CN116421144A - ESES (ESES) -related epileptic encephalopathy prognosis evaluation model, training and using method thereof - Google Patents
ESES (ESES) -related epileptic encephalopathy prognosis evaluation model, training and using method thereof Download PDFInfo
- Publication number
- CN116421144A CN116421144A CN202310119357.0A CN202310119357A CN116421144A CN 116421144 A CN116421144 A CN 116421144A CN 202310119357 A CN202310119357 A CN 202310119357A CN 116421144 A CN116421144 A CN 116421144A
- Authority
- CN
- China
- Prior art keywords
- brain
- graph
- eses
- network
- module
- 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
- 238000004393 prognosis Methods 0.000 title claims abstract description 32
- 208000014644 Brain disease Diseases 0.000 title claims abstract description 29
- 230000001037 epileptic effect Effects 0.000 title claims abstract description 28
- 238000012549 training Methods 0.000 title claims abstract description 21
- 238000013210 evaluation model Methods 0.000 title claims abstract description 20
- 238000000034 method Methods 0.000 title claims abstract description 20
- 208000032274 Encephalopathy Diseases 0.000 title claims abstract description 19
- 210000004556 brain Anatomy 0.000 claims abstract description 81
- 238000000605 extraction Methods 0.000 claims abstract description 37
- 238000003745 diagnosis Methods 0.000 claims abstract description 21
- 238000013528 artificial neural network Methods 0.000 claims abstract description 20
- 230000007246 mechanism Effects 0.000 claims abstract description 11
- 238000011156 evaluation Methods 0.000 claims abstract description 10
- 238000011002 quantification Methods 0.000 claims abstract description 9
- 238000004364 calculation method Methods 0.000 claims abstract description 7
- 238000006243 chemical reaction Methods 0.000 claims abstract description 7
- 239000012634 fragment Substances 0.000 claims abstract description 5
- 238000011176 pooling Methods 0.000 claims description 10
- 238000013527 convolutional neural network Methods 0.000 claims description 8
- 238000012952 Resampling Methods 0.000 claims description 7
- 230000004913 activation Effects 0.000 claims description 7
- 238000001914 filtration Methods 0.000 claims description 7
- 230000004927 fusion Effects 0.000 claims description 7
- 230000000877 morphologic effect Effects 0.000 claims description 6
- 230000007958 sleep Effects 0.000 claims description 6
- 238000002372 labelling Methods 0.000 claims description 5
- 230000004931 aggregating effect Effects 0.000 claims description 4
- 125000004122 cyclic group Chemical group 0.000 claims description 4
- 238000013507 mapping Methods 0.000 claims description 4
- 238000010606 normalization Methods 0.000 claims description 4
- 230000006870 function Effects 0.000 claims description 3
- 230000007274 generation of a signal involved in cell-cell signaling Effects 0.000 claims description 3
- 230000015654 memory Effects 0.000 claims description 3
- 238000007781 pre-processing Methods 0.000 claims description 3
- 238000012545 processing Methods 0.000 claims description 3
- 230000004044 response Effects 0.000 claims description 3
- 238000005070 sampling Methods 0.000 claims description 3
- 230000008034 disappearance Effects 0.000 claims description 2
- PCHJSUWPFVWCPO-UHFFFAOYSA-N gold Chemical compound [Au] PCHJSUWPFVWCPO-UHFFFAOYSA-N 0.000 claims description 2
- 230000007774 longterm Effects 0.000 claims description 2
- 210000004243 sweat Anatomy 0.000 claims description 2
- 230000001131 transforming effect Effects 0.000 claims description 2
- 238000012795 verification Methods 0.000 claims description 2
- 238000000537 electroencephalography Methods 0.000 description 6
- 206010010904 Convulsion Diseases 0.000 description 3
- 238000004458 analytical method Methods 0.000 description 3
- 230000007547 defect Effects 0.000 description 3
- 238000001514 detection method Methods 0.000 description 3
- 206010015037 epilepsy Diseases 0.000 description 3
- 208000028329 epileptic seizure Diseases 0.000 description 3
- 230000000306 recurrent effect Effects 0.000 description 3
- 238000011160 research Methods 0.000 description 3
- 208000024891 symptom Diseases 0.000 description 3
- 238000013136 deep learning model Methods 0.000 description 2
- 239000000284 extract Substances 0.000 description 2
- 230000008569 process Effects 0.000 description 2
- 230000037322 slow-wave sleep Effects 0.000 description 2
- 208000002877 Epileptic Syndromes Diseases 0.000 description 1
- 108010076504 Protein Sorting Signals Proteins 0.000 description 1
- 230000002159 abnormal effect Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000003759 clinical diagnosis Methods 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 230000001934 delay Effects 0.000 description 1
