WO2022241710A1 - 一种癫痫发作区定位系统、设备及介质 - Google Patents

一种癫痫发作区定位系统、设备及介质 Download PDF

Info

Publication number
WO2022241710A1
WO2022241710A1 PCT/CN2021/094805 CN2021094805W WO2022241710A1 WO 2022241710 A1 WO2022241710 A1 WO 2022241710A1 CN 2021094805 W CN2021094805 W CN 2021094805W WO 2022241710 A1 WO2022241710 A1 WO 2022241710A1
Authority
WO
WIPO (PCT)
Prior art keywords
epilepsy
graph
data
seizure
feature
Prior art date
Application number
PCT/CN2021/094805
Other languages
English (en)
French (fr)
Inventor
窦永霖
詹阳
Original Assignee
中国科学院深圳先进技术研究院
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by 中国科学院深圳先进技术研究院 filed Critical 中国科学院深圳先进技术研究院
Priority to PCT/CN2021/094805 priority Critical patent/WO2022241710A1/zh
Publication of WO2022241710A1 publication Critical patent/WO2022241710A1/zh

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
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/07—Endoradiosondes

Definitions

  • the present invention relates to the field of medical image processing, in particular to an epileptic seizure area positioning system, equipment and medium.
  • Epilepsy is a chronic disease in which the sudden abnormal discharge of brain neurons leads to transient brain dysfunction, and it belongs to the nervous system disease.
  • the current method of treating epilepsy is generally to locate the epileptic seizure area and remove the epileptic seizure area through surgery.
  • Data related to epilepsy include epilepsy spontaneous data and epilepsy-induced data. Doctors can obtain the following data before surgery:
  • Spontaneous epilepsy data according to the three-dimensional electroencephalogram, the patient's spontaneous data can be recorded, and the electrode electroencephalogram (EEG) signals (EEG) of sleep and waking states under epileptic seizures and interepilepsy periods (without seizures) can be obtained.
  • EEG electrode electroencephalogram
  • Cortico-cortical evoked potential can be obtained by actively stimulating the electrodes. By stimulating a pair of electrode sites in the brain to record the response of the whole brain electrodes, measure electrodes or brain regions Effective connection between.
  • the prior art Before surgery, doctors often determine the location of the epileptic seizure area based on spontaneous epilepsy data combined with observational images such as MRI and CT images, and find out the brain area that needs to be resected or thermally coagulated through the determined electrode position of the seizure area, and then perform the operation.
  • epilepsy-induced data often account for a small proportion in locating epileptic seizure areas. Therefore, the prior art generally has the following defects:
  • the existing technology is relatively simple to process the EEG data of spontaneous epilepsy, and most of them only use simple signal processing means for processing, which often misses potential characteristic information.
  • the existing technology basically adopts the self-supervision method, which requires manual labeling of the electrode site or manual judgment, which is time-consuming and laborious, and the processing efficiency is low. Due to the difference in experience among clinical professionals, the diagnosis becomes difficult to evaluate.
  • the present invention provides an epileptic seizure area positioning system, equipment and medium.
  • the specific plan is as follows:
  • a seizure zone localization system comprising,
  • Data acquisition unit used to acquire epilepsy spontaneous data, epilepsy-induced data and epilepsy image data, the epilepsy-induced data includes EEG signals at the electrode sites under the induced stimulation;
  • Feature extraction unit used to perform feature extraction on the epilepsy spontaneous data, and obtain graph feature information
  • Structure construction unit used to obtain the EEG signals of multiple electrode sites under the same evoked stimulus in the epilepsy induced data, and construct the graph structure information through the significance test, and the graph structure information includes the brain signal under the evoked stimulus The adjacency matrix diagram of the connection relationship;
  • Seizure area acquisition unit for clustering the graph structure information and the graph feature information based on an unsupervised clustering process, and acquiring seizure area electrode sites, based on the seizure area electrode sites and the epilepsy image data Get the seizure zone.
