CN111700592A - Method and system for acquiring epilepsia electroencephalogram automatic classification model and classification system - Google Patents
Method and system for acquiring epilepsia electroencephalogram automatic classification model and classification system Download PDFInfo
- Publication number
- CN111700592A CN111700592A CN202010625498.6A CN202010625498A CN111700592A CN 111700592 A CN111700592 A CN 111700592A CN 202010625498 A CN202010625498 A CN 202010625498A CN 111700592 A CN111700592 A CN 111700592A
- Authority
- CN
- China
- Prior art keywords
- neural network
- classification
- convolutional neural
- electroencephalogram
- network model
- 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.)
- Pending
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
- A61B5/40—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
- 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
- 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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR 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
Landscapes
- Health & Medical Sciences (AREA)
- Engineering & Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Physics & Mathematics (AREA)
- Artificial Intelligence (AREA)
- Biophysics (AREA)
- Theoretical Computer Science (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Biomedical Technology (AREA)
- Mathematical Physics (AREA)
- Neurology (AREA)
- Evolutionary Computation (AREA)
- Software Systems (AREA)
- Physiology (AREA)
- General Engineering & Computer Science (AREA)
- Computing Systems (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- Neurosurgery (AREA)
- Public Health (AREA)
- Computational Linguistics (AREA)
- Veterinary Medicine (AREA)
- Pathology (AREA)
- Heart & Thoracic Surgery (AREA)
- Medical Informatics (AREA)
- Surgery (AREA)
- Animal Behavior & Ethology (AREA)
- Fuzzy Systems (AREA)
- Signal Processing (AREA)
- Psychiatry (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Measurement And Recording Of Electrical Phenomena And Electrical Characteristics Of The Living Body (AREA)
Abstract
The invention relates to a method and a system for acquiring an epilepsia electroencephalogram automatic classification model, wherein the system comprises an effective electroencephalogram signal acquisition module, a time-frequency diagram acquisition module, a classification accuracy calculation module and an output module; the invention also discloses an automatic classification system of the epilepsia electroencephalogram signals, which comprises an effective electroencephalogram signal acquisition module, a time-frequency diagram acquisition module, a classification module and an output module; according to the system and the method, the time-frequency diagram is subjected to automatic feature extraction by adopting transfer learning, so that the workload of parameter debugging and parameter learning is reduced on the basis of reducing the time for processing a large number of features, the time for building and training a network model is greatly saved, and the efficiency of the system and the model for classifying the electroencephalogram signals is improved.
Description
Technical Field
The invention relates to the field of biomedical engineering signal processing, in particular to an acquisition method, a system and a classification system of an epilepsia electroencephalogram automatic classification model.
Background
Epilepsy is a common brain disease. The world health organization counts in 2019 in 6 months, about 5000 million epileptic patients account for about 0.6-0.8% of the total population worldwide, and the epileptic diseases increase at the rate of 240 million every year, have large age span and tend to be young, and seriously threaten the healthy development of human beings. Among the epileptic population, teenagers and children under 20 years of age become the high-haired population. The causes of epilepsy are complex, the pathogenesis of epilepsy has not yet been fully elucidated, and structural or metabolic abnormalities that explain the corresponding symptoms cannot be found from the brain of a patient in many cases. However, the electroencephalogram signals of epileptic patients contain a great deal of pathological information of the brain. Electroencephalography is a low-cost, non-invasive tool that can be used for long-term evaluation. Therefore, electroencephalography is the most useful tool for diagnosing epilepsy. At present, the diagnosis of epilepsy mainly depends on the examination and analysis of electroencephalograms by doctors, is time-consuming and labor-consuming, and has the problem of misjudgment, so that the automatic identification and classification of electroencephalogram signals of epilepsy have important significance for the detection of epilepsy, the burden of doctors can be reduced to a great extent, and the diagnosis efficiency can be improved.
The electroencephalogram signal is a random non-stationary signal, and the noise background is strong, and the signal is weak, so that certain difficulty exists in extracting the characteristics of the signal. The existing systems are analyzed from a time domain or a frequency domain only, the effect is not ideal, and the characteristics of the epileptic signal cannot be fully expressed due to the randomness and the non-stationarity of the signal. And the time-frequency analysis can fully reserve the time and frequency information of the signal. Therefore, the system adopting time-frequency analysis analyzes the epilepsia electroencephalogram signals, and can extract a plurality of epilepsia electroencephalogram related characteristics.
According to the traditional epilepsy electroencephalogram processing model and system based on time-frequency analysis, a time-frequency graph of an epilepsy electroencephalogram signal is obtained, numerous characteristics are extracted, and classification is carried out. However, while the classification accuracy is high and the feature extraction is less, the processing calculation time is reduced, and the two methods are difficult to balance. Therefore, the scheme of the invention adopts machine learning to automatically extract and classify the time-frequency diagram obtained by processing the signals, and realizes the classification and automatic identification of the epileptic electroencephalogram, thereby reducing the burden of doctors and improving the efficiency of the system for diagnosing the electroencephalogram.
Disclosure of Invention
The invention aims to provide an acquisition method, a system and a classification system of an epilepsia electroencephalogram automatic classification model, which are used for automatically extracting characteristics of a time-frequency graph by adopting transfer learning, reduce the workload of parameter debugging and parameter learning on the basis of reducing the processing time of a large number of characteristics, and greatly save the time for building and training a network model, thereby improving the efficiency of the system and the model for classifying electroencephalograms.
In order to achieve the purpose, the invention provides the following scheme:
an epilepsia electroencephalogram automatic classification system comprises an effective electroencephalogram signal acquisition module, a time-frequency diagram acquisition module, a classification module and an output module;
the effective electroencephalogram signal acquisition module is used for performing wavelet transformation on electroencephalogram signals to be classified to obtain electroencephalogram signals in an effective frequency band range and recording the electroencephalogram signals as effective electroencephalogram signals;
the time-frequency image acquisition module is used for carrying out short-time Fourier transform on the effective electroencephalogram signal to obtain a time-frequency image of reaction time and frequency;
the classification module is used for classifying the time-frequency diagram by utilizing a TensorFlow frame to obtain a model classification result; the TensorFlow frame utilizes an initial inclusion-V3 convolutional neural network model to perform transfer learning on the time-frequency diagram to obtain a feature vector, and then the feature vector is input into the optimized inclusion-V3 convolutional neural network model for classification;
the output module is used for outputting the classification result of the optimized inclusion-V3 convolutional neural network model.