- 238000013399 early diagnosis Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 230000001787 epileptiform Effects 0.000 description 1
- 230000007787 long-term memory Effects 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 230000004660 morphological change Effects 0.000 description 1
- 230000003557 neuropsychological effect Effects 0.000 description 1
- 230000002085 persistent effect Effects 0.000 description 1
- 230000006403 short-term memory Effects 0.000 description 1
- 208000005809 status epilepticus Diseases 0.000 description 1
- 230000002459 sustained effect Effects 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
- 230000007704 transition Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
- G16H50/00—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics
- G16H50/30—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics for calculating health indices; for individual health risk assessment
-
- 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
- A61B5/369—Electroencephalography [EEG]
- A61B5/372—Analysis of electroencephalograms
-
- 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
- A61B5/7203—Signal processing specially adapted for physiological signals or for diagnostic purposes for noise prevention, reduction or removal
-
- 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
- A61B5/725—Details of waveform analysis using specific filters therefor, e.g. Kalman or adaptive filters
-
- 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
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
- G16H50/00—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics
- G16H50/20—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics for computer-aided diagnosis, e.g. based on medical expert systems
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B2503/00—Evaluating a particular growth phase or type of persons or animals
- A61B2503/06—Children, e.g. for attention deficit diagnosis
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
- Y02A90/00—Technologies having an indirect contribution to adaptation to climate change
- Y02A90/10—Information and communication technologies [ICT] supporting adaptation to climate change, e.g. for weather forecasting or climate simulation
Landscapes
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Engineering & Computer Science (AREA)
- Biomedical Technology (AREA)
- Medical Informatics (AREA)
- Public Health (AREA)
- General Health & Medical Sciences (AREA)
- Physics & Mathematics (AREA)
- Pathology (AREA)
- Molecular Biology (AREA)
- Biophysics (AREA)
- Heart & Thoracic Surgery (AREA)
- Veterinary Medicine (AREA)
- Animal Behavior & Ethology (AREA)
- Surgery (AREA)
- Psychiatry (AREA)
- Physiology (AREA)
- Artificial Intelligence (AREA)
- Signal Processing (AREA)
- Neurology (AREA)
- Data Mining & Analysis (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Psychology (AREA)
- Databases & Information Systems (AREA)
- Primary Health Care (AREA)
- Neurosurgery (AREA)
- Theoretical Computer Science (AREA)
- Epidemiology (AREA)
- General Engineering & Computer Science (AREA)
- Computing Systems (AREA)
- Evolutionary Computation (AREA)
- Computational Linguistics (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Measurement And Recording Of Electrical Phenomena And Electrical Characteristics Of The Living Body (AREA)
Abstract
The invention belongs to the technical field of epileptic encephalopathy auxiliary diagnosis, and particularly relates to an ESES-related epileptic encephalopathy automatic quantification and prognosis evaluation model and a training and using method thereof. The brain automatic quantification and prognosis evaluation model comprises a brain network feature extraction module, an electroencephalogram depth feature extraction module and a recognition diagnosis module; the brain network feature extraction module comprises a calculation module, a graph convolution neural network and an attention mechanism network, so as to obtain brain network graph theory features and graph features, and the brain network graph theory features and the graph features are fused to obtain information of state conversion between data fragments; the electroencephalogram depth feature extraction module is used for extracting electroencephalogram depth features; the identification and diagnosis module is used for fusing the brain network characteristics extracted by the brain depth characteristic extraction module and the brain depth characteristics extracted by the brain depth characteristic extraction module, and carrying out the identification of ESES characteristic waves and the evaluation of prognosis grades. The invention not only realizes automatic quantification of ESES, but also realizes prognosis evaluation of ESES related epileptic encephalopathy, thereby greatly improving diagnosis and treatment efficiency of ESES.