  • the feature extraction unit specifically includes:
  • a preprocessing module used to acquire spontaneous EEG signals from the epilepsy spontaneous data, and preprocess the spontaneous EEG signals
  • Feature extraction module used for feature extraction of the preprocessed spontaneous EEG signal to obtain time-frequency features
  • An information acquisition module used to process the time-frequency features based on a deep learning network to obtain time-frequency features after dimensionality reduction, and use the time-frequency features after dimensionality reduction as the graph feature information.
  • the seizure area acquisition unit is provided with a clustering module
  • Clustering module for constructing a graph filter according to the graph structure information, and constructing a graph representation feature according to the graph feature information;
  • Each graph convolution includes matrix multiplication of the graph filter and the graph representation features, and the number of graph convolutions is performed according to the intra-cluster distance. adaptive judgment;
  • a plurality of clusters are obtained by clustering the fusion features, and an seizure area cluster is obtained by screening, and the seizure area cluster only includes the electrode sites of the seizure area.
  • the epilepsy image data includes CT images and MRI images
  • the seizure area acquisition unit is also provided with a positioning module
  • Positioning module used for performing image registration on the CT image and the MRI image to obtain first image data; judging the electrode position of the electrode site in the attack area in the image according to the first image data; matching the Electrode positions and preset image templates were used to capture seizure areas in the brain.
  • two types of clustering are performed on the fusion features to obtain ictal area clusters and non-ictal area clusters, the ictal area clusters only include the ictal area electrode sites, and the non-ictal area clusters only include Including non-ictal area electrode sites;
  • the spontaneous data include the spontaneous EEG signals of epileptic patients during and during epileptic seizures;
  • the evoked data includes when a pair of electrode sites are stimulated, EEG signals of all electrode sites in the brain under the stimulation.
  • the preprocessing includes removing artifacts, removing power frequency interference, filtering, downsampling, and castration;
  • the feature extraction module performs feature extraction on the preprocessed spontaneous EEG signal through Fourier transform or wavelet transform to obtain time-frequency features.
  • the construction process of the graph filter includes:
  • a graph filter is constructed from the Laplacian matrix.
  • the expression of the Laplacian matrix is
  • L s represents the Laplacian matrix
  • D represents the degree matrix
  • A represents the adjacency matrix
  • G represents the graph filter
  • I represents the identity matrix
  • L s represents the Laplacian matrix
  • the deep learning network includes an autoencoder model
  • the information acquisition module processes the time-frequency features based on the autoencoder model.
  • the process of obtaining graph feature information specifically includes:
  • a computer device comprising:
  • processors one or more processors
  • memory for storing one or more programs
  • the one or more processors implement the following processing:
  • the data acquisition unit acquires epilepsy spontaneous data, epilepsy-induced data and epilepsy image data
  • the feature extraction unit performs feature extraction on the epileptic spontaneous data to obtain graph feature information
  • the structure construction unit acquires the EEG signals of multiple electrode sites under the same evoked stimulus from the epilepsy-induced data, and constructs graph structure information through a significance test, and the graph structure information includes the brain connection relationship under the evoked stimulus adjacency matrix graph;
  • the seizure zone acquisition unit clusters the graph structure information and the graph feature information based on an unsupervised clustering process, acquires seizure zone electrode sites, and acquires epileptic seizures according to the seizure zone electrode sites and the epilepsy image data. Area.
  • a computer-readable storage medium on which a computer program is stored, and when the program is executed by a processor, the following processes are realized:
  • the data acquisition unit acquires epilepsy spontaneous data, epilepsy-induced data and epilepsy image data
  • the feature extraction unit performs feature extraction on the epileptic spontaneous data to obtain graph feature information
  • the structure construction unit acquires the EEG signals of multiple electrode sites under the same evoked stimulus from the epilepsy-induced data, and constructs graph structure information through a significance test, and the graph structure information includes the brain connection relationship under the evoked stimulus adjacency matrix graph;
  • the seizure zone acquisition unit clusters the graph structure information and the graph feature information based on an unsupervised clustering process, acquires seizure zone electrode sites, and acquires epileptic seizures according to the seizure zone electrode sites and the epilepsy image data. Area.
