An epilepsy electroencephalogram signal recognition method, system, device and storage medium

Abstract

The application discloses a kind of epilepsy electroencephalogram signal identification method, system, equipment and storage medium, comprising: obtaining electroencephalogram signal, and the phase synchronization information between different channels is calculated as edge feature according to electroencephalogram signal;Filtering electroencephalogram signal, respectively calculate and extract the time domain feature, dynamics feature and phase amplitude coupling feature corresponding to each channel different wave band, and fusion is node feature;The edge feature and node feature obtained are integrated into a graph, classified using a graph attention network model, and the electroencephalogram signal is judged and identified according to the classification result.The application proposes a high-performance, prior-explainable classification method for epilepsy signal identification and prediction, which not only considers the interaction between different channels, but also focuses on and extracts the deep features of the time domain, dynamics, cross-frequency coupling and other features of the node itself, and is more accurate for classifying electroencephalogram signals in different periods.

Description

TECHNICAL FIELD

[0001] The application belongs to the field of cognitive neuroscience and information technology, and relates to an epilepsy electroencephalogram signal recognition method, system, device and storage medium. BACKGROUND

[0002] The diagnosis of epilepsy has always been a hot issue of concern. As a complex nervous system disease, the accurate and timely diagnosis of epilepsy is of great significance to the quality of life of patients, the development of treatment plans and prognosis evaluation. Traditional epilepsy diagnosis highly depends on electroencephalogram (EEG), among which, electroencephalogram data is an important basis for judging epilepsy. The diagnosis of epilepsy based on electroencephalogram generally requires experts to make judgments based on professional knowledge, which requires a lot of manpower and material resources as a cost, and is affected by many factors such as doctor's experience, fatigue degree and subjective judgment, which may lead to inconsistency and delay of the diagnosis results, and it is difficult to predict in advance.

[0003] In recent years, with the rapid development of computer science and artificial intelligence technology, the introduction of machine learning, especially deep learning, has brought the possibility of automatic diagnosis and prediction of epilepsy. Deep learning models can automatically learn complex feature representations from massive EEG data and make predictions accordingly, which is expected to significantly improve the accuracy and efficiency of diagnosis and reduce the burden of medical resources. However, despite the significant progress in this field, there are still many challenges and limitations of existing technologies.

[0004] Firstly, EEG data itself has a high degree of complexity and nonlinearity, and the functional connection between different brain regions is the key to understanding the mechanism of epilepsy seizure. However, existing researches treat different channels of electroencephalogram signals as independent of each other, or only consider the spatial position relationship between different channels of electroencephalogram signals, without paying attention to the functional connection between different brain regions that spans multiple brain regions. This simplified processing method is difficult to fully capture the dynamic evolution process of epilepsy activity in the brain, limiting the accuracy and reliability of the diagnosis model. Secondly, each EEG channel contains rich physiological and pathological information, which is crucial for identifying the unique electro-physiological patterns of epilepsy. However, most of the researches on epilepsy electroencephalogram diagnosis and prediction based on graph neural networks (GNN) focus more on the connection relationship between different channels, and pay little attention to the information of the channels themselves. This way, it is easy to overlook the spatial and functional connections between different channels. This may cause the model to lose key information in the feature extraction stage, and thus affect the subsequent diagnosis and prediction performance. Furthermore, the model has poor interpretability. Although deep learning models can achieve remarkable prediction results, their decision-making process is often like a "black box", and clinicians can hardly directly use the results as a reference. SUMMARY

[0005] The purpose of this invention is to provide a method, system, device, and storage medium for recognizing epilepsy EEG signals, thereby addressing the lack of accuracy and reliability in diagnosing based on existing deep learning models.

[0006] To achieve the above objectives, the present invention employs the following technical solution:

[0007] A method for recognizing epilepsy electroencephalogram (EEG) signals, comprising:

[0008] Acquire EEG signals and calculate phase synchronization information between different channels as side features based on the EEG signals;

[0009] The EEG signal was filtered, and the temporal features, dynamic features, and phase amplitude coupling features corresponding to different bands of each channel were calculated and extracted, and then fused into node features.

[0010] The obtained edge features and node features are integrated into a graph, and a graph attention network model is used for classification. The EEG signals are then judged and identified based on the classification results.

[0011] Furthermore, the phase synchronization information between the different channels is as follows:

[0012]

[0013] In the formula, Indicates the length of the time series. , Represents a node The instantaneous phase of the corresponding time series, Represents the imaginary part of the phase difference. It is a symbolic function.