Description
Technical Field
The invention belongs to the technical field of epileptic encephalopathy auxiliary diagnosis, and particularly relates to a prognosis evaluation model of ESES (ESES) related epileptic encephalopathy, and a training and using method thereof.
Background
The status of epilepticus (Electrical Status Epilepticus during Sleep, ESES) in sleep is an epileptic symptom of sustained spike-slow wave occurring during slow wave sleep, characterized by a strongly activated status epileptiform activity occurring during sleep, which in electroencephalogram consists of a series of consecutive spikes and slow wave anomalies. The occurrence frequency of the pediatric epileptic seizure is high, but the pediatric epileptic seizure generally shows idiopathic, cryptogenic and symptomatic properties, and abnormal discharge is mostly carried out in a sleep state, so that the pediatric epileptic seizure is easily ignored, is not easily found in clinic, and delays treatment and handling. Although most of the children patients tend to improve clinical symptoms and electroencephalogram before and after puberty, since neuropsychological damage caused by ESES is not recoverable, the method is important for early diagnosis and prognosis evaluation of ESES-related epileptic encephalopathy.
Clinically, the Slow wave sleep spike Index (SWI) based on the long Cheng Nao electrographic analysis is an important standard for diagnosing and treating ESES related epileptic syndromes, and is also an important Index for prognosis evaluation of infants. The common basis of clinical diagnosis is to quantify ESES-related epileptic activities and estimate SWI index by electroencephalography in combination with manual identification of epileptic waves. However, the electroencephalogram recording time related to ESES is often up to several hours to tens of hours, and the manual analysis is time-consuming and labor-consuming, and requires a long time (usually up to more than 1 hour) from completion of electroencephalogram examination to the conclusion of the result; the manual estimation is not accurate enough, and the analysis results of different doctors are different, so that the factors are unfavorable for timely diagnosis and effective treatment of the children. Most of the current researches usually identify and quantify ESES by realizing a peak detection mode, extract morphological characteristics, time domain or frequency domain characteristics from signals, and then use a template matching method or a machine learning method to realize classification. The limitation is that the feature extraction step relies on a priori knowledge and requires extensive feature engineering and complex parameter tuning. In addition, most methods only implement spike detection, and ignore slow wave detection, presenting difficulties in achieving accurate identification of ESES signatures. In addition to ESES signature recognition, no study is currently underway for ESES-related epileptic encephalopathy prognosis evaluation.
In summary, the interpretation of ESES epileptic waves has the defects of high professional requirements, large manual interpretation variability and the like, and the existing research proposes a method for ignoring overlapped slow waves and has the defects of inaccurate identification and the like. Therefore, the diagnosis and treatment efficiency of ESES can be greatly improved by researching the ESES automatic quantification and ESES related epileptic encephalopathy prognosis evaluation system.
Disclosure of Invention
Aiming at the defects of the prior art, the invention aims to provide an ESES-related epileptic encephalopathy prognosis evaluation model integrating brain depth characteristics and brain network characteristics, and a training and using method thereof.
