further: the specific steps of the step 1 comprise:
step 1.1: acquiring a plurality of original electroencephalogram data;
step 1.2: performing data cleaning on all original brain wave data to obtain a plurality of pieces of intermediate brain wave data;
step 1.3: and performing data dimensionality reduction on all the intermediate brain wave data by adopting a principal component analysis method to obtain a plurality of target brain wave data.
Further: the specific steps of the step 2 comprise:
step 2.1: making all target brain wave data into a data set, and randomly dividing the data set into a training set, a testing set and a verification set;
step 2.2: based on the capsule network learning method, a training model is built by using a capsule network, and the training model is trained by using the training set according to preset iterative training times to obtain a first detection model;
step 2.3: inputting the test set into the first detection model for detection, acquiring a first accuracy rate of the first detection model, judging whether the first accuracy rate reaches an expected value, if so, determining the first detection model as the classification model, and if not, executing the step 2.4;
step 2.4: training the training model by using the test set according to the preset iterative training times to obtain a second detection model;
step 2.5: inputting the training set into the second detection model for detection to obtain a second accuracy of the second detection model;
step 2.6: judging whether the second accuracy reaches the expected value, if so, determining the second detection model as the classification model, if not, returning to the step 2.1, randomly dividing the data set into a new training set, a new testing set and a new verification set again, repeating the step 2.2 to the step 2.5 until the first accuracy or the second accuracy reaches the expected value, and determining the first detection model corresponding to the first accuracy reaching the expected value as the classification model, or determining the second detection model corresponding to the second accuracy reaching the expected value as the classification model.
Further: in the step 2.2, the specific steps of constructing the training model include:
step 2.2.1: constructing a capsule network structure based on the capsule network learning method;
step 2.2.2: and respectively carrying out data format conversion on the training set and the test set, inputting the training set and the test set after the data format conversion into the capsule network structure, and classifying each capsule layer of the capsule network structure by using a dynamic routing method to obtain the training model.
Further: the specific steps of the step 3 comprise:
step 3a.1: when the first accuracy in the step 2.6 reaches the expected value, inputting the training set and the test set corresponding to the first accuracy reaching the expected value into the corresponding classification models, inputting parameter ranges of the corresponding classification models, and performing parameter tuning on the corresponding classification models by using a grid searching method to obtain optimized classification models corresponding to the corresponding classification models;
step 3a.2: verifying the corresponding optimized classification model by using the verification set corresponding to the first accuracy reaching the expected value, if the verification is passed, determining the corresponding optimized classification model as the target optimized model, and if the verification is not passed, returning to the step 2.1;
alternatively, the first and second electrodes may be,
step 3b.1: when the second accuracy in the step 2.6 reaches the expected value, inputting the training set and the test set corresponding to the second accuracy reaching the expected value into the corresponding classification models, inputting the parameter ranges of the corresponding classification models, and performing parameter tuning on the corresponding classification models by using a grid search method to obtain optimized classification models corresponding to the corresponding classification models;
step 3b.2: and verifying the corresponding optimized classification model by using the verification set corresponding to the second accuracy reaching the expected value, if the verification is passed, determining the corresponding optimized classification model as the target optimized model, and if the verification is not passed, returning to the step 2.1.
According to another aspect of the invention, an epileptic discharge recognition system based on a capsule network is provided, which comprises a data acquisition module, a data processing module, a model acquisition module, a model optimization module and a recognition module;
the data acquisition module is used for acquiring a plurality of original brain wave data;
the data processing module is used for preprocessing all the original brain wave data to obtain a plurality of target brain wave data;
the model acquisition module is used for making all the target brain wave data into a data set and acquiring a classification model according to the data set based on a capsule network learning method;
the model optimization module is used for sequentially carrying out parameter tuning and verification on the classification model according to the data set to obtain a target optimization model;
and the identification module is used for identifying the electroencephalogram data to be identified according to the target optimization model to obtain an identification result.
The invention has the beneficial effects that: the original brain wave data acquired by the data acquisition module is preprocessed by the data processing module, so that data containing missing values can be screened out, more useful information data and information with larger influence on identification and classification can be analyzed, a data set is manufactured according to the acquired target brain wave data, and a classification model with higher identification and classification accuracy can be acquired conveniently; the classification model obtained based on the capsule network learning method can effectively improve the reliability and stability of classifying epileptic discharge, the parameter of the classification model is adjusted, optimized and verified through the model optimization module, the obtained target optimization model can further improve the accuracy of epileptic discharge identification, the defects that the workload is large, the efficiency is low and the accuracy is insufficient in traditional epileptic discharge identification through manual visual inspection reading are overcome, manual visual inspection reading can be replaced, a large amount of brain wave data can be automatically and intelligently detected and identified, the brain wave data of epileptic discharge can be detected in real time, and timely intervention can be made for clinic, so that the efficiency is high, the stability is strong, and the accuracy is high.
On the basis of the technical scheme, the invention can be further improved as follows:
further: the data processing module comprises a data cleaning unit and a data dimension reduction unit;
the data cleaning unit is used for performing data cleaning on all the original brain wave data to obtain a plurality of pieces of intermediate brain wave data;
and the data dimension reduction unit is used for performing data dimension reduction on all the intermediate brain wave data by adopting a principal component analysis method to obtain a plurality of target brain wave data.