A method for acquiring an epileptic brain electrical automatic classification model, the method comprising:
performing multi-layer wavelet decomposition on the electroencephalogram signals to be classified to obtain electroencephalogram signals in an effective frequency band range, and recording the electroencephalogram signals as effective electroencephalogram signals; the number of layers of the wavelet decomposition is determined by the sampling frequency of the electroencephalogram signals to be classified;
carrying out short-time Fourier transform on the effective electroencephalogram signal to obtain a time-frequency diagram of reaction time and frequency;
classifying the time-frequency diagram by using a TensorFlow frame to obtain a model classification result; the TensorFlow frame utilizes an initial inclusion-V3 convolutional neural network model to perform transfer learning on the time-frequency diagram to obtain a feature vector, and then the feature vector is input into the optimized inclusion-V3 convolutional neural network model for classification;
calculating the classification accuracy of the optimized inclusion-V3 convolutional neural network model according to the model classification result, judging whether the classification accuracy reaches a preset threshold value, and stopping training when the classification accuracy reaches the preset threshold value to obtain a trained electroencephalogram epilepsy automatic classification model; otherwise, continuing training the optimized inclusion-V3 convolutional neural network model until the classification accuracy reaches a preset threshold.
An acquisition system of an epilepsia electroencephalogram automatic classification model comprises an effective electroencephalogram signal acquisition module, a time-frequency diagram acquisition module, a classification accuracy calculation module and an output module;
the effective electroencephalogram signal acquisition module is used for carrying out multilayer wavelet decomposition on electroencephalogram signals to be classified to obtain electroencephalogram signals within an effective frequency band range and recording the electroencephalogram signals as effective electroencephalogram signals; the number of layers of the wavelet decomposition is determined by the sampling frequency of the electroencephalogram signals to be classified;
the time-frequency image acquisition module is used for carrying out short-time Fourier transform on the effective electroencephalogram signal to obtain a time-frequency image of reaction time and frequency;
the classification module is used for classifying the time-frequency diagram by utilizing a TensorFlow frame to obtain a model classification result; the TensorFlow frame utilizes an initial inclusion-V3 convolutional neural network model to perform transfer learning on the time-frequency diagram to obtain a feature vector, and then the feature vector is input into the optimized inclusion-V3 convolutional neural network model for classification;
the classification accuracy calculation module is used for calculating the classification accuracy of the optimized inclusion-V3 convolutional neural network model according to the model classification result, judging whether the classification accuracy reaches a preset threshold value, and stopping training when the classification accuracy reaches the preset threshold value to obtain a trained epilepsia electroencephalogram automatic classification model; otherwise, continuing training the optimized inclusion-V3 convolutional neural network model until the classification accuracy reaches a preset threshold;
the output module is used for outputting the classification accuracy of the optimized inclusion-V3 convolutional neural network model.
According to the specific embodiment provided by the invention, the invention discloses the following technical effects:
1. according to the method and the system for obtaining the epilepsia electroencephalogram automatic classification model, the inclusion-V3 convolutional neural network model is adopted for transfer learning, the complexity of the training model is effectively reduced, the time for building and training the model is saved, the efficiency of the system for electroencephalogram diagnosis is improved, and particularly, the advantages are more obvious under the condition that test data are insufficient.
2. According to the automatic epilepsia electroencephalogram classification system, the electroencephalogram signals to be classified are actually classified by adopting the trained automatic epilepsia electroencephalogram classification model, so that the electroencephalogram diagnosis efficiency of the system is improved.
3. According to the method and the system for acquiring the epilepsia electroencephalogram automatic classification model, the time-frequency diagram of the electroencephalogram signal is extracted through the time-frequency diagram acquisition module, and meanwhile, the time-frequency diagram and the frequency information of the signal are contained, so that the method and the system are more comprehensive compared with the prior art that only the time information or the frequency information is extracted.
4. According to the method and the system for acquiring the epilepsia electroencephalogram automatic classification model, the time-frequency diagram of the electroencephalogram signals is extracted through the time-frequency diagram acquisition module, the time and the frequency of the signals are included, the complex artificial feature extraction is omitted, and a machine learning system is adopted for automatic feature extraction learning, so that the method and the system are simple and effective.
5. The epilepsia electroencephalogram signal automatic classification model is used for classification, and good classification effect can be obtained easily.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required to be used in the embodiments will be briefly described below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and it is obvious for those skilled in the art that other drawings can be obtained according to the drawings without inventive labor.
Fig. 1 shows a block diagram of a terminal device applicable to all embodiments of the present invention;
fig. 2 is a flowchart of an acquiring method of an epilepsia electroencephalogram automatic classification model according to an embodiment of the present invention;
fig. 3 is a schematic diagram of decomposition of 3 layers of wavelets in the method for acquiring an epileptic brain electrical automatic classification model according to an embodiment of the present invention;
fig. 4 is a flowchart of classifying a time-frequency diagram by using a tensrflow frame in the method for acquiring an epileptic electroencephalogram automatic classification model according to the embodiment of the present invention, so as to obtain a model classification result;
fig. 5 is a schematic structural diagram of an acquiring system of an epilepsia electroencephalogram automatic classification model according to a second embodiment of the present invention;
FIG. 6 is a schematic structural diagram of a classification module in the second and third embodiments of the present invention;
fig. 7 is a schematic structural diagram of an epilepsia electroencephalogram automatic classification system provided by a third embodiment of the present invention;
fig. 8 is a schematic structural diagram of an effective electroencephalogram signal acquisition module in the automatic epileptic electroencephalogram classification system provided by the third embodiment of the present invention;
description of the symbols: the method comprises the following steps of 1-an effective electroencephalogram signal acquisition module, 2-a time-frequency diagram acquisition module, 3-a classification module, 4-a classification accuracy calculation module, 5-an output module, 11-an electroencephalogram signal import unit, 12-a wavelet decomposition unit, 31-a data set division unit, 32-a feature extraction unit, 33-a model optimization unit, 34-a model training unit, 110-a memory, 120-a storage controller, 130-a processor, 140-a peripheral interface and 150-an input and output unit.