The invention provides an ESES related epileptic encephalopathy prognosis evaluation model which fuses brain depth characteristics and brain network characteristics, comprising the following steps: the brain network feature extraction module, the brain electrical depth feature extraction module and the identification diagnosis module; wherein:
the brain network feature extraction module is used for extracting brain network features, including brain network graph theory features calculated by the calculation module and graph features extracted by the graph convolution neural network, and the brain network feature extraction module and the graph feature extraction module are fused to extract state conversion information between data fragments through the attention mechanism network; wherein:
the calculation module calculates and obtains brain network graph theory features according to the originally acquired brain electrical signals, wherein the brain network graph theory features comprise node degree, node betweenness centrality, node strength, clustering coefficients, weighted feature path length, weighted global efficiency, local efficiency and the like;
the graph convolution neural network is used for extracting graph characteristics of the brain network, namely extracting connectivity among graph signal nodes; and extracting original brain electrical signal characteristics of each node of the graph signal as graph signal node information for graph convolution operation.
Further, the graph convolution neural network specifically comprises a graph signal generation module, wherein the graph signal generation module comprises a node characteristic extraction layer and a topological structure extraction layer and generates a graph signal; the graph convolution neural network comprises 1 graph convolution layer and 1 pooling layer; the graph convolution layer is used for aggregating information among layers, the pooling layer is used for aggregating information of all nodes in the layers, and finally the graph characteristics are output through a reading layer.
The attention mechanism network is used for extracting effective information of conversion among the characteristics of each data segment, such as effective information of brain network characteristic conversion between appearance and disappearance of ESES sample waves.
Further, the attention mechanism network specifically comprises a feature fusion module and an attention mechanism module; the feature fusion module is used for fusing the brain network graph theory features and graph features obtained by graph convolution; the attention mechanism module is composed of an encoder and a decoder and is used for mapping the sequence of the data segment characteristics into a diagnosis-assisting type space, and finally, classification is completed through a full connection layer and an activation function.
The electroencephalogram depth feature extraction module is used for extracting electroencephalogram depth features and consists of a double-scale convolutional neural network and a cyclic neural network; wherein:
the double-scale convolutional neural network is used for extracting morphological characteristics of brain electrical signals;
further, the double-scale convolutional neural network specifically comprises two groups of parallel feature extraction modules, wherein each group consists of 3 convolutional layers, 2 pooling layers, a batch normalization layer and a nonlinear activation layer. The convolution layer is used for reducing dimension to extract features, the pooling layer is used for further reducing dimension and reducing calculation amount, the batch normalization layer is used for preventing gradient from disappearing, and the nonlinear activation layer is used for mapping and transforming the features to high dimension and extracting depth information. The first convolution layer of the double-scale convolution neural network adopts different step sizes and convolution kernels, and morphological features with different frequencies are extracted from the electroencephalogram signals and fused, so that the robustness of the model is improved.
The circulating neural network is used for extracting the time sequence characteristics of the brain electrical signals;
further, the recurrent neural network specifically comprises a long and short-term memory network, and compared with the common recurrent neural network, the recurrent neural network is added with a forgetting gate, an input gate, an output gate and a memory unit. Through the control of three gates, the unit state in time sequence learning is changed, information is transmitted to the next moment, and the long-term dependence problem, namely the problem that the gradient disappears after multi-stage propagation is solved while the time sequence signal characteristics are memorized and learned.
The recognition and diagnosis module is used for fusing the brain network characteristics extracted by the brain depth characteristic extraction module and the brain depth characteristics extracted by the brain depth characteristic extraction module, and carrying out recognition, automatic quantification and prognosis evaluation of the ESES characteristic waves.
Further, the identification and diagnosis module is composed of a convolutional neural network and specifically comprises 3 convolutional layers, 2 pooling layers and 2 nonlinear activation layers.