Further: the model acquisition module comprises a data set making unit, a training unit, a detection unit and an analysis unit;
the data set making unit is used for making all the target electroencephalogram data into data sets and randomly dividing the data sets into a training set, a testing set and a verification set;
the training unit is used for constructing a training model by using a capsule network based on the capsule network learning method, and training the training model by using the training set according to preset iterative training times to obtain a first detection model;
the detection unit is used for inputting the test set into the first detection model for detection, and acquiring a first accuracy of the first detection model;
the analysis unit is used for judging whether the first accuracy reaches an expected value or not, and if so, determining the first detection model as the classification model;
the training unit is further configured to train the training model by using the test set according to a preset iterative training frequency to obtain a second detection model when the analysis unit judges that the first accuracy rate does not reach the expected value;
the detection unit is further configured to input the training set into the second detection model for detection, so as to obtain a second accuracy of the second detection model;
the analysis unit is further configured to determine whether the second accuracy reaches the expected value, determine the second detection model as the classification model if the second accuracy reaches the expected value, and randomly divide the data set into a new training set, a new test set, and a new verification set again if the second accuracy does not reach the expected value;
wherein the training unit, the detection unit and the analysis unit form a closed loop.
According to another aspect of the present invention, there is provided a capsule network-based epileptic discharge recognition system, comprising a processor, a memory and a computer program stored in the memory and executable on the processor, wherein the computer program when executed implements the steps of a capsule network-based epileptic discharge recognition method of the present invention.
The beneficial effects of the invention are: the computer program stored in the memory is run on the processor to realize the epileptic discharge recognition of the invention, and based on the capsule network learning method, the defects of large workload, low efficiency and insufficient accuracy rate of the traditional epileptic discharge recognition through manual visual inspection reading are overcome, the manual visual inspection reading can be replaced, a large amount of electroencephalogram data can be automatically and intelligently detected and recognized, the electroencephalogram data of epileptic discharge can be detected in real time, and the clinical intervention can be carried out in time, so that the efficiency is high, the stability is strong, and the accuracy is high.
In accordance with another aspect of the present invention, there is provided a computer storage medium, including: at least one instruction which, when executed, performs a step in a method of epileptiform discharge identification and classification of the present invention.
The invention has the beneficial effects that: the computer storage medium containing at least one instruction is executed to realize the epileptic discharge identification of the invention, the defects of large workload, low efficiency and insufficient accuracy rate of the traditional epileptic discharge identification by manual visual inspection reading are overcome based on the capsule network learning method, the manual visual inspection reading can be replaced, intelligent detection and identification can be automatically carried out on a large amount of electroencephalogram data, the electroencephalogram data of epileptic discharge can be detected in real time, timely intervention can be made for clinic, the efficiency is high, the stability is strong, and the accuracy is high.
Drawings
Fig. 1 is a schematic flowchart of an epileptic discharge identification method based on a capsule network according to an embodiment of the present invention;
FIG. 2 is a schematic flow chart of obtaining multiple target electroencephalogram data according to a first embodiment of the present invention;
FIG. 3 is a waveform diagram of one of the original brain wave data according to the first embodiment of the present invention;
fig. 4 is a waveform schematic diagram of target brain wave data corresponding to one of the original brain wave data in the first embodiment of the present invention;
FIG. 5 is a schematic flow chart of obtaining a classification model according to a first embodiment of the present invention;
6-1 and 6-2 are schematic flow charts of obtaining a target optimization model in the first embodiment of the invention;
fig. 7 is a schematic structural diagram of an epileptic discharge identification system based on a capsule network according to a second embodiment of the present invention;
fig. 8 is a schematic structural diagram of another epilepsy discharge recognition system based on capsule network according to the second embodiment of the present invention.
Detailed Description
The principles and features of this invention are described below in conjunction with the following drawings, which are set forth by way of illustration only and are not intended to limit the scope of the invention.
The present invention will be described with reference to the accompanying drawings.
In a first embodiment, as shown in fig. 1, a method for recognizing epileptic discharge based on a capsule network includes the following steps:
s1: acquiring a plurality of original brain wave data, and preprocessing all the original brain wave data to obtain a plurality of target brain wave data;
s2: making all target brain wave data into a data set, and obtaining a classification model according to the data set based on a capsule network learning method;
s3: sequentially carrying out parameter tuning and verification on the classification model according to the data set to obtain a target optimization model;
s4: and identifying the electroencephalogram data to be identified according to the target optimization model to obtain an identification result.
The acquired original brain wave data is preprocessed, so that data containing missing values can be screened out, more useful information data and information with larger influence on identification and classification can be analyzed, a data set can be conveniently manufactured according to the acquired target brain wave data, and a classification model with higher identification and classification accuracy can be conveniently acquired; the classification model obtained based on the capsule network learning method can effectively improve the reliability and stability of epileptic discharge classification, and the obtained target optimization model can further improve the epileptic discharge identification accuracy through parameter optimization and verification of the classification model, overcomes the defects of large workload, low efficiency and insufficient accuracy of traditional manual visual inspection reading for epileptic discharge identification, can replace manual visual inspection reading, automatically and intelligently detect and identify a large amount of electroencephalogram data, and detect the electroencephalogram data of epileptic discharge in real time, so that timely intervention can be made for clinic, and the method has the advantages of high efficiency, strong stability and high accuracy.