Detailed Description
The technical solutions in the embodiments of the present invention will be described clearly and completely with reference to the accompanying drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
The invention aims to provide an acquisition method, a system and a classification system of an epilepsia electroencephalogram automatic classification model, which are used for automatically extracting characteristics of a time-frequency graph by adopting transfer learning, reduce the workload of parameter debugging and parameter learning on the basis of reducing the processing time of a large number of characteristics, and greatly save the time for building and training a network model, thereby improving the efficiency of the system and the model for classifying electroencephalograms.
In order to make the aforementioned objects, features and advantages of the present invention comprehensible, embodiments accompanied with figures are described in further detail below.
Fig. 1 shows a block schematic diagram of a terminal device that can be applied to embodiments one, two and three of the present invention. The terminal device 10 includes a memory 110, a memory controller 120, a processor 130, a peripheral interface 140, and an input-output unit 150. For example, the terminal device 10 may be a Personal Computer (PC), a tablet PC, a smart phone, a Personal Digital Assistant (PDA), or the like.
The memory 110, the memory controller 120, the processor 130, the peripheral interface 140, and the input/output unit 150 are electrically connected to each other directly or indirectly to realize data transmission or interaction. For example, the components may be electrically connected to each other via one or more communication buses or signal lines.
The memory 110 may be configured to store a software program and a module, such as a program instruction/module corresponding to the method, system and classification system for acquiring an epileptic electroencephalogram signal automatic classification model in the embodiment of the present invention, and the processor 130 executes various functional applications and data processing by operating the software program and the module stored in the memory 110, that is, implements the method for acquiring an epileptic electroencephalogram signal automatic classification model in the embodiment of the present invention. The memory 110 may include high speed random access memory, and may also include non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory.
The processor 130 may be an integrated circuit chip having signal processing capabilities. The Processor 130 may be a general-purpose Processor, and includes a Central Processing Unit (CPU), a Network Processor (NP), and the like; but may also be a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), an off-the-shelf programmable gate array (FPGA) or other programmable logic device, discrete gate or transistor logic, discrete hardware components. The various methods, steps and logic blocks disclosed in the embodiments of the present invention may be implemented or performed. A general purpose processor may be a microprocessor or the processor may be any conventional processor or the like.
The peripheral interface 140 couples various input/output devices to the processor 130 as well as to the memory 110. In some embodiments, the peripheral interface 140, the processor 130, and the memory controller 110 may be implemented in a single chip. In other examples, they may be implemented separately from the individual chips.
The input/output unit 150 is used for providing input data for a user to realize the interaction of the user with the terminal device 10. The input/output unit 150 may be, but is not limited to, a mouse, a keyboard, and the like.
It will be appreciated that the configuration shown in fig. 1 is merely illustrative and that the terminal device 10 may include more or fewer components than shown in fig. 1 or may have a different configuration than shown in fig. 1. The components shown in fig. 1 may be implemented in hardware, software, or a combination thereof.
Example one
As shown in fig. 2, the method for acquiring an epileptic electroencephalogram automatic classification model shown in this embodiment is specifically implemented according to the following method:
performing multilayer wavelet decomposition on an electroencephalogram signal to be classified to obtain an electroencephalogram signal in an effective frequency band range, and recording the electroencephalogram signal as an effective electroencephalogram signal; the number of layers of wavelet decomposition is determined by the sampling frequency of the electroencephalogram signals to be classified;
specifically, the electroencephalogram data to be classified used in this embodiment is derived from the CHB-MIT scalp electroencephalogram database, which is collected in the Boston Children hospital and consists of electroencephalogram records of pediatric patients with refractory seizures.
The sampling frequency of the electroencephalogram signals in the embodiment is 256Hz, and the electrode placement mode is international 10-20 electrode positions. And (3) leading the electroencephalogram signals obtained from the database into MATLAB, and reading signals of the epileptic onset period of the patient. Because the effective frequency band of the brain electrical signal is within 0.5-32Hz and the sampling frequency of the signal is 256Hz, 3-layer wavelet decomposition is carried out on the signal. The number of layers of the wavelet decomposition will also adapt to the change for different sampling frequencies until the decomposition reaches the range of the effective band. And in the decomposition process, the high-frequency part is not processed, the low-frequency part is continuously decomposed, but the decomposition degree of the low-frequency part is half of the original decomposition degree, so that the signal is decomposed layer by layer, the signal is changed into two groups of coefficients, a detail coefficient and an approximate coefficient, the detail coefficient represents the high-frequency information of the signal, and the approximate coefficient represents the low-frequency information of the signal.
As shown in FIG. 3, the low frequency of the first layer is 0-128Hz, the low frequency of the second layer is 0-64Hz, the low frequency of the third layer is 0-32Hz, the low frequency coefficient of the third layer is reserved, effective signals within 32Hz are obtained, the filtering effect on high-frequency noise is achieved, and useful electroencephalogram signals are effectively extracted. The db4 wavelet waveform has the highest approximation degree with the spike wave in the epileptic characteristic wave, so the db4 wavelet is adopted as the mother wavelet in the wavelet decomposition.
Secondly, carrying out short-time Fourier transform on the effective electroencephalogram signals to obtain a time-frequency graph of reaction time and frequency;
because the time-frequency diagram contains time information and frequency information, the signal characteristics can be more represented by adopting the time-frequency diagram for analysis than the signal characteristics can be represented by adopting pure time domain or frequency domain analysis.
In the embodiment, the short-time fourier transform time window function adopts hamming, the window width is 70, signals move for 2 seconds each time, fourier transform is performed every 2 seconds, and a time-frequency graph is output. And storing the time-frequency diagram of the epileptic brain electricity signal and the time-frequency diagram of the normal brain electricity signal separately.
Thirdly, classifying the time-frequency diagram by using a TensorFlow frame to obtain a model classification result; the TensorFlow framework is used for carrying out transfer learning on the time-frequency diagram by utilizing an initial inclusion-V3 convolutional neural network model to obtain a characteristic vector, and then inputting the characteristic vector into the optimized inclusion-V3 convolutional neural network model for classification;
as shown in fig. 4, classifying the time-frequency diagram by using the tensrflow frame to obtain a model classification result specifically includes:
s301, forming a data set by a time-frequency graph of the electroencephalogram signals in the epileptic seizure period and a time-frequency graph of the normal electroencephalogram signals, dividing the data set into a training set, a verification set and a testing set by adopting a random function, and adding labels to the data of the training set, the testing set and the verification set, wherein the label of the time-frequency graph of the electroencephalogram signals in the epileptic seizure period is 1, and the label of the time-frequency graph of the normal electroencephalogram signals is 0.