Training the ESES-related epileptic encephalopathy prognosis evaluation model which is constructed by the invention and integrates the electroencephalogram depth feature and the brain network feature, and specifically comprises the following steps:
(1) Acquiring original brain electrical signals of N children in a sleep state, and carrying out ESES characteristic labeling on the brain electrical signals of each tested person; ESES characteristic labeling results carried out on the original EEG signals are used as gold standards for model performance verification; generally, N is required to be 20 or more;
(2) Cutting data of the original EEG signals obtained in the step 1 and labels carried by the original EEG signals, and performing data equalization processing to meet the model data specification and requirements, so that various training data samples are relatively balanced; specifically, the electroencephalogram signal is cut into signals of 30s, whether each section of signals contains ESES characteristic waves or not is used as a classification basis, and the number of most types of samples is equal to that of few types of samples in a downsampling mode;
(3) Resampling and filtering the standardized brain electrical signals obtained in the step 2 to remove background noise, various artifact signals and the like, wherein the specific steps are as follows:
(3.1) resampling the marked electroencephalogram signals to a sampling rate required by the model;
(3.2) filtering background noise by using a Butterworth band-pass filter;
(3.3) removing artifacts caused by sweat, movement, electrode interference, etc. of the subject using a finite length unit impulse response (FIR) filter;
(4) Training a brain disease prognosis evaluation model by using the preprocessed training data; the input of the model is the preprocessed brain wave signals, and the output is whether ESES characteristic waves are contained or not, the starting and stopping positions of the ESES characteristic waves and the prognosis level of the related epileptic brain diseases.
Firstly, an electroencephalogram signal is converted into a graph signal form, the electroencephalogram signal is input to an electroencephalogram depth feature extraction module, the electroencephalogram signal in the graph signal form is input to a brain network feature extraction module, features obtained by the two modules are combined together and input to a recognition and diagnosis module, the output of the module is divided into two parts, wherein a classification problem is whether ESES feature waves are contained or not, and a regression problem is the start and stop positions of the feature waves. And overlapping two loss functions of cross entropy loss and mean square error loss for training. After multiple iterations, a trained encephalopathy prognosis evaluation model is obtained.
The application method of the trained encephalopathy prognosis evaluation model comprises the following specific steps:
(1) Preprocessing the newly acquired original brain electrical signals as training data to obtain a graph signal form required by a model, inputting the graph signal form into a brain network feature extraction module, and inputting the brain electrical signals into a brain electrical depth feature extraction module;
(2) Fusing the obtained brain network characteristics and brain electrical depth characteristics;
(3) And inputting the obtained fusion characteristics into a recognition diagnosis module to obtain a recognition result, and carrying out automatic quantification and prognosis evaluation.
In the present invention, the labeling includes: the start position of ESES characteristic wave and the end position of characteristic wave.
In the present invention, the filters are butterworth band-pass filters and finite length impulse response (FIR) filters.
In the invention, the resampling is to up-sample or down-sample the original signal sequence, thereby reaching a specific sampling frequency.
The invention has the characteristics and beneficial effects that:
by extracting the topological structure and the characteristics of the original electroencephalogram signals, the invention can greatly utilize various information contained in the electroencephalogram signals, greatly reduce the data quantity and the calculated quantity in the subsequent processing process and simultaneously maintain higher robustness.
By extracting the morphological characteristics and the time sequence characteristics of the original electroencephalogram signals, the invention can acquire the time-space connection of the front and back morphological changes of the electroencephalogram signals and can reduce the data dimension and the calculated amount.
The invention extracts EEG signals from brain network, not only utilizes the overall characteristics of the original EEG signal change, but also focuses on the local characteristics of each node, and combines the interconnectivity of each node and the front-back transition information among different data fragments. Meanwhile, the brain network characteristics and the brain depth characteristics are fused, so that the signal characteristics are higher in universality, the robustness of an auxiliary diagnosis model is ensured, and the interpretation and recognition research progress of brain electrical signals are facilitated.
Drawings
FIG. 1 is a schematic flow chart of an ESES-related epileptic encephalopathy automatic quantification and prognosis evaluation model construction method based on fusion brain network characteristics and brain electrical depth characteristics.
Detailed Description
The invention is described in further detail below with reference to the drawings and specific examples. While exemplary embodiments of the invention are shown in the drawings, it should be understood that the present disclosure may be embodied in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the invention to those skilled in the art.