[0014] Furthermore, the time-domain features include the maximum amplitude and energy of the EEG signal;

[0015] The maximum amplitude of the EEG signal is

[0016]

[0017] In the formula, This represents the time series corresponding to the band. These represent four different frequency bands;

[0018] The energy of the EEG signal is

[0019]

[0020] In the formula, Indicates the sampling frequency.

[0021] Furthermore, the method for calculating the dynamic characteristics is as follows:

[0022] reconstructing into a vector sequence of dimension :

[0023]

[0024] wherein, based on the reconstructed , calculate its corresponding :

[0025]

[0026] wherein, denotes a set threshold value, is a Heaviside function, and obtains After that, calculate the of the entire vector sequence:

[0027]

[0028] set the dimension to , repeat the above process to obtain , and calculate the time sequence dimension L sample entropy under the threshold condition:

[0029]

[0030] Calculate the sample entropy of a single channel under different wave bands , and finally obtain the entropy feature matrix of the channel .

[0031] Further, the calculation method of the phase amplitude coupling feature is:

[0032] Filter the original time sequence x into high frequency sequence and low frequency sequence Two frequency bands, then use Hilbert transform to calculate the instantaneous phase of the low frequency sequence and the amplitude envelope surface of the high frequency sequence , divide the range into K partitions, for the , calculate the average value of the corresponding partition, and obtain the normalized value :

[0033]

[0034] obtain the sequence , and calculate the Shannon entropy of the sequence:

[0035]

[0036] The following formula is used to obtain... :

[0037]

[0038] Calculate the MI index of the low-frequency delta band for the θ, α, and β bands respectively. The phase amplitude coupling characteristic matrix of each channel is obtained. .

[0039] Furthermore, the graph attention network model includes a graph attention network and a node feedforward network;

[0040] The working mechanism of the graph attention network is as follows:

[0041] First, calculate the attention index for different nodes.

[0042]

[0043] In the formula, W∈RF′×F is a learnable parameter matrix. For a shared attention mechanism;

[0044]

[0045] In the formula, ∈R 2F′, For transpose, For merging operations;

[0046]

[0047] In the formula, Represents a node The first-order nearest neighbor node;

[0048]

[0049] In the formula, h represents the h-th attention operation. This indicates a merge operation. Indicates the activation function;

[0050] The working mechanism of the node feedforward network is as follows:

[0051]

[0052] In the formula, W1, W2 ∈ RF×F′, RF′×F′′, Let ∈RF′ and RF′′ be the learnable coefficient matrix and bias, respectively, where F represents the dimension of the input data and F′ represents the dimension of the output data. This is represented by the Sigmoid activation function.

[0053] Furthermore, the EEG signals include non-ictal signals, preictal signals, and ictal signals.

[0054] An epilepsy EEG signal recognition system, comprising:

[0055] The edge feature calculation module is used to acquire EEG signals and calculate phase synchronization information between different channels as edge features based on the EEG signals.

[0056] The node feature calculation module is used to filter the EEG signal, calculate and extract the temporal features, dynamic features and phase amplitude coupling features corresponding to different bands of each channel, and fuse them into node features.

[0057] The judgment and recognition module is used to integrate the obtained edge features and node features into a graph, classify them using a graph attention network model, and judge and recognize the EEG signals based on the classification results.

[0058] A terminal device includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the method.

[0059] A computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the method.

[0060] Compared with the prior art, the present invention has the following beneficial effects:

[0061] This invention provides a method for recognizing epilepsy EEG signals. It calculates phase synchronization information between different channels of EEG signal data as side features, and simultaneously filters and divides the EEG signal into multiple frequency bands. For each channel, it calculates and extracts temporal, dynamic, and phase amplitude coupling features in different frequency bands, and merges these into node features. The calculated side and node features are integrated into a graph, and a graph attention network model is used for classification to determine whether the EEG signal represents the preictal, ictal, or epileptic phase, thus aiding in epilepsy prediction. This invention utilizes a GAT-based deep learning model to propose a high-performance, prior-interpretable classification method for epilepsy signal detection and prediction. Compared with existing epilepsy detection research, this invention not only considers the interaction between different channels in the EEG signal but also focuses on and extracts deep-level features of the nodes themselves, such as temporal, dynamic, and cross-frequency coupling, integrating these two types of information into a graph signal. This allows for a more comprehensive mining of deep-level information in the EEG signal, making the model more accurate in classifying EEG signals at different stages. Attached Figure Description

[0062] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as a limitation on the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0063] Figure 1 This is a schematic diagram of the overall structure of the present invention.