  • the present invention provides an epileptic seizure area positioning system, equipment and medium, which organically combines epilepsy-induced data, epilepsy spontaneous data and epilepsy image data, and accurately and quickly obtains the epilepsy data for different patients through signal processing and machine learning methods.
  • the position of the electrode or brain area of the epileptic seizure area can help the doctor locate the epileptic seizure area before the operation, and assist the doctor to make a judgment.
  • the feature relationship is represented by the epilepsy spontaneous data
  • the structural connection relationship is represented by the epilepsy-induced data
  • the location and judgment of the epileptic seizure area is performed by integrating the feature relationship and the structural connection relationship.
  • the unsupervised clustering method compared with supervised and semi-supervised learning, it does not require time-consuming training, nor does it need to use a large number of manually labeled samples for pre-training or supervised learning, shortening the detection time and improving detection efficiency.
  • the model is an adaptive graph convolution, it can adaptively judge the number of convolution iterations, so it can perform targeted detection according to the actual condition of different patients, with strong generalization ability and high detection accuracy.
  • Fig. 1 is a structural diagram of an epileptic seizure area positioning system according to Embodiment 1 of the present invention
  • Fig. 2 is a schematic diagram of the principle of the epileptic seizure area positioning system in Embodiment 1 of the present invention
  • Fig. 3 is a schematic diagram of an epilepsy-induced data electrode according to Embodiment 1 of the present invention.
  • Fig. 4 is a schematic diagram of electrode induction of epilepsy-induced data in Embodiment 1 of the present invention.
  • Fig. 5 is the flow chart of feature extraction of embodiment 1 of the present invention.
  • Fig. 6 is a schematic diagram of the autoencoder model of Embodiment 1 of the present invention.
  • Fig. 7 is an electrode site adjacency matrix diagram of Example 1 of the present invention.
  • Fig. 8 is a clustering flowchart of Embodiment 1 of the present invention.
  • FIG. 9 is a flow chart of adaptive graph convolution in Embodiment 1 of the present invention.
  • Fig. 10 is a diagram of the experimental results of the epileptic seizure area positioning system in Example 1 of the present invention.
  • Fig. 11 is a schematic structural diagram of a computer device according to Embodiment 2 of the present invention.
  • the present invention studies the connection relationship of the epilepsy network through CCEP-induced points, and constructs graph structure information.
  • the map feature information is obtained by extracting the EEG feature information in the epilepsy spontaneous data.
  • Combining graph feature information with graph structure information using machine learning-based adaptive graph convolution method for unsupervised clustering, to obtain two categories of epileptic seizures and epileptic non-seizures, helping doctors locate the location of seizures before surgery, Assist doctors in making judgments.
  • the self-adaptive unsupervised clustering is a new type of clustering algorithm, which is suitable for attribute graphs and utilizes
  • the edge connection relationship in the graph structure is used as a structural feature, and a better feature representation is obtained by aggregating the features of each node and its k-order neighbors through the graph convolution operation.
  • the invention provides an epileptic seizure area positioning system, equipment and medium.
  • the seizure area is located, the spontaneous data is used to represent the characteristic relationship, and the induced data is used to represent the structural connection relationship, fully considering the impact of the brain connection relationship on epilepsy.
  • the adaptive graph convolution method does not require pre-training or supervised learning with manually labeled samples, nor does it need to train and learn a large number of existing samples, which greatly reduces labor costs, shortens detection time, and improves detection efficiency. Since the model is an adaptive graph convolution, it can adaptively judge the number of convolution iterations, so it can perform targeted detection according to different patients, and has strong generalization ability.
  • This embodiment proposes an epileptic seizure area positioning system.
  • the block diagram of the system is shown in Figure 1 of the specification, and the principle diagram is shown in Figure 2 of the specification.
  • the specific plan is as follows:
  • An epileptic seizure area positioning system includes a data acquisition unit 1 , a feature extraction unit 2 , a structure construction unit 3 and an seizure area acquisition unit 4 .