The invention provides an auxiliary diagnosis method for epileptic encephalopathy related to epileptic electric persistent state in children sleeping, the whole flow is shown in figure 1, and the method comprises the following steps:
1. the method comprises the steps of obtaining original electroencephalogram signals of at least 20 children in the sleeping process, enabling the signal duration of each child to be at least about 2 hours (at least one sleeping period is needed), and enabling an expert to sign symptoms of each tested electroencephalogram signal to obtain training data with signs. In this embodiment, the original electroencephalogram signal acquired by the medical electroencephalogram acquisition device does not include a feature wave position tag, so that in order to enable the proposed model to learn features, an expert is required to manually mark the start and end positions of the corresponding feature wave in the signal, and the marked data are used for training the proposed deep learning model.
2. Preprocessing the obtained marked data, mainly comprising the following steps: data slicing, filtering, resampling, data equalization, and the like.
(1) Cutting data: cutting the original electroencephalogram signals to generate standardized data fragments;
(2) And (3) filtering: removing larger background noise in the original electroencephalogram signal by using a Butterworth band-pass filter to obtain a purer electroencephalogram signal; removing the artifacts by using an FIR filter;
(3) Resampling: the brain electrical data was resampled to 100Hz.
3. Training the auxiliary diagnostic model using the preprocessed training data.
4. The trained model performance is tested using test set samples. Model performance is defined as the accuracy of the comparison manual annotation.
Compared with the existing ESES auxiliary diagnosis method, the method provided by the invention can improve the utilization rate of information contained in the electroencephalogram signal and greatly reduce the data quantity and the operand on the premise of effectively realizing auxiliary diagnosis, and meanwhile, the robustness of model use among different crowds is maintained. The invention adopts a deep learning model method, which can greatly reduce the time and workload of manual identification.
Claims (6)
1. An ESES-related epileptic brain disease prognosis evaluation model integrating brain depth features and brain network features, comprising: the brain network feature extraction module, the brain electrical depth feature extraction module and the identification diagnosis module; wherein:
the brain network feature extraction module comprises a calculation module, a graph convolution neural network and an attention mechanism network, wherein:
the calculation module calculates and obtains brain network graph theory features according to the originally acquired brain electrical signals, wherein the brain network graph theory features comprise node degree, node betweenness centrality, node strength, clustering coefficients, weighted feature path length, weighted global efficiency and local efficiency;
the graph convolution neural network is used for extracting graph characteristics of the brain network, namely extracting connectivity among graph signal nodes; extracting original electroencephalogram signal characteristics of each node of the graph signal at the same time, and using the characteristics as graph signal node information for graph convolution operation;
the attention mechanism network is used for fusing brain network graph theory characteristics and map characteristics, extracting information of state conversion among data fragments, and comprises effective information of brain network characteristic conversion between appearance and disappearance conversion of ESES (ESES) sample waves;
the electroencephalogram depth feature extraction module is used for extracting electroencephalogram depth features and consists of a double-scale convolutional neural network and a cyclic neural network; wherein:
the double-scale convolutional neural network is used for extracting morphological characteristics of brain electrical signals;
the circulating neural network is used for extracting the time sequence characteristics of the brain electrical signals;
the recognition and diagnosis module is used for fusing the brain network characteristics extracted by the brain depth characteristic extraction module and the brain depth characteristics extracted by the brain depth characteristic extraction module, and carrying out automatic quantification of ESES characteristic waves and prognosis evaluation of related epileptic encephalopathy.
2. The ese related epileptic brain disease prognosis evaluation model according to claim 1, wherein in the brain network feature extraction module:
the graph convolution neural network specifically comprises a graph signal generation module which comprises a node characteristic extraction layer and a topological structure extraction layer and generates a graph signal; the graph convolution neural network comprises 1 graph convolution layer and 1 pooling layer; the graph convolution layer is used for aggregating information among layers, the pooling layer is used for aggregating information of all nodes in the layers, and finally, the graph characteristics are output through a reading layer;
the attention mechanism network specifically comprises a feature fusion module and an attention mechanism module; the feature fusion module is used for fusing the brain network graph theory features and graph features obtained by graph convolution; the attention mechanism module is composed of an encoder and a decoder and is used for mapping and outputting the sequence of the data segment characteristics into the diagnosis-assisting type space.