[0064] Figure 2 This is a diagram illustrating the calculation process of the phase amplitude coupling index of the present invention.

[0065] Figure 3 This is a schematic diagram of the network structure of the present invention.

[0066] Figure 4 This is a graph showing the loss rate and accuracy of the model in Embodiment 1 of the present invention.

[0067] Figure 5 This is the accuracy confusion matrix diagram of the last round of training model in Embodiment 1 of the present invention.

[0068] Figure 6 This is a schematic diagram of the structure of a preferred embodiment of the epilepsy EEG signal recognition system of the present invention.

[0069] Figure 7 This is a schematic diagram of the electronic device structure according to a preferred embodiment of the present invention. Detailed Implementation

[0070] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations.

[0071] Therefore, the following detailed description of the embodiments of the invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the invention without inventive effort are within the scope of protection of the invention.

[0072] The following description, in conjunction with the accompanying drawings, illustrates exemplary embodiments of this application, including various details to aid understanding. These should be considered merely exemplary. Therefore, those skilled in the art will recognize that various changes and modifications can be made to the embodiments described herein without departing from the scope and spirit of this application. Similarly, for clarity and brevity, descriptions of well-known functions and structures are omitted in the following description.

[0073] Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0074] It should be noted that the terminals involved in the embodiments of this application may include, but are not limited to, mobile phones, personal digital assistants (PDAs), wireless handheld devices, tablet computers, personal computers (PCs), MP3 players, MP4 players, wearable devices (e.g., smart glasses, smartwatches, smart bracelets), smart home devices, and other smart devices.

[0075] Furthermore, the term "and / or" in this article is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, or B existing alone. Additionally, the character " / " in this article generally indicates that the preceding and following related objects have an "or" relationship.

[0076] The present invention will now be described in further detail with reference to the accompanying drawings:

[0077] See Figure 1 This invention provides a method for recognizing epilepsy electroencephalogram (EEG) signals, specifically including the following steps:

[0078] Step 1: First, calculate the phase synchronization information between different channels based on the raw EEG data sample. As an edge feature:

[0079]

[0080] In the formula, Indicates the length of the time series. , Represents a node The instantaneous phase of the corresponding time series, Represents the imaginary part of the phase difference. The sign function is PLI. The range of PLI is between 0 and 1. The larger the value, the stronger the phase synchronization between the two regions, that is, the closer the functional connection between the two regions. Conversely, the smaller the value, the more independent the relationship between the two regions.

[0081] The connectivity relationships between different nodes are calculated based on the PLI index, and a functional connectivity binary matrix corresponding to the original EEG signal is constructed. Each element in the matrix The calculation formula is as follows:

[0082]

[0083] in, A threshold is set; when the PLI between two nodes exceeds the threshold, a connection is considered to exist between the two channels; otherwise, no connection is considered to exist. Matrix ∈RN×N, where N is the number of channels of the EEG signal.

[0084] Step 2: Filter and divide the raw EEG signal samples into... For each of the four bands, the temporal features, dynamic features, and phase amplitude coupling features corresponding to different bands of each channel are calculated and extracted, and then fused into nodal features.

[0085] Temporal characteristics:

[0086] For EEG signals of different frequency bands, calculate their maximum amplitude values ​​respectively. With energy The calculation methods for the time-domain features of the signal are as follows:

[0087]

[0088]

[0089] In the formula, For the time series of the corresponding band, the maximum absolute value is taken as the amplitude information of that band. ; For the sampling frequency, take The integral of the square of the value over time is used as the energy information for that band.

[0090] Dynamic characteristics:

[0091] For a time series Reconstruct it into A sequence of vectors of dimension:

[0092]

[0093] in, Based on the reconstruction Calculate its corresponding :

[0094]

[0095] in, The threshold value set, For the Heaviside function, we get Then, calculate the entire vector sequence. The calculation formula is as follows:

[0096]

[0097] Next, set the dimension to Repeat the above steps to obtain And calculate the time series of length L. dimension, The sample entropy under the threshold condition is:

[0098]

[0099] set up It is 2. Time series Calculate the sample entropy of a single channel in different bands using 0.2 times the standard deviation. Finally, the entropy feature matrix of this channel is obtained. The subscript indicates the different bands of the channel.