When the picture names are separated according to a training set, a verification set and a test set, a random function is adopted to randomly generate a score within 100, and the name of the picture is judged to be classified into which category according to a randomly obtained score value (score): score <10: validation set; score <20: test set; else, training set.
S302, calling an initial inclusion-V3 convolutional neural network model to perform transfer learning feature processing on the time-frequency graph in the data set and outputting feature vectors.
When the initial inclusion-V3 model is called to perform feature calculation on all the time-frequency graphs, the finally generated feature vector is [ None, 2048], wherein None is the number of the time-frequency graphs, and 2048 is 2048 feature values obtained by calculating the time-frequency graphs by the model. [ None, 2048] is taken as input to the fully connected layer and classified.
Among them, the inclusion-V3 model is well-trained by Google based on ImageNet dataset. The Incep-V3 convolutional neural network model has 46 layers, 11 Incep modules and 96 convolutional layers, and as the number of model layers of the convolutional neural network increases and the complexity increases, more and more labeled data are needed for training the model. For example, the ResNet model, which has 152 layers deep, is trained using 120 million images with labels in the ImageNet dataset to achieve 96.5% accuracy. Although this is a good result, it takes a long time to train a complex convolutional neural network using this data considering that it is very difficult to collect this much picture data in real applications and even if it takes much manpower and material resources to collect this much picture data. Based on the above consideration, the embodiment adopts the inclusion-V3 model trained by Google to perform the feature processing of transfer learning on the time-frequency diagram.
S303, optimizing the initial Incepration-V3 convolutional neural network model, and replacing the full-connection layer into two categories to obtain the optimized Incepration-V3 convolutional neural network model.
The initial inclusion-V3 convolutional neural network model is a 1000-class classification, and this embodiment classifies the processed epilepsy electroencephalogram time-frequency map to distinguish epilepsy from a normal state, so that the last full-link layer needs to be replaced, and the classification is changed into a two-class classification. Therefore, the process of model optimization is a process of replacing the fully connected layer with two classes. For the replaced fully connected layer, a network layer before the replaced fully connected layer is called a Bottleneck layer (Bottleneck corresponds to the last Dropout layer in the Incep-V3 model framework, the third last layer is a Dropout layer, and the second last layer is a fully connected layer, which needs to be replaced). In the inclusion-V3 model, a Dropout layer is connected to a single-layer fully-connected neural network, and after some operations of the fully-connected neural network layer, the results are processed by a Softmax layer to be classified, but in the embodiment, the fully-connected layer needs to be replaced by two classifications.
S304, training, verifying and testing the optimized inclusion-V3 convolutional neural network model by utilizing a training set, a verifying set and a testing set, and training the optimized inclusion-V3 convolutional neural network model by using a cross entropy loss function and a random gradient descent algorithm in the training process.
Wherein, the cross entropy formula:
the cross entropy describes how accurately probability distribution Q estimates probability distribution P, so when using the cross entropy loss function, it is generally assumed that P represents the correct answer, Q represents the predicted result value, and x is the event of calculating the probability. Therefore, in this embodiment, P represents an actual classification result, Q represents a classification result of the optimized inclusion-V3 convolutional neural network model, and x is an accuracy rate of classifying the epileptic electroencephalogram signal by the optimized inclusion-V3 convolutional neural network model.
Solving the total loss function in a machine learning algorithm can generally be understood as calculating the sum of the loss functions for all samples. For example, the cross entropy of each sample is lost by-p (x)i)log q(xi) Is denoted as L (p)i,qiω), then for the total loss function, the gradient descent needs to be calculated:
as the size of the training data set increases, the calculation of the total loss function needs to pay more operation cost, and the calculation of each step of gradient consumes a quite long time. The core of the gradient descent algorithm is to approximate the gradient by using small-scale samples. Specifically, at each step of the algorithm, we uniformly extract a small batch of samples X' ═ { X ] from the training set1,x2,x3,L,xi′}, small batchIs typically a relatively small number, from one to several hundred. Importantly, as the training set size i grows, i' is typically fixed. Therefore, we may fit billions of samples, using only a few hundred samples per update calculation. The estimation of the gradient can be expressed as:
using samples from the small lot X', the random gradient descent algorithm will use the following gradient descent:
ω-σ·g→ωi′
wherein σ is the learning rate; ω is an input parameter to a function, and in deep neural networks, ω generally refers to a parameter in the neural network.
Calculating the classification accuracy of the optimized inclusion-V3 convolutional neural network model according to the model classification result, judging whether the classification accuracy reaches a preset threshold value, and stopping training when the classification accuracy reaches the preset threshold value to obtain a trained electroencephalogram automatic classification model for epilepsy; otherwise, continuing training the optimized inclusion-V3 convolutional neural network model until the classification accuracy reaches a preset threshold.
The classification accuracy of the optimized inclusion-V3 convolutional neural network model is calculated through the accuracy, the recall rate and the accuracy;
the accuracy calculation formula is as follows:
the ACC represents the classification accuracy of the optimized inclusion-V3 convolutional neural network model, and the TP is a true case and represents the number of correctly classified electroencephalogram signals in the epileptic seizure period; TN is a true negative example, which represents the number of correctly classified samples of normal electroencephalogram signals; FP is a false positive example and represents the number of samples of the electroencephalogram signals in the epileptic seizure period which are classified in error; FN is false negative, and represents the number of samples of misclassified normal brain wave number;
the recall ratio calculation formula is:
SEN represents the classification recall rate of the optimized inclusion-V3 convolutional neural network model;
the accuracy calculation formula is as follows:
wherein PRE represents the classification accuracy of the optimized inclusion-V3 convolutional neural network model.