Preferably, as shown in fig. 2, the specific steps of S1 include:
s1.1: acquiring a plurality of original electroencephalogram data;
s1.2: performing data cleaning on all original brain wave data to obtain a plurality of pieces of intermediate brain wave data;
s1.3: and (3) performing data dimensionality reduction on all the intermediate brain wave data by adopting a principal component analysis method to obtain a plurality of target brain wave data.
Through data cleaning, data with poor quality and missing values can be screened out, and a plurality of pieces of intermediate brain wave data with better quality are obtained; the data dimensionality reduction is carried out by a principal component analysis method, so that the main characteristics which have larger influence on epilepsy classification can be obtained conveniently, the calculation amount of subsequent steps is reduced, and the subsequent capsule network can obtain higher accuracy rate only by needing less training data;
a Principal Component Analysis (PCA) is also called Principal Component Analysis, and aims to convert multiple indexes into a small number of comprehensive indexes (Principal components) by using the idea of dimension reduction, wherein each Principal Component can reflect most information of an original variable and the contained information is not repeated; the method leads the complex factors to be a plurality of main components while introducing multi-aspect variables, simplifies the problem and obtains more scientific and effective data information; the method is a mathematical transformation method, which converts a given group of related variables into another group of uncorrelated variables through linear transformation, and the new variables are arranged according to the sequence that the variance is decreased in turn; keeping the total variance of the variables unchanged in the mathematical transformation, so that the first variable has the largest variance and is called a first principal component, and the second variable has the second largest variance and is irrelevant to the first variable and is called a second principal component; by analogy, n variables have n principal components; the specific operation steps of the principal component analysis method are the prior art, and are not described in detail.
Specifically, in the embodiment, original electroencephalogram data of 150 normal persons and epileptic patients are acquired, and the 150 original electroencephalogram data include 50 normal ripples, 50 ripples and 50 fast ripples; the 150 original electroencephalogram data are described by using a data format (150, 500), wherein 150 represents 150 original electroencephalogram data, and 500 represents the dimension of each original electroencephalogram data, namely the size of each original electroencephalogram data is 500 multiplied by 1, wherein 1 represents single-channel sampling, 500 is the total number of single-channel sampling points, for any original electroencephalogram data, characteristic values of some dimensions can generate a plurality of columns of missing values in the column direction, therefore, in the process of data cleaning, all columns of the missing values in the characteristic values are deleted, and corresponding intermediate electroencephalogram data can be obtained.
Specifically, in this embodiment, a waveform schematic diagram of one of the original brain wave data is shown in fig. 3, the dimension is 500, the data dimension reduction is performed on the original brain wave data by using a principal component analysis method, the dimension is reduced to 300, and a waveform schematic diagram of the corresponding target brain wave data is shown in fig. 4.
Preferably, as shown in fig. 5, the specific step of S2 includes:
s2.1: making all target electroencephalogram data into a data set, and randomly dividing the data set into a training set, a testing set and a verification set;
s2.2: based on the capsule network learning method, a training model is built by using a capsule network, and the training model is trained by using the training set according to preset iterative training times to obtain a first detection model;
s2.3: inputting the test set into the first detection model for detection, obtaining a first accuracy of the first detection model, judging whether the first accuracy reaches an expected value, if so, determining the first detection model as the classification model, and if not, executing S2.4;
s2.4: training the training model by using the test set according to the preset iterative training times to obtain a second detection model;
s2.5: inputting the training set into the second detection model for detection to obtain a second accuracy of the second detection model;
s2.6: and judging whether the second accuracy reaches the expected value, if so, determining the second detection model as the classification model, otherwise, returning to S2.1, randomly dividing the data set into a new training set, a new testing set and a new verification set again, repeating S2.2 to S2.5 until the first accuracy or the second accuracy reaches the expected value, and determining the first detection model corresponding to the first accuracy reaching the expected value as the classification model, or determining the second detection model corresponding to the second accuracy reaching the expected value as the classification model.
The data are randomly divided into a training set, a testing set and a verification set, so that the objectivity of the data can be ensured, human factors are reduced, and the accuracy of a subsequent classification model is effectively improved; meanwhile, the first detection model and the second detection model obtained based on the capsule network learning method can ensure higher classification accuracy and obtain a classification model meeting expectations; when the first accuracy of the first detection model does not reach an expected value, the test set is trained to obtain a second detection model, and the second accuracy is obtained by detecting the first detection model by using the training set, which is equivalent to exchanging the training set and the test set, so that expected classification models can be further obtained; when the second accuracy rate does not reach the expected value, the data set is randomly divided into a new training set, a new testing set and a new verifying set again, the first accuracy rate and the second accuracy rate are repeatedly detected until the first accuracy rate or the second accuracy rate reaches the expected value, the accuracy rate of detecting the brain wave data to be identified by the classification model corresponding to the expected value can be kept at a higher level all the time, and the stability and the reliability of epileptic discharge identification are high.