[0100] Phase-amplitude coupling characteristics:

[0101] Filter the original time series x into a high-frequency series. With low frequency sequences Two frequency bands were used, and then the instantaneous phase of the low-frequency sequence was calculated using the Hilbert transform. amplitude envelope of high-frequency sequences The instantaneous phase ranges from [-180°, 180°]. The range is divided into K equal partitions. Calculate the corresponding partitions average And obtain the normalized value. The calculation process is as follows: Figure 2 As shown.

[0102]

[0103] This yields the sequence. And calculate the Shannon entropy of the sequence:

[0104]

[0105] Finally, we obtain the result using the following formula. :

[0106]

[0107] In the formula, The instantaneous phase of the low-frequency sequence, The amplitude envelope of the high-frequency sequence. With K=18, the MI index of the low-frequency delta band for the θ, α, and β bands was calculated. The phase amplitude coupling characteristic matrix of each channel is obtained. .

[0108] Based on the functional connectivity and temporal, dynamic, and phase amplitude coupling characteristics obtained from the above calculations, the original EEG signal is reconstructed into graph-structured data. , among which the side The binary PLI function matrix calculated in step 1 ,node ,in .

[0109] Step 3: Integrate the calculated edges and nodes into a graph and input it into the graph attention network model. The graph attention network model mainly includes a graph attention network and a node feedforward network. The specific mechanism of the graph attention network is shown in the following formula:

[0110] First, calculate the attention index for different nodes.

[0111]

[0112] In the formula, W∈RF′×F is a learnable parameter matrix. For a shared attention mechanism.

[0113]

[0114] In the formula, ∈R 2F′, For transpose, For merging operations.

[0115]

[0116] In the formula, Represents a node First-order nearest neighbor

[0117]

[0118] In the formula, h represents the h-th attention operation. This indicates a merge operation. This represents the activation function.

[0119] The working mechanism of the node feedforward network is shown in the following equation:

[0120]

[0121] In the formula, W1, W2 ∈ RF×F′, RF′×F′′, Let ∈RF′ and RF′′ be the learnable coefficient matrix and bias, respectively, where F represents the dimension of the input data and F′ represents the dimension of the output data. This is represented by the Sigmoid activation function.

[0122] Step 4: A schematic diagram of the specific structure of the classification algorithm is shown below. Figure 3 As shown, the data processed through the above steps is input into the graph attention network model to obtain the final classification result, which determines whether the original EEG signal is a signal from the non-ictal period, the preictal period, or the ictal period.

[0123] Example 1:

[0124] This embodiment provides a method for recognizing epilepsy electroencephalogram (EEG) signals, including the following steps:

[0125] 1) Obtaining preprocessed EEG signal data. In this embodiment, the dataset used is the CHB-MIT dataset, an open-source scalp EEG signal dataset recorded at Children's Hospital Boston. These scalp EEG signals were measured by multiple electrodes placed at fixed positions on the subject's scalp to record multi-channel electrophysiological signals of brain activity. The dataset records a total of 198 epileptic recordings from 22 patients aged 1.5 to 22 years, including 17 female and 5 male patients. The CHB21 data was obtained by re-examining the CHB01 patient 1.5 years later, and some information about CHB24 is unknown. The data was measured at 256 Hz with 16-bit resolution, and the electrode positions followed the international 10-20 system. In this embodiment, we selected the CHB01 patient as the research subject, and for the selection of brain regions, we selected 22 channels that were present in most of the recordings.

[0126] 2) Bandpass filtering. In this embodiment, a third-order Butterworth filter is used to perform bandpass filtering of 0.1~40Hz on the original EEG signal to eliminate DC and high-frequency signal interference in the EEG signal.

[0127] 3) In this embodiment, a window size of 5 seconds and a sliding step size of 3 seconds are used to segment the EEG signal data at different time periods, resulting in nearly 10,000 EEG signal samples of different categories with a duration of 5 seconds.

[0128] 4) Calculate the binary matrix of functional connectivity between different channels in the original EEG signal data sample. As edge features The raw EEG sample signals were filtered and divided into... For each of the four bands, time-domain, dynamic, and phase-amplitude coupling information is calculated and extracted, and these are then fused into nodal features. Based on the calculated node and edge information, the original samples are reconstructed into graph structure data samples for subsequent model training. .

[0129] 5) Integrate the calculated edges and nodes into a graph and input it into a graph attention-based classification algorithm. Its structural diagram is shown below. Figure 3 As shown, the final classification results were obtained, and the original EEG signals were identified as signals from the non-ictal period, the preictal period, and the ictal period.