Example two
As shown in fig. 6, the system for acquiring an epileptic brain electrical automatic classification model shown in this embodiment includes: the device comprises an effective electroencephalogram signal acquisition module 1, a time-frequency diagram acquisition module 2, a classification module 3, a classification accuracy calculation module 4 and an output module 5.
The effective electroencephalogram signal acquisition module 1 is used for performing multi-layer wavelet decomposition on electroencephalogram signals to be classified to obtain electroencephalogram signals within an effective frequency band range, and recording the electroencephalogram signals as effective electroencephalogram signals. The number of layers of wavelet decomposition is determined by the sampling frequency of the electroencephalogram signals to be classified.
The time-frequency image acquisition module 2 is used for carrying out short-time Fourier transform on the effective electroencephalogram signal to obtain a time-frequency image of the reaction time and the frequency.
And the classification module 3 is used for classifying the time-frequency diagram by utilizing a Tensorflow frame to obtain a model classification result. The TensorFlow frame utilizes an initial inclusion-V3 convolutional neural network model to perform transfer learning on a time-frequency graph to obtain a feature vector, and then the feature vector is input into the optimized inclusion-V3 convolutional neural network model to be classified;
as shown in fig. 7, as an alternative embodiment, the classification module 3 includes a data set partitioning unit 31, a feature extraction unit 32, a model optimization unit 33, and a model training unit 34;
the data set dividing unit 31 is configured to combine the time-frequency diagram of the electroencephalogram signal in the epileptic seizure stage and the time-frequency diagram of the normal electroencephalogram signal into a data set, and divide the data set into a training set, a verification set and a test set by using a random function.
The data set dividing unit 31 is further configured to add a label to the training set, the test set, and the verification set, where the label of the time-frequency diagram of the electroencephalogram in the epileptic seizure period is 1, and the label of the time-frequency diagram of the normal electroencephalogram signal is 0.
The feature extraction unit 32 is configured to invoke an initial inclusion-V3 convolutional neural network model to perform feature processing of transfer learning on a time-frequency graph in a data set and output a feature vector;
the model optimization unit 33 is used for optimizing the initial inclusion-V3 convolutional neural network model, and replacing the full connection layer with two classes to obtain the optimized inclusion-V3 convolutional neural network model.
The model training unit 34 is configured to train, verify and test the optimized inclusion-V3 convolutional neural network model by using a training set, a verification set and a test set, and the model training unit 34 trains the optimized inclusion-V3 convolutional neural network model by using a cross entropy loss function and a random gradient descent algorithm in a training process to obtain a classification result of the optimized inclusion-V3 convolutional neural network model.
The classification accuracy calculation module 4 is used for calculating the classification accuracy of the optimized inclusion-V3 convolutional neural network model according to the model classification result, judging whether the classification accuracy reaches a preset threshold value, and stopping training when the classification accuracy reaches the preset threshold value to obtain a trained epilepsia electroencephalogram automatic classification model; otherwise, continuing training the optimized inclusion-V3 convolutional neural network model until the classification accuracy reaches a preset threshold;
as an optional implementation manner, the classification accuracy calculation module 4 calculates the classification accuracy of the optimized inclusion-V3 convolutional neural network model through the accuracy, the recall rate and the precision;
the output module 5 is used for outputting the classification accuracy of the optimized inclusion-V3 convolutional neural network model.
It should be noted that, for the system disclosed in the second embodiment, since it corresponds to the method disclosed in the first embodiment, the description is relatively simple, and for the relevant points, refer to the description of the method part.
EXAMPLE III
As shown in fig. 8, the automatic epilepsia electroencephalogram classification system shown in this embodiment includes: the device comprises an effective electroencephalogram signal acquisition module 1, a time-frequency diagram acquisition module 2, a classification module 3 and an output module 5.
The effective electroencephalogram signal acquisition module 1 is used for performing wavelet transformation on an electroencephalogram signal to be classified to obtain an electroencephalogram signal in an effective frequency band range, and recording the electroencephalogram signal as an effective electroencephalogram signal.
As an alternative embodiment, as shown in fig. 8, the effective electroencephalogram signal acquisition module 1 includes a electroencephalogram signal introduction unit 11 and a wavelet decomposition unit 12.
The electroencephalogram signal importing unit 11 is used for importing the electroencephalogram signals to be classified into the wavelet decomposition unit 12.
In this embodiment, the electroencephalograms to be classified are from the CHB-MIT scalp electroencephalogram database, which is collected in boston children's hospital and is composed of electroencephalogram records of pediatric patients with refractory seizures.
The wavelet decomposition unit 12 is used for performing multi-layer wavelet decomposition on the electroencephalogram signals to be classified to obtain effective electroencephalogram signals; the number of layers of wavelet decomposition is determined by the sampling frequency of the electroencephalogram signals to be classified.
The time-frequency image acquisition module 2 is used for carrying out short-time Fourier transform on the effective electroencephalogram signal to obtain a time-frequency image of the reaction time and the frequency.
And the classification module 3 is used for classifying the time-frequency diagram by utilizing a Tensorflow frame to obtain a model classification result. The TensorFlow frame utilizes an initial inclusion-V3 convolutional neural network model to perform transfer learning on a time-frequency graph to obtain a feature vector, and then the feature vector is input into the optimized inclusion-V3 convolutional neural network model to be classified;
the classification module 3 comprises a data set dividing unit 31, a feature extraction unit 32, a model optimization unit 33 and a model training unit 34;
the data set dividing unit 31 is configured to combine a time-frequency diagram of the electroencephalogram signal in the epileptic seizure stage and a time-frequency diagram of the normal electroencephalogram signal into a data set, and divide the data set into a training set, a verification set, and a test set by using a random function.
The data set dividing unit 31 is further configured to add a label to the training set, the test set, and the verification set, where the label of the time-frequency diagram of the electroencephalogram in the epileptic seizure period is 1, and the label of the time-frequency diagram of the normal electroencephalogram signal is 0, so as to allow the neural network to perform identification and classification.
The feature extraction unit 32 is configured to invoke an initial inclusion-V3 convolutional neural network model to perform feature processing of transfer learning on a time-frequency graph in a data set and output a feature vector;
the model optimization unit 33 is used for optimizing the initial inclusion-V3 convolutional neural network model, and replacing the full connection layer with two classes to obtain the optimized inclusion-V3 convolutional neural network model.