  • the data acquisition unit 1 is connected to the feature extraction unit 2 and the structure construction unit 3 respectively
  • the seizure area acquisition unit 4 is connected to the feature extraction unit 2 and the structure construction unit 3 respectively.
  • Data acquisition unit 1 used to acquire epilepsy spontaneous data, epilepsy-induced data and epilepsy image data, the epilepsy-induced data includes EEG signals at electrode sites under evoked stimulation;
  • Feature extraction unit 2 used to obtain graph feature information by performing feature extraction on epilepsy spontaneous data
  • Structure construction unit 3 used to obtain the EEG signals of multiple electrode sites under the same evoked stimulus in the epilepsy-induced data, and construct the graph structure information through the significance test.
  • the graph structure information includes the brain connection relationship under the evoked stimulus adjacency matrix graph;
  • Seizure area acquisition unit 4 used to cluster the graph structure information and graph feature information based on the unsupervised clustering process, obtain the seizure area electrode sites, and obtain the seizure area according to the seizure area electrode sites and epilepsy image data.
  • the data acquisition unit 1 is used to acquire epilepsy image data, epilepsy spontaneous data and epilepsy induced data.
  • the epilepsy image data includes CT images and MRI images.
  • the epilepsy spontaneous data is the spontaneous EEG signal data collected by the hospital during the epileptic seizures and during the epileptic seizures (including awake and sleeping states), and the data during the epileptic seizures is the most important.
  • CCEP Cortico-cortical evoked potential
  • cortical-cortical evoked potential By stimulating a pair of electrode sites, the EEG signals of all intracranial electrode sites under this evoked stimulation can be obtained. Stimulation of all electrode pairs yields an evoked brain connectivity relationship that allows exploration of efficient and functional connectivity.
  • the brain electrodes are set in pairs, and the EEG responses of all electrode sites are obtained by stimulating a pair of electrodes. As shown in Figure 3 of the specification, when a pair of electrode sites in Figure 3 is stimulated, the EEG responses of all electrode sites in Figure 4 of the specification can be obtained.
  • Feature extraction unit 2 used to obtain graph feature information by performing feature extraction on epilepsy spontaneous data.
  • the feature extraction unit 2 specifically includes a preprocessing module 21 , a feature extraction module 22 and an information acquisition module 23 .
  • Preprocessing module 21 used to obtain spontaneous EEG signals from epilepsy spontaneous data, and preprocess the spontaneous EEG signals;
  • Feature extraction module 22 for performing feature extraction on the preprocessed spontaneous EEG signal to obtain time-frequency features
  • Information acquisition module 23 used to process the time-frequency feature based on the deep learning network to obtain the time-frequency feature after dimensionality reduction, and use the time-frequency feature after dimensionality reduction as graph feature information.
  • the preprocessing module is required to preprocess the original spontaneous EEG signal.
  • Preprocessing includes signal processing methods such as removing artifacts, removing power frequency interference, filtering, downsampling, castration, etc.; after obtaining the preprocessed signal, feature extraction
  • the module extracts signal features through Fourier transform or wavelet transform, and obtains time-frequency features. For example, using wavelet power spectrum can obtain various power features in different frequency bands; the information acquisition module uses automatic coding based on deep learning for time-frequency features The time-frequency features after dimensionality reduction are obtained through unsupervised methods, and the time-frequency features are used as graph feature information.
  • the acquisition process of graph feature information specifically includes: constructing feature vectors based on time-frequency features; performing dimensionality reduction and feature learning through the encoding part of the autoencoder model to obtain a low-dimensional hidden feature representation of an intermediate state;
  • the decoding part converts the low-dimensional hidden feature representation into a feature vector; adjusts the network parameters based on the loss function; iterates the above steps until the iteration stop condition is met, and uses the low-dimensional feature representation that satisfies the iteration stop condition as graph feature information.
  • time-domain features can be obtained after preprocessing operations, or time-frequency features can be obtained by using Fourier transform and wavelet transform.