[0130] like Figure 4 As shown, the loss rate and accuracy of the model are represented. "Training loss rate" and "validation loss rate" represent the model's loss rate on the training and validation sets, respectively. These are metrics that measure the difference between the model's predictions and the actual results. A lower loss rate indicates that the model's predictions are closer to the actual results, and the better the model's performance. "Training accuracy" and "validation accuracy" represent the model's accuracy on the training and validation sets, respectively. These are also metrics used to measure the difference between the model's predictions and the actual results. A value closer to 1 indicates a more accurate classification result. Figure 4 The loss rate of the model eventually plateaus and stabilizes at a relatively small value, while the accuracy of the model eventually plateaus and stabilizes at a relatively high value. From this, we can conclude the following:

[0131] 1. The model has converged: During the training process, the value of the loss function gradually decreases and tends to stabilize, indicating that the model has learned the main features of the data and no longer improves significantly.

[0132] 2. Training is complete: During the optimization process, the value of the loss function becomes smaller and smaller, eventually stabilizing at a relatively small value, which means that the model parameters have found a good solution.

[0133] 3. Stable model performance: The stability of the loss rate indicates that the model's performance has reached a stable level, meaning that the error has been minimized.

[0134] 4. Good generalization ability: The stability of loss rate and accuracy mainly reflects the model's performance, because overfitted models usually exhibit unstable loss rate and accuracy during training.

[0135] 5. Reasonable parameter settings: Stable loss rate and accuracy indicate that the parameters (learning rate, batch size, etc.) are set properly, and the model can converge to a good solution within a limited number of rounds.

[0136] like Figure 5 The diagram shows the confusion matrix of the model. In the confusion matrix, "True Value" represents the actual label of the data, and "Predicted Value" represents the label predicted by the model. The numbers in the nine cells represent the number of samples whose predictions match the corresponding labels. The "Accuracy" metric in the matrix represents the model's ability to distinguish between each class of samples; the model's average accuracy is 93.81%. The "Precision" metric in the matrix measures the accuracy of the model's predictions for positive samples. Figure 5 The accuracy and precision of the model are both above 90%, indicating that the model has good predictive performance on the dataset and high reliability. It can capture positive samples well and thus better distinguish EEG signals at different times.

[0137] Example 2:

[0138] Embodiment 2 provided by the present invention is an embodiment of the epilepsy EEG signal recognition system provided by the present invention, such as... Figure 6 As shown, an embodiment of the system includes: an edge feature calculation module, a node feature calculation module, and a judgment and recognition module.

[0139] The edge feature calculation module is used to acquire EEG signals and calculate phase synchronization information between different channels as edge features based on the EEG signals.

[0140] The node feature calculation module is used to filter the EEG signal, calculate and extract the temporal features, dynamic features and phase amplitude coupling features corresponding to different bands of each channel, and fuse them into node features.

[0141] The judgment and recognition module is used to integrate the obtained edge features and node features into a graph, classify them using a graph attention network model, and judge and recognize the EEG signals based on the classification results.

[0142] It is understood that the epilepsy EEG signal recognition system provided by the present invention corresponds to the epilepsy EEG signal recognition method provided in the foregoing embodiments. The relevant technical features of the epilepsy EEG signal recognition system can be referred to the relevant technical features of the epilepsy EEG signal recognition method, and will not be repeated here.

[0143] Another object of the present invention is to provide an electronic device, such as... Figure 7 As shown, it includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor performs the steps of the epilepsy EEG signal recognition method.

[0144] The method for recognizing epilepsy EEG signals includes the following steps:

[0145] Acquire EEG signals and calculate phase synchronization information between different channels as side features based on the EEG signals;

[0146] The EEG signal was filtered, and the temporal features, dynamic features, and phase amplitude coupling features corresponding to different bands of each channel were calculated and extracted, and then fused into node features.

[0147] The obtained edge features and node features are integrated into a graph, and a graph attention network model is used for classification. The EEG signals are then judged and identified based on the classification results.

[0148] A fourth objective of this invention is to provide a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the epilepsy EEG signal recognition method.

[0149] The method for recognizing epilepsy EEG signals includes the following steps:

[0150] Acquire EEG signals and calculate phase synchronization information between different channels as side features based on the EEG signals;

[0151] The EEG signal was filtered, and the temporal features, dynamic features, and phase amplitude coupling features corresponding to different bands of each channel were calculated and extracted, and then fused into node features.