The model training unit 34 is configured to train, verify and test the optimized inclusion-V3 convolutional neural network model by using a training set, a verification set and a test set, and the model training unit 34 trains the optimized inclusion-V3 convolutional neural network model by using a cross entropy loss function and a random gradient descent algorithm in a training process to obtain a classification result of the optimized inclusion-V3 convolutional neural network model.
As an optional implementation manner, the model optimization unit 33 is further configured to invoke the classification result to calculate the classification accuracy of the optimized inclusion-V3 convolutional neural network model; the model optimization unit 33 calculates the classification accuracy of the optimized inclusion-V3 convolutional neural network model according to the accuracy, the recall rate and the precision;
the output module 5 is used for outputting the classification result of the optimized inclusion-V3 convolutional neural network model.
According to the invention, a large amount of feature extraction of epilepsia electroencephalogram signals is abandoned, and the time-frequency graph is subjected to automatic feature extraction by adopting transfer learning, so that the workload of parameter debugging and parameter learning is reduced on the basis of reducing the processing time of a large amount of features, the time for building and training a network model is greatly saved, and the effectiveness of the invention is verified.
For the system disclosed in the third embodiment, it is actually an application system for actually classifying the electroencephalogram signal to be classified by using the optimized inclusion-V3 convolutional neural network model obtained in the first embodiment and the second embodiment, so that the same parts can be referred to each other.
The principles and embodiments of the present invention have been described herein using specific examples, which are provided only to help understand the method and the core concept of the present invention; meanwhile, for a person skilled in the art, according to the idea of the present invention, the specific embodiments and the application range may be changed. In view of the above, the present disclosure should not be construed as limiting the invention.
Claims (10)
1. An epilepsia electroencephalogram automatic classification system is characterized by comprising an effective electroencephalogram signal acquisition module, a time-frequency diagram acquisition module, a classification module and an output module;
the effective electroencephalogram signal acquisition module is used for performing wavelet transformation on electroencephalogram signals to be classified to obtain electroencephalogram signals in an effective frequency band range and recording the electroencephalogram signals as effective electroencephalogram signals;
the time-frequency image acquisition module is used for carrying out short-time Fourier transform on the effective electroencephalogram signal to obtain a time-frequency image of reaction time and frequency;
the classification module is used for classifying the time-frequency diagram by utilizing a TensorFlow frame to obtain a model classification result; the TensorFlow frame utilizes an initial inclusion-V3 convolutional neural network model to perform transfer learning on the time-frequency diagram to obtain a feature vector, and then the feature vector is input into the optimized inclusion-V3 convolutional neural network model for classification;
the output module is used for outputting the classification result of the optimized inclusion-V3 convolutional neural network model.
2. The system for automatically classifying epilepsia electroencephalogram according to claim 1, wherein the effective electroencephalogram signal acquisition module comprises an electroencephalogram signal importing unit and a wavelet decomposition unit;
the electroencephalogram signal leading-in unit is used for leading the electroencephalogram signals to be classified into the wavelet decomposition unit;
the wavelet decomposition unit is used for carrying out multilayer wavelet decomposition on the electroencephalogram signals to be classified to obtain the effective electroencephalogram signals; the number of layers of the wavelet decomposition is determined by the sampling frequency of the electroencephalogram signals to be classified.
3. The system for automatically classifying epilepsia electroencephalogram according to claim 1, wherein the classification module comprises a data set partitioning unit, a feature extraction unit, a model optimization unit and a model training unit;
the data set dividing unit is used for forming a data set by a time-frequency graph of the electroencephalogram signals in the epileptic seizure period and a time-frequency graph of the normal electroencephalogram signals, dividing the data set into a training set, a verification set and a test set by adopting a random function, and adding labels to the data of the training set, the test set and the verification set, wherein the label of the time-frequency graph of the electroencephalogram signals in the epileptic seizure period is 1, and the label of the time-frequency graph of the normal electroencephalogram signals is 0;
the feature extraction unit is used for calling an initial inclusion-V3 convolutional neural network model to perform transfer learning feature processing on the time-frequency graph in the data set and outputting a feature vector;
the model optimization unit is used for optimizing the initial inclusion-V3 convolutional neural network model, replacing the full connection layer with two classes to obtain the optimized inclusion-V3 convolutional neural network model;
the model training unit is used for training, verifying and testing the optimized inclusion-V3 convolutional neural network model by utilizing the training set, the verifying set and the testing set, and the model training unit uses a cross entropy loss function and a random gradient descent algorithm to train the optimized inclusion-V3 convolutional neural network model in the training process to obtain the classification result of the optimized inclusion-V3 convolutional neural network model.
4. The system of claim 3, wherein the model optimization unit is further configured to call the classification result to calculate the classification accuracy of the optimized inclusion-V3 convolutional neural network model; the model optimization unit is used for calculating the classification accuracy of the optimized inclusion-V3 convolutional neural network model according to the accuracy, the recall rate and the precision;
the accuracy calculation formula is as follows:
the ACC represents the classification accuracy of the optimized inclusion-V3 convolutional neural network model, and the TP is a real example and represents the number of correctly classified electroencephalogram signals in the epileptic seizure period; TN is a true negative example and represents the number of correctly classified samples of the normal electroencephalogram signals; FP is a false positive case and represents the number of samples of the electroencephalogram signals in the epileptic seizure period which are classified in error; FN is false negative example, and represents the number of samples of the normal electroencephalogram signals which are classified in error;
the recall ratio calculation formula is as follows:
wherein SEN represents the classification recall rate of the optimized inclusion-V3 convolutional neural network model;
the accuracy calculation formula is as follows:
wherein PRE represents the classification accuracy of the optimized inclusion-V3 convolutional neural network model.