It should be noted that, because the training set, the test set, and the verification set are randomly divided each time, the random proportion of each time is different, and the train _ test _ split function may be invoked to perform random division.
Specifically, in the process of making 150 target electroencephalogram data into a data set in this embodiment, since the 150 target electroencephalogram data include 50 normal ripples, 50 ripples and 50 fast ripples, in the data set, sample tags are respectively labeled on the three types of data, and the corresponding sample tags are respectively N, F and R, so that the subsequent training process is facilitated through the sample tags.
Preferably, in S2.2, the specific step of constructing the training model includes:
s2.2.1: constructing a capsule network structure based on the capsule network learning method;
s2.2.2: and respectively carrying out data format conversion on the training set and the test set, inputting the training set and the test set after the data format conversion into the capsule network structure, and classifying each capsule layer of the capsule network structure by using a dynamic routing method to obtain the training model.
Through data format conversion, the establishment of a training model can be facilitated, the subsequent training based on a capsule network learning method is further facilitated, and the training efficiency and the training effect are improved; the dynamic routing method is the routing selection of the node, is determined by depending on the current state information of the network, adjusts the routing table of the dynamic routing method according to the change of the network flow and the topological structure, and finds out the optimal routing, so that each capsule layer is classified by using the dynamic network routing method, the classification difference between the classification result of the obtained classification model after the obtained training model is trained and the classification result of the original data is smaller, and the classification effect is good; the specific operation steps of the dynamic routing method are the prior art, and are not described in detail.
Specifically, in this embodiment s2.2.2, the training set and the test set after the data format conversion are all input into the capsule network structure, and the three-dimensional tensor obtained by one layer of standard convolution (the number of convolutions is 32, the convolution kernel is 1 × 1, the step size is 1, and the activation function is relu) is (None, 100, 32), and the number of parameters therein is 64; then, performing convolution with a square activation function (the number of capsules is 4, the number of channels is 32, a convolution kernel is 1 multiplied by 1, and the step length is 1) to obtain a reshape (None, 3200, 8); and entering a core capsule layer through a squash activation function, wherein the dynamic routing method is to output a classification result in the core capsule layer, and label marking is carried out on the corresponding classification result to complete the construction of a training model.
Preferably, as shown in fig. 6-1 and 6-2, the specific steps of S3 include:
s3a.1: when the first accuracy reaches the expected value in S2.6, inputting the training set and the test set corresponding to the first accuracy reaching the expected value into the corresponding classification model, inputting the parameter range of the corresponding classification model, and performing parameter tuning on the corresponding classification model by using a grid search method to obtain an optimized classification model corresponding to the corresponding classification model;
s3, 3a.2: verifying the corresponding optimized classification model by using the verification set corresponding to the first accuracy reaching the expected value, if the verification is passed, determining the corresponding optimized classification model as the target optimized model, and if the verification is not passed, returning to S2.1;
alternatively, the first and second electrodes may be,
s3b.1: when the second accuracy reaches the expected value in S2.6, inputting the training set and the test set corresponding to the second accuracy reaching the expected value into the corresponding classification model, inputting the parameter range of the corresponding classification model, and performing parameter tuning on the corresponding classification model by using a grid search method to obtain an optimized classification model corresponding to the corresponding classification model;
s3b.2: and verifying the corresponding optimized classification model by using the verification set corresponding to the second accuracy reaching the expected value, if the verification is passed, determining the corresponding optimized classification model as the target optimized model, and if the verification is not passed, returning to S2.1.
The grid searching method is used for parameter tuning, so that the optimal parameters corresponding to the classification model can be obtained, and the parallelism of the automatic tuning process is high; the optimized classification model is verified by using a verification set, and the verified optimized classification model is determined as a target optimized model, so that the classification accuracy of the target optimized model on epileptic discharge can be further ensured, the brain wave data of epileptic sample discharge can be detected in real time, and the clinical intervention can be performed in time; the specific operation steps of the grid search method are the prior art, and are not described in detail.
It should be noted that, in this embodiment, parameter tuning and verification are performed on the classification model corresponding to the first accuracy reaching the expected value, or parameter tuning and verification are performed on the classification model corresponding to the second accuracy reaching the expected value; when the first accuracy is the expected value, performing parameter tuning by using a training set and a test set corresponding to the first accuracy reaching the expected value, and verifying by using a corresponding verification set; and if the second accuracy is the expected value, performing parameter optimization by using the training set and the test set corresponding to the second accuracy, and verifying by using the corresponding verification set.
Specifically, the parameters to be tuned and optimized in S3.1 of this embodiment include the learning rate, the iteration number, and the like in the classification model, and the parameter ranges of these parameters are customized and input into the classification model, and then these parameters are subjected to parameter tuning and optimization by using a grid search method to obtain an optimal parameter combination, and then the optimal parameter combination is input into the classification model, so that the optimal classification model is obtained. Meanwhile, when the data set is too large, parameter tuning can be performed by combining a coordinate descending method, the specific operation step of coordinate descending is the prior art, and details are not repeated in this embodiment.