[0152] The obtained edge features and node features are integrated into a graph, and a graph attention network model is used for classification. The EEG signals are then judged and identified based on the classification results.

[0153] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0154] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0155] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0156] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0157] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the specific implementation of the present invention. Any modifications or equivalent substitutions that do not depart from the spirit and scope of the present invention should be covered within the protection scope of the claims of the present invention.

Claims

1. A method for recognizing epileptic electroencephalogram (EEG) signals, characterized in that, include: Acquire EEG signals and calculate phase synchronization information between different channels as side features based on the EEG signals; The EEG signal was filtered, and the temporal features, dynamic features, and phase amplitude coupling features corresponding to different bands of each channel were calculated and extracted, and then fused into node features. The obtained edge features and node features are integrated into a graph, and a graph attention network model is used for classification. The EEG signals are then judged and identified based on the classification results.

2. The method for recognizing epilepsy EEG signals according to claim 1, characterized in that, The phase synchronization information between different channels is In the formula, Indicates the length of the time series. , Represents a node The instantaneous phase of the corresponding time series, Represents the imaginary part of the phase difference. It is a symbolic function.

3. The method for recognizing epilepsy EEG signals according to claim 1, characterized in that, The time-domain features include the maximum amplitude and energy of the EEG signal; The maximum amplitude of the EEG signal is In the formula, This represents the time series corresponding to the band. These represent four different frequency bands; The energy of the EEG signal is In the formula, Indicates the sampling frequency.

4. The method for recognizing epilepsy EEG signals according to claim 1, characterized in that, The method for calculating the dynamic characteristics is as follows: Time series Reconstructed A sequence of vectors of dimension: in, Based on the reconstruction Calculate its corresponding : in, This indicates the set threshold. For the Heaviside function, we get Then, calculate the entire vector sequence. : Set the dimension to Repeat the above process to obtain And calculate the time series of length L. dimension, Sample entropy under threshold conditions: Calculate the sample entropy of a single channel in different bands. Finally, the entropy feature matrix of this channel is obtained. .

5. The method for recognizing epilepsy EEG signals according to claim 1, characterized in that, The method for calculating the phase amplitude coupling characteristic is as follows: Filter the original time series x into a high-frequency series. With low frequency sequences Two frequency bands were used, and then the instantaneous phase of the low-frequency sequence was calculated using the Hilbert transform. amplitude envelope of high-frequency sequences Divide the range into K equal partitions. For each partition... Calculate the corresponding partitions average And obtain the normalized value. : Obtain the sequence And calculate the Shannon entropy of the sequence: The following formula is used to obtain... : Calculate the MI index of the low-frequency delta band for the θ, α, and β bands respectively. The phase amplitude coupling characteristic matrix of each channel is obtained. .

6. The method for recognizing epilepsy EEG signals according to claim 1, characterized in that, The graph attention network model includes a graph attention network and a node feedforward network; The working mechanism of the graph attention network is as follows: First, calculate the attention index for different nodes. In the formula, W∈RF′×F is a learnable parameter matrix. For a shared attention mechanism; In the formula, ∈R 2F′, For transpose, For merging operations; In the formula, Represents a node The first-order nearest neighbor node; In the formula, h represents the h-th attention operation. This indicates a merge operation. Indicates the activation function; The working mechanism of the node feedforward network is as follows: In the formula, W1, W2 ∈ RF×F′, RF′×F′′, Let ∈RF′ and RF′′ be the learnable coefficient matrix and bias, respectively, where F represents the dimension of the input data and F′ represents the dimension of the output data. This is represented by the Sigmoid activation function.

7. The method for recognizing epileptic EEG signals according to claim 1, characterized in that, The EEG signals include signals during the non-ictal period, signals during the preictal period, and signals during the ictal period.

8. A system for recognizing epilepsy EEG signals, characterized in that, include: The edge feature calculation module is used to acquire EEG signals and calculate phase synchronization information between different channels as edge features based on the EEG signals. The node feature calculation module is used to filter the EEG signal, calculate and extract the temporal features, dynamic features and phase amplitude coupling features corresponding to different bands of each channel, and fuse them into node features. The judgment and recognition module is used to integrate the obtained edge features and node features into a graph, classify them using a graph attention network model, and judge and recognize the EEG signals based on the classification results.

9. A terminal device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 7.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 7.

Images