3. The ese related epileptic brain disease prognosis evaluation model according to claim 2, wherein in the electroencephalogram depth feature extraction module:
the double-scale convolutional neural network specifically comprises two groups of parallel feature extraction modules, wherein each group consists of 3 convolutional layers, 2 pooling layers, a batch normalization layer and a nonlinear activation layer; the convolution layer is used for reducing dimension to extract features, the pooling layer is used for further reducing dimension and reducing calculation amount, the batch normalization layer is used for preventing gradient from disappearing, and the nonlinear activation layer is used for mapping and transforming the features to high dimension and extracting depth information; the first convolution layer of the double-scale convolution neural network adopts different step sizes and convolution kernels, and morphological features with different frequencies are extracted from the electroencephalogram signals and fused, so that the robustness of a model is improved;
the cyclic neural network specifically comprises a long and short time memory network, wherein a forgetting gate, an input gate, an output gate and a memory unit are added compared with the cyclic neural network; through the control of three gates, the unit state in time sequence learning is changed, information is transmitted to the next moment, and the long-term dependence problem, namely the problem that the gradient disappears after multi-stage propagation is solved while the time sequence signal characteristics are memorized and learned.
4. The ese related epileptic brain disease prognosis evaluation model according to claim 3, wherein the identification and diagnosis module is specifically composed of a convolutional neural network, and specifically comprises 3 convolutional layers, 2 pooling layers and 2 nonlinear activation layers.
5. A method of training a prognosis evaluation model for ese-related epileptic brain diseases according to any one of claims 1 to 4, comprising the specific steps of:
(1) Acquiring original brain electrical signals of N children in a sleep state, and carrying out ESES characteristic labeling on the brain electrical signals of each tested person; ESES characteristic labeling results carried out on the original EEG signals are used as gold standards for model performance verification;
(2) Cutting data of the original EEG signals obtained in the step 1 and labels carried by the original EEG signals, and performing data equalization processing to meet the model data specification and requirements, so that various training data samples are relatively balanced; specifically, the electroencephalogram signal is cut into signals of 30s, whether each section of signals contains ESES characteristic waves or not is used as a classification basis, and the number of most types of samples is equal to that of few types of samples in a downsampling mode;
(3) Resampling and filtering the standardized brain electrical signal obtained in the step 2, and filtering background noise and various artifact signals, wherein the specific steps are as follows:
(3.1) resampling the marked electroencephalogram signals to a sampling rate required by the model;
(3.2) filtering background noise by using a Butterworth band-pass filter;
(3.3) removing artifacts due to sweat, movement and electrode interference of the subject using a finite length unit impulse response (FIR) filter;
(4) Training a brain disease prognosis evaluation model by using the preprocessed training data; the input of the model is the preprocessed brain wave signals, and the output is whether ESES characteristic waves are contained or not, and the starting and stopping positions of the ESES characteristic waves and the prognosis level of the related epileptic brain diseases;
firstly, converting an electroencephalogram signal into a graph signal form, inputting the electroencephalogram signal into an electroencephalogram depth feature extraction module, inputting the electroencephalogram signal in the graph signal form into a brain network feature extraction module, respectively obtaining features by the two modules, combining the features together, inputting the features into a recognition and diagnosis module, and dividing the output of the module into two parts, wherein a classification problem is whether ESES feature waves are contained or not, and a regression problem is the start and stop positions of the feature waves; training by overlapping two loss functions of cross entropy loss and mean square error loss; after multiple iterations, a trained encephalopathy prognosis evaluation model is obtained.