Specifically, in the process of verifying the optimized classification model in this embodiment S3.2, the verification set is input into the optimized classification model, and then four evaluation indexes of the optimized classification model are calculated, which are respectively Recall (Recall), accuracy (Accuracy), precision (Precision) and comprehensive evaluation index (F-Measure, or F-Score), and whether the optimized classification model passes the verification is comprehensively evaluated according to specific values of the four evaluation indexes.
In a second embodiment, as shown in fig. 7, a system for recognizing epileptic discharge based on a capsule network includes a data acquisition module, a data processing module, a model acquisition module, a model optimization module, and a recognition module;
the data acquisition module is used for acquiring a plurality of original brain wave data;
the data processing module is used for preprocessing all the original brain wave data to obtain a plurality of target brain wave data;
the model acquisition module is used for making all target electroencephalogram data into a data set and acquiring a classification model according to the data set based on a capsule network learning method;
the model optimization module is used for sequentially carrying out parameter tuning and verification on the classification model according to the data set to obtain a target optimization model;
and the identification module is used for identifying the brain wave data to be identified according to the target optimization model to obtain an identification result.
The data processing module is used for preprocessing the original brain wave data acquired by the data acquisition module, so that data containing missing values can be screened out, more useful information data and information with larger influence on identification and classification can be analyzed, a data set is manufactured according to the acquired target brain wave data, and a classification model with higher identification and classification accuracy can be acquired conveniently; the classification model obtained based on the capsule network learning method can effectively improve the reliability and stability of classifying epileptic discharge, and the obtained target optimization model can further improve the accuracy of epileptic discharge identification through the parameter optimization and verification of the classification model by the model optimization module, overcomes the defects of large workload, low efficiency and insufficient accuracy of epileptic discharge identification through the traditional manual visual inspection reading, can replace manual visual inspection reading, automatically and intelligently detect and identify a large amount of electroencephalogram data, and detects the electroencephalogram data of epileptic discharge in real time.
Preferably, as shown in fig. 8, the data processing module includes a data cleaning unit and a data dimension reduction unit;
the data cleaning unit is used for performing data cleaning on all the original brain wave data to obtain a plurality of pieces of intermediate brain wave data;
and the data dimension reduction unit is used for performing data dimension reduction on all the intermediate brain wave data by adopting a principal component analysis method to obtain a plurality of target brain wave data.
Through data cleaning, data with poor quality and missing values can be screened out, and a plurality of pieces of intermediate brain wave data with better quality are obtained; the data dimensionality reduction is carried out by a principal component analysis method, so that the main characteristics which have larger influence on epilepsy classification can be obtained conveniently, the calculation amount of subsequent steps is reduced, and the subsequent capsule network can obtain higher accuracy rate only by needing less training data.
Preferably, as shown in fig. 8, the model acquisition module includes a data set making unit, a training unit, a detection unit and an analysis unit;
the data set making unit is used for making all the target electroencephalogram data into data sets and randomly dividing the data sets into a training set, a testing set and a verification set;
the training unit is used for constructing a training model by using a capsule network based on the capsule network learning method, and training the training model by using the training set according to preset iterative training times to obtain a first detection model;
the detection unit is used for inputting the test set into the first detection model for detection, and acquiring a first accuracy of the first detection model;
the analysis unit is used for judging whether the first accuracy reaches an expected value or not, and if so, determining the first detection model as the classification model;
the training unit is further configured to train the training model by using the test set according to a preset iterative training frequency when the analysis unit determines that the first accuracy does not reach the expected value, so as to obtain a second detection model;
the detection unit is further configured to input the training set into the second detection model for detection, and obtain a second accuracy of the second detection model;
the analysis unit is further configured to determine whether the second accuracy reaches the expected value, determine the second detection model as the classification model if the second accuracy reaches the expected value, and randomly divide the data set into a new training set, a new test set, and a new verification set again if the second accuracy does not reach the expected value;
wherein the training unit, the detection unit and the analysis unit form a closed loop.
Through the model acquisition module formed by the units, the objectivity of data can be ensured, human factors are reduced, and the accuracy of a subsequent classification model is effectively improved; meanwhile, the first detection model and the second detection model obtained based on the capsule network learning method can ensure higher classification accuracy, obtain a classification model according with expectation, and have high stability and reliability for epileptic discharge recognition.
Preferably, as shown in fig. 8, the model optimization module includes a parameter tuning unit and a verification unit;
the parameter tuning unit is configured to, when the first accuracy reaches the expected value, input both the training set and the test set corresponding to the first accuracy reaching the expected value into the corresponding classification model, and input a parameter range of the corresponding classification model, and perform parameter tuning on the corresponding classification model by using a grid search method, to obtain an optimized classification model corresponding to the corresponding classification model; or, when the second accuracy reaches the expected value, inputting the training set and the test set corresponding to the second accuracy reaching the expected value into the corresponding classification model, inputting a parameter range of the corresponding classification model, and performing parameter tuning on the corresponding classification model by using a grid search method to obtain an optimized classification model corresponding to the corresponding classification model;
the verification unit is used for verifying the corresponding optimized classification model by using the verification set corresponding to the first accuracy reaching the expected value, and if the verification is passed, determining the corresponding optimized classification model as the target optimized model; or, the verification set corresponding to the second accuracy reaching the expected value is used for verifying the corresponding optimized classification model, and if the verification passes, the corresponding optimized classification model is determined as the target optimized model.