  • a feature vector that can represent the original waveform is constructed. Afterwards, dimensionality reduction and feature learning are performed through the encoding part of the autoencoder model. After the encoding process, a low-dimensional hidden feature representation of an intermediate state will be obtained, and then the low-dimensional feature representation will be converted back to the original feature vector through the decoding process. Through the loss function Adjust the network parameters, and finally the input features obtained after multiple rounds of iterations are close to the output features, and the low-dimensional feature representation of the intermediate state can be used as the graph feature information.
  • a deep learning based autoencoder model is used at the end of the feature extraction process.
  • This model is an unsupervised feature extraction method, which is suitable for the unsupervised clustering process, so that the whole system is an unsupervised learning process.
  • the autoencoder model (Autoencoder) is a neural network model whose goal is to copy the input features to the output. By compressing the input features into a low-dimensional hidden space feature representation, and then reconstructing the low-dimensional feature representation to increase the dimension, so that the reconstructed output features are closer to the input features, it is considered that the low-dimensional hidden space feature representation is Can represent the original high-dimensional features.
  • the network structure is shown in Figure 6 of the instruction manual.
  • Each layer can be connected by a fully connected layer, and activation functions such as ReLu and Sigmoid are added in the middle to realize nonlinear transformation.
  • the loss function can choose mean square error, etc.; Calculate the distance between the input feature and the output feature.
  • it can be realized by using commonly used deep learning frameworks such as TensorFlow or Pytorch.
  • autoencoder model or its variant models such as the variable autoencoder model, can be used as a means of feature extraction.
  • Structure construction unit 3 used to obtain the EEG signals of multiple electrode sites under the same evoked stimulus in the epilepsy-induced data, and pass the significance test to construct the graph structure information, which includes the brain connection relationship under the evoked stimulus The adjacency matrix graph of .
  • the recorded evoked EEG signals are generally repeated records.
  • the epilepsy-induced data include the EEG signals of all electrode sites in the brain under the evoked stimulation obtained when a pair of electrode sites are stimulated.
  • the EEG signals under the same evoked stimulus that is, the EEG signals under the stimulation of the same electrode pair are screened out.
  • the structure construction unit 3 selects the baseline and EEG responses for the significance test of the screened EEG signals, and selects the part whose significance level meets the preset conditions to obtain the adjacency matrix diagram of the brain connection relationship under the evoked stimulation,
  • the welcome matrix graph is the graph structure information.
  • the electrode site adjacency matrix is shown in Figure 7 of the specification, where the vertical axis represents the evoked stimulus, that is, the pair of electrodes that give the stimulus, the horizontal axis represents all the responding electrodes, and different gray levels represent different significance levels.
  • the significance test is to make a hypothesis about the parameters or the overall distribution form of the population (random variable) in advance, and then use the sample information to judge whether the hypothesis (alternative hypothesis) is reasonable, that is, to judge the real situation of the population and the original hypothesis Is there any significant difference.
  • the significance test is to judge whether the difference between the sample and the hypothesis is purely chance variation, or is caused by the inconsistency between the hypothesis and the overall real situation.
  • the significance test is to test the assumptions made by the population, and its principle is to accept or deny the assumptions based on the "principle of practical impossibility of small probability events".
  • Seizure area acquisition unit 4 cluster the graph structure information and graph feature information based on the unsupervised clustering process, obtain the seizure area electrode positions, and obtain the seizure area according to the seizure area electrode positions and preset image data.
  • the unsupervised clustering process the adaptive graph convolution method is used to process the graph structure signal and graph feature information. Compared with self-supervised learning and semi-supervised learning, it does not need to use a large number of samples marked by doctors for pre-training or supervised learning.
  • the seizure area acquisition unit 4 includes a clustering module 41 and a positioning module 42 .
  • the clustering module 41 specifically includes: using the graph structure information to construct a graph filter through Laplace transform, and constructing graph representation features according to the graph feature information; performing feature fusion through graph convolution operations to obtain fusion features, and each graph convolution includes The matrix multiplication of the graph filter and the graph representation feature is performed, and the number of graph convolutions is adaptively judged according to the distance within the cluster; multiple clusters are obtained by clustering according to the fusion feature, and the electrode sites of the seizure area are obtained by screening.