6. A method of using the brain disease prognosis evaluation model trained by claim 5, comprising the specific steps of:
(1) Preprocessing the acquired original brain electrical signals as training data to obtain a graph signal form required by a model, inputting the graph signal form into a brain network feature extraction module, and inputting the brain electrical signals into a brain electrical depth feature extraction module;
(2) Fusing the obtained brain network characteristics and brain electrical depth characteristics;
(3) And inputting the obtained fusion characteristics into a recognition diagnosis module to obtain a recognition result, and carrying out automatic quantification and prognosis evaluation.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202310119357.0A CN116421144A (en) | 2023-02-15 | 2023-02-15 | ESES (ESES) -related epileptic encephalopathy prognosis evaluation model, training and using method thereof |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202310119357.0A CN116421144A (en) | 2023-02-15 | 2023-02-15 | ESES (ESES) -related epileptic encephalopathy prognosis evaluation model, training and using method thereof |
Publications (1)
Publication Number | Publication Date |
---|---|
CN116421144A true CN116421144A (en) | 2023-07-14 |
Family
ID=87089660
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202310119357.0A Pending CN116421144A (en) | 2023-02-15 | 2023-02-15 | ESES (ESES) -related epileptic encephalopathy prognosis evaluation model, training and using method thereof |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN116421144A (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN116712089A (en) * | 2023-07-26 | 2023-09-08 | 华南师范大学 | Epileptiform discharge enriching epileptiform interval and method for predicting focus |
-
2023
- 2023-02-15 CN CN202310119357.0A patent/CN116421144A/en active Pending
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN116712089A (en) * | 2023-07-26 | 2023-09-08 | 华南师范大学 | Epileptiform discharge enriching epileptiform interval and method for predicting focus |
CN116712089B (en) * | 2023-07-26 | 2024-03-22 | 华南师范大学 | Epileptiform discharge enriching epileptiform interval and method for predicting focus |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Kulkarni et al. | Extracting salient features for EEG-based diagnosis of Alzheimer's disease using support vector machine classifier | |
US20240023886A1 (en) | Noninvasive method and system for sleep apnea detection | |
CN105956623A (en) | Epilepsy electroencephalogram signal classification method based on fuzzy entropy | |
CN110353673B (en) | Electroencephalogram channel selection method based on standard mutual information | |
CN108511055B (en) | Ventricular premature beat recognition system and method based on classifier fusion and diagnosis rules | |
CN100493445C (en) | Automatic testing method for traditional Chinese medical pulse manifestation characteristics parameter | |
CN109009102B (en) | Electroencephalogram deep learning-based auxiliary diagnosis method and system | |
CN104569694B (en) | Electric signal feature extraction and recognition system oriented to aircraft flying process | |
Emmanuel | A review of signal processing techniques for heart sound analysis in clinical diagnosis | |
CN110141216B (en) | Identification method, training method and system for QRS (QRS) characteristic waves of electrocardiosignals | |
CN106618560B (en) | Method and device for processing brain wave signals | |
CN110432895B (en) | Training data processing method, electrocardiographic waveform detection method and electronic equipment | |
CN113925459B (en) | Sleep stage method based on electroencephalogram feature fusion | |
CN116421144A (en) | ESES (ESES) -related epileptic encephalopathy prognosis evaluation model, training and using method thereof | |
CN112869716B (en) | Pulse feature identification system and method based on two-channel convolutional neural network | |
Huang et al. | Automatic epileptic seizure detection via attention-based CNN-BiRNN | |
CN113768519A (en) | Method for analyzing consciousness level of patient based on deep learning and resting state electroencephalogram data | |
CN103405225A (en) | Method, apparatus and device for obtaining pain feeling evaluation indexes | |
CN104095627A (en) | Electrocardiogram digitized signal quality soft-decision method and device | |
CN103267652B (en) | Intelligent online diagnosis method for early failures of equipment | |
CN111528821A (en) | Method for identifying characteristic points of counterpulsation waves in pulse waves | |
CN116269212A (en) | Multi-mode sleep stage prediction method based on deep learning | |
CN103425983A (en) | Brain network topology difference fast extracting method based on network synchronicity | |
CN116881762A (en) | Emotion recognition method based on dynamic brain network characteristics | |
CN110811591A (en) | Heart failure grading method based on heart rate variability |
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 |