The grid searching method is used for parameter tuning, so that the optimal parameters corresponding to the classification model can be obtained, and the parallelism of the automatic tuning process is high; the optimized classification model is verified by using a verification set, and the verified optimized classification model is determined to be a target optimization model, so that the accuracy of the target optimization model in classifying epileptic discharge can be further ensured, the brain wave data of epileptic sample discharge can be detected in real time, and timely intervention can be made for clinic.
In a third embodiment, based on the first embodiment and the second embodiment, the present embodiment further discloses an epileptic discharge identification system based on a capsule network, which includes a processor, a memory, and a computer program stored in the memory and executable on the processor, and the computer program implements the specific steps S1 to S4 shown in fig. 1 when running.
The computer program stored in the memory is run on the processor to realize the epileptic discharge recognition of the invention, and based on the capsule network learning method, the defects of large workload, low efficiency and insufficient accuracy rate of the traditional epileptic discharge recognition through manual visual inspection reading are overcome, the manual visual inspection reading can be replaced, a large amount of electroencephalogram data can be automatically and intelligently detected and recognized, the electroencephalogram data of epileptic discharge can be detected in real time, and the clinical intervention can be carried out in time, so that the efficiency is high, the stability is strong, and the accuracy is high.
The present embodiment also provides a computer storage medium, where at least one instruction is stored on the computer storage medium, and when executed, the instruction implements the specific steps of S1 to S4.
The computer storage medium containing at least one instruction is executed to realize the epileptic discharge identification of the invention, the defects of large workload, low efficiency and insufficient accuracy rate of the traditional epileptic discharge identification by manual visual inspection reading are overcome based on the capsule network learning method, the manual visual inspection reading can be replaced, intelligent detection and identification can be automatically carried out on a large amount of electroencephalogram data, the electroencephalogram data of epileptic discharge can be detected in real time, timely intervention can be made for clinic, the efficiency is high, the stability is strong, and the accuracy is high.
Details of S1 to S4 in this embodiment are not described in detail in the first embodiment and the specific descriptions in fig. 1 to 6-2, which are not repeated herein.
The above description is only for the purpose of illustrating the preferred embodiments of the present invention and should not be taken as limiting the scope of the present invention, which is intended to cover any modifications, equivalents, improvements, etc. within the spirit and scope of the present invention.
Claims (8)
1. An epileptic discharge identification method based on a capsule network is characterized by comprising the following steps:
step 1: acquiring a plurality of original brain wave data, and preprocessing all the original brain wave data to obtain a plurality of target brain wave data;
step 2: making all target brain wave data into a data set, and obtaining a classification model according to the data set based on a capsule network learning method;
and step 3: sequentially carrying out parameter tuning and verification on the classification model according to the data set to obtain a target optimization model;
and 4, step 4: recognizing the electroencephalogram data to be recognized according to the target optimization model to obtain a recognition result;
the specific steps of the step 2 comprise:
step 2.1: making all target electroencephalogram data into a data set, and randomly dividing the data set into a training set, a testing set and a verification set;
step 2.2: based on the capsule network learning method, a training model is built by using a capsule network, and the training model is trained by using the training set according to preset iterative training times to obtain a first detection model;
step 2.3: inputting the test set into the first detection model for detection, obtaining a first accuracy of the first detection model, judging whether the first accuracy reaches an expected value, if so, determining the first detection model as the classification model, and if not, executing the step 2.4;
step 2.4: training the training model by using the test set according to the preset iterative training times to obtain a second detection model;
step 2.5: inputting the training set into the second detection model for detection, and acquiring a second accuracy of the second detection model;
step 2.6: judging whether the second accuracy reaches the expected value, if so, determining the second detection model as the classification model, if not, returning to the step 2.1, randomly dividing the data set into a new training set, a new testing set and a new verification set again, repeating the step 2.2 to the step 2.5 until the first accuracy or the second accuracy reaches the expected value, and determining the first detection model corresponding to the first accuracy reaching the expected value as the classification model, or determining the second detection model corresponding to the second accuracy reaching the expected value as the classification model.
2. The method for recognizing epileptic discharge based on capsule network as claimed in claim 1, wherein the specific steps of step 1 include:
step 1.1: acquiring a plurality of original brain wave data;
step 1.2: performing data cleaning on all original brain wave data to obtain a plurality of pieces of intermediate brain wave data;
step 1.3: and (3) performing data dimensionality reduction on all the intermediate brain wave data by adopting a principal component analysis method to obtain a plurality of target brain wave data.
3. The method for recognizing epileptic discharge based on capsule network as claimed in claim 1, wherein in the step 2.2, the specific step of constructing the training model comprises:
step 2.2.1: constructing a capsule network structure based on the capsule network learning method;
step 2.2.2: and respectively carrying out data format conversion on the training set and the test set, inputting the training set and the test set after the data format conversion into the capsule network structure, and classifying each capsule layer of the capsule network structure by using a dynamic routing method to obtain the training model.