5. A method for acquiring an epilepsia electroencephalogram automatic classification model is characterized by comprising the following steps:
performing multi-layer wavelet decomposition on the electroencephalogram signals to be classified to obtain electroencephalogram signals in an effective frequency band range, and recording the electroencephalogram signals as effective electroencephalogram signals; the number of layers of the wavelet decomposition is determined by the sampling frequency of the electroencephalogram signals to be classified;
carrying out short-time Fourier transform on the effective electroencephalogram signal to obtain a time-frequency diagram of reaction time and frequency;
classifying the time-frequency diagram by using a TensorFlow frame to obtain a model classification result; the TensorFlow frame utilizes an initial inclusion-V3 convolutional neural network model to perform transfer learning on the time-frequency diagram to obtain a feature vector, and then the feature vector is input into the optimized inclusion-V3 convolutional neural network model for classification;
calculating the classification accuracy of the optimized inclusion-V3 convolutional neural network model according to the model classification result, judging whether the classification accuracy reaches a preset threshold value, and stopping training when the classification accuracy reaches the preset threshold value to obtain a trained electroencephalogram epilepsy automatic classification model; otherwise, continuing training the optimized inclusion-V3 convolutional neural network model until the classification accuracy reaches a preset threshold.
6. The method for acquiring the epilepsia electroencephalogram automatic classification model, according to claim 5, is characterized in that the classifying the time-frequency diagram by using a Tensorflow framework to obtain a model classification result specifically comprises:
forming a data set by a time-frequency graph of an electroencephalogram signal in a seizure period and a time-frequency graph of a normal electroencephalogram signal, dividing the data set into a training set, a verification set and a test set by adopting a random function, and adding labels to the training set, the test set and the verification set, wherein the label of the time-frequency graph of the electroencephalogram signal in the seizure period is 1, and the label of the time-frequency graph of the normal electroencephalogram signal is 0;
calling an initial inclusion-V3 convolutional neural network model to perform transfer learning feature processing on the time-frequency graph in the data set and outputting a feature vector;
optimizing the initial Incepration-V3 convolutional neural network model, and replacing the full-connection layer with two classes to obtain the optimized Incepration-V3 convolutional neural network model;
and training, verifying and testing the optimized inclusion-V3 convolutional neural network model by utilizing the training set, the verifying set and the testing set, and training the optimized inclusion-V3 convolutional neural network model by utilizing a cross entropy loss function and a random gradient descent algorithm in the training process.
7. The method for acquiring the epilepsia electroencephalogram automatic classification model, according to claim 5, is characterized in that the classification accuracy of the optimized inclusion-V3 convolutional neural network model is calculated through a correct rate, a recall rate and an accuracy rate;
the accuracy calculation formula is as follows:
the ACC represents the classification accuracy of the optimized inclusion-V3 convolutional neural network model, and the TP is a real example and represents the number of correctly classified electroencephalogram signals in the epileptic seizure period; TN is a true negative example and represents the number of correctly classified samples of the normal electroencephalogram signals; FP is a false positive case and represents the number of samples of the electroencephalogram signals in the epileptic seizure period which are classified in error; FN is false negative example, and represents the number of samples of the normal electroencephalogram signals which are classified in error;
the recall ratio calculation formula is as follows:
wherein SEN represents the classification recall rate of the optimized inclusion-V3 convolutional neural network model;
the accuracy calculation formula is as follows:
wherein PRE represents the classification accuracy of the optimized inclusion-V3 convolutional neural network model.
8. The system for acquiring the epilepsia electroencephalogram automatic classification model is characterized by comprising an effective electroencephalogram signal acquisition module, a time-frequency image acquisition module, a classification accuracy calculation module and an output module;
the effective electroencephalogram signal acquisition module is used for carrying out multilayer wavelet decomposition on electroencephalogram signals to be classified to obtain electroencephalogram signals within an effective frequency band range and recording the electroencephalogram signals as effective electroencephalogram signals; the number of layers of the wavelet decomposition is determined by the sampling frequency of the electroencephalogram signals to be classified;
the time-frequency image acquisition module is used for carrying out short-time Fourier transform on the effective electroencephalogram signal to obtain a time-frequency image of reaction time and frequency;
the classification module is used for classifying the time-frequency diagram by utilizing a TensorFlow frame to obtain a model classification result; the TensorFlow frame utilizes an initial inclusion-V3 convolutional neural network model to perform transfer learning on the time-frequency diagram to obtain a feature vector, and then the feature vector is input into the optimized inclusion-V3 convolutional neural network model for classification;
the classification accuracy calculation module is used for calculating the classification accuracy of the optimized inclusion-V3 convolutional neural network model according to the model classification result, judging whether the classification accuracy reaches a preset threshold value, and stopping training when the classification accuracy reaches the preset threshold value to obtain a trained epilepsia electroencephalogram automatic classification model; otherwise, continuing training the optimized inclusion-V3 convolutional neural network model until the classification accuracy reaches a preset threshold;
the output module is used for outputting the classification accuracy of the optimized inclusion-V3 convolutional neural network model.
9. The system for acquiring the epilepsia electroencephalogram automatic classification model, according to claim 8, is characterized in that the classification module comprises a data set division unit, a feature extraction unit, a model optimization unit and a model training unit;
the data set dividing unit is used for forming a data set by a time-frequency graph of the electroencephalogram signals in the epileptic seizure period and a time-frequency graph of the normal electroencephalogram signals, dividing the data set into a training set, a verification set and a test set by adopting a random function, and adding labels to the data of the training set, the test set and the verification set, wherein the label of the time-frequency graph of the electroencephalogram signals in the epileptic seizure period is 1, and the label of the time-frequency graph of the normal electroencephalogram signals is 0;
the feature extraction unit is used for calling an initial inclusion-V3 convolutional neural network model to perform transfer learning feature processing on the time-frequency graph in the data set and outputting a feature vector;
the model optimization unit is used for optimizing the initial inclusion-V3 convolutional neural network model, replacing the full connection layer with two classes to obtain the optimized inclusion-V3 convolutional neural network model;
the model training unit is used for training, verifying and testing the optimized inclusion-V3 convolutional neural network model by utilizing the training set, the verifying set and the testing set, and the model training unit trains the optimized inclusion-V3 convolutional neural network model by using a cross entropy loss function and a random gradient descent algorithm in the training process to obtain the classification result of the optimized inclusion-V3 convolutional neural network model.