4. The method for recognizing epileptic discharge based on capsule network as claimed in claim 1, wherein the specific steps of step 3 comprise:
step 3a.1: when the first accuracy in the step 2.6 reaches the expected value, inputting the training set and the test set corresponding to the first accuracy reaching the expected value into the corresponding classification model, inputting the parameter range of the corresponding classification model, and performing parameter tuning on the corresponding classification model by using a grid search method to obtain an optimized classification model of the corresponding classification model;
step 3a.2: verifying the corresponding optimized classification model by using the verification set corresponding to the first accuracy reaching the expected value, if the verification is passed, determining the corresponding optimized classification model as the target optimized model, and if the verification is not passed, returning to the step 2.1;
alternatively, the first and second electrodes may be,
step 3b.1: when the second accuracy in the step 2.6 reaches the expected value, inputting the training set and the test set corresponding to the second accuracy reaching the expected value into the corresponding classification models, inputting the parameter ranges of the corresponding classification models, and performing parameter tuning on the corresponding classification models by using a grid search method to obtain optimized classification models corresponding to the corresponding classification models;
step 3b.2: and verifying the corresponding optimized classification model by using the verification set corresponding to the second accuracy rate reaching the expected value, determining the corresponding optimized classification model as the target optimized model if the verification is passed, and returning to the step 2.1 if the verification is not passed.
5. An epileptic discharge recognition system based on a capsule network is characterized by comprising a data acquisition module, a data processing module, a model acquisition module, a model optimization module and a recognition module;
the data acquisition module is used for acquiring a plurality of original brain wave data;
the data processing module is used for preprocessing all the original brain wave data to obtain a plurality of target brain wave data;
the model acquisition module is used for making all target electroencephalogram data into a data set and acquiring a classification model according to the data set based on a capsule network learning method;
the model optimization module is used for sequentially carrying out parameter tuning and verification on the classification model according to the data set to obtain a target optimization model;
the identification module is used for identifying the electroencephalogram data to be identified according to the target optimization model to obtain an identification result;
the model acquisition module comprises a data set making unit, a training unit, a detection unit and an analysis unit;
the data set making unit is used for making all the target brain wave data into data sets and randomly dividing the data sets into a training set, a test set and a verification set;
the training unit is used for constructing a training model by using a capsule network based on the capsule network learning method, and training the training model by using the training set according to preset iterative training times to obtain a first detection model;
the detection unit is used for inputting the test set into the first detection model for detection to obtain a first accuracy of the first detection model;
the analysis unit is used for judging whether the first accuracy reaches an expected value or not, and if so, determining the first detection model as the classification model;
the training unit is further configured to train the training model by using the test set according to a preset iterative training frequency to obtain a second detection model when the analysis unit judges that the first accuracy rate does not reach the expected value;
the detection unit is further configured to input the training set into the second detection model for detection, so as to obtain a second accuracy of the second detection model;
the analysis unit is further configured to determine whether the second accuracy reaches the expected value, determine the second detection model as the classification model if the second accuracy reaches the expected value, and randomly divide the data set into a new training set, a new test set, and a new verification set again if the second accuracy does not reach the expected value;
wherein the training unit, the detection unit and the analysis unit form a closed loop.
6. The capsule network-based epileptic discharge recognition system of claim 5, wherein the data processing module comprises a data cleaning unit and a data dimension reduction unit;
the data cleaning unit is used for performing data cleaning on all the original brain wave data to obtain a plurality of pieces of intermediate brain wave data;
and the data dimension reduction unit is used for performing data dimension reduction on all the intermediate brain wave data by adopting a principal component analysis method to obtain a plurality of target brain wave data.
7. An epileptic discharge recognition system based on a capsule network, characterized by comprising a processor, a memory and a computer program stored in the memory and executable on the processor, the computer program when executed realizing the method steps according to any of claims 1 to 4.