10. The system for acquiring the epilepsia electroencephalogram automatic classification model, according to claim 8, wherein the classification accuracy calculation module is used for calculating the classification accuracy of the optimized inclusion-V3 convolutional neural network model through the accuracy, the recall rate and the precision rate;
the accuracy calculation formula is as follows:
the ACC represents the classification accuracy of the optimized inclusion-V3 convolutional neural network model, and the TP is a real example and represents the number of correctly classified electroencephalogram signals in the epileptic seizure period; TN is a true negative example and represents the number of correctly classified samples of the normal electroencephalogram signals; FP is a false positive case and represents the number of samples of the electroencephalogram signals in the epileptic seizure period which are classified in error; FN is false negative example, and represents the number of samples of the normal electroencephalogram signals which are classified in error;
the recall ratio calculation formula is as follows:
wherein SEN represents the classification recall rate of the optimized inclusion-V3 convolutional neural network model;
the accuracy calculation formula is as follows:
wherein PRE represents the classification accuracy of the optimized inclusion-V3 convolutional neural network model.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010625498.6A CN111700592A (en) | 2020-07-02 | 2020-07-02 | Method and system for acquiring epilepsia electroencephalogram automatic classification model and classification system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010625498.6A CN111700592A (en) | 2020-07-02 | 2020-07-02 | Method and system for acquiring epilepsia electroencephalogram automatic classification model and classification system |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN111700592A true CN111700592A (en) | 2020-09-25 |
Family
ID=72545544
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010625498.6A Pending CN111700592A (en) | 2020-07-02 | 2020-07-02 | Method and system for acquiring epilepsia electroencephalogram automatic classification model and classification system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111700592A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113951898A (en) * | 2021-10-15 | 2022-01-21 | 浙江大学 | P300 electroencephalogram signal detection method and device for data migration, electronic device and medium |
| CN114010205A (en) * | 2021-10-11 | 2022-02-08 | 杭州电子科技大学 | Auxiliary analysis method for 3D (three-dimensional) attention residual error deep network children epilepsy syndrome |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106909784A (en) * | 2017-02-24 | 2017-06-30 | 天津大学 | Epileptic electroencephalogram (eeg) recognition methods based on two-dimentional time-frequency image depth convolutional neural networks |
| CN110236533A (en) * | 2019-05-10 | 2019-09-17 | 杭州电子科技大学 | Epileptic seizure prediction method based on the study of more deep neural network migration features |
| CN110960191A (en) * | 2019-11-29 | 2020-04-07 | 杭州电子科技大学 | Epilepsia electroencephalogram signal classification method based on frequency spectrum energy diagram |
-
2020
- 2020-07-02 CN CN202010625498.6A patent/CN111700592A/en active Pending
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106909784A (en) * | 2017-02-24 | 2017-06-30 | 天津大学 | Epileptic electroencephalogram (eeg) recognition methods based on two-dimentional time-frequency image depth convolutional neural networks |
| CN110236533A (en) * | 2019-05-10 | 2019-09-17 | 杭州电子科技大学 | Epileptic seizure prediction method based on the study of more deep neural network migration features |
| CN110960191A (en) * | 2019-11-29 | 2020-04-07 | 杭州电子科技大学 | Epilepsia electroencephalogram signal classification method based on frequency spectrum energy diagram |
Non-Patent Citations (1)
| Title |
|---|
| 曲桂果等: "基于深度网络迁移学习的致痫区脑电识别", 《仪器仪表学报》 * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114010205A (en) * | 2021-10-11 | 2022-02-08 | 杭州电子科技大学 | Auxiliary analysis method for 3D (three-dimensional) attention residual error deep network children epilepsy syndrome |
| CN113951898A (en) * | 2021-10-15 | 2022-01-21 | 浙江大学 | P300 electroencephalogram signal detection method and device for data migration, electronic device and medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111657935B (en) | Epilepsia electroencephalogram recognition system based on hierarchical graph convolutional neural network, terminal and storage medium | |
| WO2021226778A1 (en) | Epileptic electroencephalogram recognition system based on hierarchical graph convolutional neural network, terminal, and storage medium | |
| Brihadiswaran et al. | EEG-based processing and classification methodologies for autism spectrum disorder: A review | |
| CN109009102B (en) | Electroencephalogram deep learning-based auxiliary diagnosis method and system | |
| CN108742697B (en) | Heart sound signal classification method and terminal equipment | |
| Supakar et al. | A deep learning based model using RNN-LSTM for the Detection of Schizophrenia from EEG data | |
| CN112641451B (en) | Multi-scale residual error network sleep staging method and system based on single-channel electroencephalogram signal | |
| CN111700592A (en) | Method and system for acquiring epilepsia electroencephalogram automatic classification model and classification system | |
| WO2024040797A1 (en) | Electroencephalogram-based autism evaluation apparatus and method, terminal device, and medium | |
| WO2021120007A1 (en) | Infrared image sequence-based sleep quality evaluation system and method | |
| Mahato et al. | Analysis of region of interest (RoI) of brain for detection of depression using EEG signal | |
| Hassan et al. | Review of EEG Signals Classification Using Machine Learning and Deep-Learning Techniques | |
| CN112446307B (en) | Local constraint-based non-negative matrix factorization electrocardiogram identity recognition method and system | |
| Deivasigamani et al. | Computer Aided Automatic Detection and Classification of EEG Signals for Screening Epilepsy Disorder. | |
| Rahman et al. | An End-to-End Deep Learning Model for Mental Arithmetic Task Classification from Multi-Channel EEG | |
| Wei et al. | Epileptic seizure prediction from multivariate EEG data using Multidimensional convolution network | |
| CN113014881A (en) | Neurosurgical patient daily monitoring method and system | |
| Divya et al. | Identification of epileptic seizures using autoencoders and convolutional neural network | |
| Rafiammal et al. | A low power and high performance hardware design for automatic epilepsy seizure detection | |
| CN114649071A (en) | Real world data-based peptic ulcer treatment scheme prediction system | |
| Alam et al. | Field programmable gate array‐based energy‐efficient and fast epileptic seizure detection using support vector machine and quadratic discriminant analysis classifier | |
| Singh et al. | Emotion recognition using deep convolutional neural network on temporal representations of physiological signals | |
| CN116616800B (en) | Scalp electroencephalogram high-frequency oscillation signal identification method and device based on meta-shift learning | |
| Luckett | Nonlinear methods for detection and prediction of epileptic seizures | |
| US20230315203A1 (en) | Brain-Computer Interface Decoding Method and Apparatus Based on Point-Position Equivalent Augmentation |
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 | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20200925 |
|
| RJ01 | Rejection of invention patent application after publication |