8. A computer storage medium, the computer storage medium comprising: at least one instruction which when executed performs the method steps of any one of claims 1 to 4.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911309016.XA CN111134664B (en) | 2019-12-18 | 2019-12-18 | Epileptic discharge identification method and system based on capsule network and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911309016.XA CN111134664B (en) | 2019-12-18 | 2019-12-18 | Epileptic discharge identification method and system based on capsule network and storage medium |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111134664A CN111134664A (en) | 2020-05-12 |
| CN111134664B true CN111134664B (en) | 2022-11-01 |
Family
ID=70518711
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201911309016.XA Active CN111134664B (en) | 2019-12-18 | 2019-12-18 | Epileptic discharge identification method and system based on capsule network and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111134664B (en) |
Families Citing this family (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112057068A (en) * | 2020-08-27 | 2020-12-11 | 武汉工程大学 | Epilepsia pathological data classification method and device and storage medium |
| CN112365901A (en) * | 2020-11-03 | 2021-02-12 | 武汉工程大学 | Mechanical audio fault detection method and device |
| CN113361654A (en) * | 2021-07-12 | 2021-09-07 | 广州天鹏计算机科技有限公司 | Image identification method and system based on machine learning |
| CN114098764B (en) * | 2021-11-05 | 2024-02-20 | 北京大学 | Data processing method, device, electronic equipment and storage medium |
| CN114224288B (en) * | 2021-12-13 | 2023-10-31 | 中国人民解放军军事科学院军事医学研究院 | Microcapsule neural network training method and equipment for detecting epileptic brain electrical signals |
| CN114886440B (en) * | 2022-07-13 | 2022-09-30 | 武汉工程大学 | Epileptic sample discharge classification model training and recognition method, system and electronic equipment |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106408030A (en) * | 2016-09-28 | 2017-02-15 | 武汉大学 | SAR image classification method based on middle lamella semantic attribute and convolution neural network |
| CN108108583A (en) * | 2016-11-24 | 2018-06-01 | 南京理工大学 | A kind of adaptive SVM approximate models parameter optimization method |
| CN109815344A (en) * | 2019-01-29 | 2019-05-28 | 华南师范大学 | Network model training system, method, apparatus and medium based on parameter sharing |
| WO2019144081A1 (en) * | 2018-01-19 | 2019-07-25 | Mars, Incorporated | Biomarkers and classification algorithms for chronic kidney disease in cats |
Family Cites Families (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US11010902B2 (en) * | 2018-06-04 | 2021-05-18 | University Of Central Florida Research Foundation, Inc. | Capsules for image analysis |
| CN109645990B (en) * | 2018-08-30 | 2020-11-27 | 北京航空航天大学 | Computer mode identification method for electroencephalogram signals of epileptics |
-
2019
- 2019-12-18 CN CN201911309016.XA patent/CN111134664B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106408030A (en) * | 2016-09-28 | 2017-02-15 | 武汉大学 | SAR image classification method based on middle lamella semantic attribute and convolution neural network |
| CN108108583A (en) * | 2016-11-24 | 2018-06-01 | 南京理工大学 | A kind of adaptive SVM approximate models parameter optimization method |
| WO2019144081A1 (en) * | 2018-01-19 | 2019-07-25 | Mars, Incorporated | Biomarkers and classification algorithms for chronic kidney disease in cats |
| CN109815344A (en) * | 2019-01-29 | 2019-05-28 | 华南师范大学 | Network model training system, method, apparatus and medium based on parameter sharing |
Non-Patent Citations (3)
| Title |
|---|
| Capsulenet-Based Spatial–Spectral Classifier for Hyperspectral Images;Arun PV，等;《IEEE JOURNAL OF SELECTED TOPICS IN APPLIED EARTH OBSERVATIONS AND REMOTE SENSING》;20190630;第12卷(第6期);全文 * |
| 基于生成对抗网络的人脸图像修复;丁阳，等;《大连民族大学学报》;20190930;第21卷(第5期);全文 * |
| 自优化转导支持向量机并行化;王海涛，等;《计算机应用》;20171220;第37卷;全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN111134664A (en) | 2020-05-12 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111134664B (en) | Epileptic discharge identification method and system based on capsule network and storage medium | |
| CN111657935B (en) | Epilepsia electroencephalogram recognition system based on hierarchical graph convolutional neural network, terminal and storage medium | |
| US20210358611A1 (en) | Method for Detecting Epileptic Spike, Method for Training Network Model, and Computer Device | |
| EP3777674A1 (en) | Time series data learning and analyzing method using artificial intelligence | |
| KR20210011361A (en) | Time constant classification | |
| CN112294341B (en) | Sleep electroencephalogram spindle wave identification method and system based on light convolutional neural network | |
| CN113855038B (en) | Electrocardiosignal critical value prediction method and device based on multi-model integration | |
| Olesen et al. | Deep residual networks for automatic sleep stage classification of raw polysomnographic waveforms | |
| CN111956212A (en) | Inter-group atrial fibrillation identification method based on frequency domain filtering-multi-mode deep neural network | |
| CN113392894A (en) | Cluster analysis method and system for multi-group mathematical data | |
| CN116740426A (en) | Classification prediction system for functional magnetic resonance images | |
| Nhu et al. | Deep learning for automated epileptiform discharge detection from scalp EEG: A systematic review | |
| Kumar et al. | Comparison of Machine learning models for Parkinson’s Disease prediction | |
| Liu et al. | Using deep learning and machine learning to detect epileptic seizure with electroencephalography (EEG) data | |
| Wang et al. | A weighted overlook graph representation of eeg data for absence epilepsy detection | |
| CN113128585A (en) | Deep neural network based multi-size convolution kernel method for realizing electrocardiographic abnormality detection and classification | |
| CN112057068A (en) | Epilepsia pathological data classification method and device and storage medium | |
| CN116484290A (en) | Depression recognition model construction method based on Stacking integration | |
| CN114224354B (en) | Arrhythmia classification method, arrhythmia classification device, and readable storage medium | |
| Mur et al. | Unsupervised event detection and classification of multichannel signals | |
| CN112465054B (en) | FCN-based multivariate time series data classification method | |
| CN114742107A (en) | Method for identifying perception signal in information service and related equipment | |
| Akila et al. | Parkinson classification neural network with mass algorithm for processing speech signals | |
| Aguilera-Martos et al. | TSFEDL: A Python Library for Time Series Spatio-Temporal Feature Extraction and Prediction using Deep Learning (with Appendices on Detailed Network Architectures and Experimental Cases of Study) | |
| Mamidi et al. | A Hybrid Approach to Parkinson's Disease Detection using Speech Attributes: The Combination of SMOTE and Active Learning |
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 |