  • the map structure information is obtained from the epilepsy-induced data, which contains a large number of relevant information on electrode sites.
  • Each electrode site is regarded as a node in the graph structure, each node has its own feature, and the feature is associated with the graph feature information, and the feature information of all nodes together constitute the complete graph feature information.
  • the adjacency matrix obtained from the epilepsy-induced data reflects the connection relationship of all nodes.
  • the adaptive unsupervised clustering provided in this embodiment is a new type of clustering algorithm, which is suitable for attribute graphs, and uses the edge connection relationship in the graph structure as a structural feature, and aggregates each node and Its k-order neighbor features to get better feature representation.
  • V ⁇ V 1 , V 2 ,...,V n ⁇ is a set of nodes
  • E is a set of edges
  • L s represents the Laplacian matrix
  • D represents the degree matrix
  • A represents the adjacency matrix
  • I represents the identity matrix
  • L s represents the Laplacian matrix.
  • the graph structure information is used to construct the graph filter in the form of Laplace transform, and the graph representation features are constructed according to the graph feature information.
  • the multiplication of the two matrices of the graph filter and the graph representation feature is regarded as a graph convolution process, and the adaptive graph convolution operation is performed to combine the two features to obtain the fusion feature, where the number of convolutions is adaptive according to the distance within the cluster judge.
  • Clustering is performed according to fusion features to obtain multiple clusters, and each cluster contains multiple nodes. Seizure clusters are selected from multiple clusters, the seizure clusters only include seizure zone electrode sites, and corresponding seizure zones are obtained according to the seizure zone electrode sites.
  • the seizure area cluster also includes multiple sub-clusters. Since the model is an adaptive graph convolution, it can adaptively judge the number of convolution iterations, so it can perform targeted detection according to the condition of different patients, and has strong generalization.
  • the process of adaptive graph convolution is shown in Figure 9 of the specification.
  • the maximum number of iterations max_iter is defined, and t is defined as the number of cycles.
  • intra(C (t) ) is the sum of the intra-cluster distances of all clusters.
  • the adaptive convolution The process is to decide whether to jump out of the loop by comparing the distance within the cluster in each iteration. When the loop is jumped out, the current result is considered to be the best, and the value of k obtained at the same time is the number of convolutions of the graph convolution. Finally, the k-order graph signal G k X is obtained, and then the clustering result is obtained through spectral clustering. Iterate according to the distance change in the cluster, and perform a convolution in each iteration to obtain the characteristics of the k-order neighbors of the node
  • two clusters are obtained by clustering, namely the ictal cluster and the non-ictal cluster.
  • Clustering based on fusion features can only get two clusters, which avoids the screening process.
  • the seizure zone cluster only includes the seizure zone electrode sites
  • the non-seizure zone cluster only includes the non-seizure zone electrode sites
  • the position of the seizure zone electrode sites in the image corresponds to the seizure zone.
  • each cluster contains multiple nodes, and each node corresponds to an electrode site
  • the ictal area cluster and the non-ictal area cluster here are both electrode site sets, which contain multiple electrode sites
  • the ictal area cluster Contains one or more electrode sites in the ictal area
  • the non-ictal area contains one or more electrode sites in the non-ictal area.
  • clustering results in more than two clusters.
  • the category is divided into multiple areas such as epilepsy initiation area, rapid dissemination area, irritating area, and non-epileptic area.
  • the specific category selection can be selected according to the actual situation.
  • the positioning module 42 After obtaining the electrode sites of the seizure area, electrode positioning is required.
  • the positioning module 42 combines the epilepsy image data and the seizure area electrode sites to obtain the position of the seizure area electrode sites, and then combines the preset image template to obtain the seizure area of the brain.
  • the positioning module specifically includes: obtaining a CT image and an MRI image of an epileptic patient, and performing image registration on the CT image and the MRI image to obtain first image data. According to the first image data, the electrode position of the electrode site of the seizure area in the image is judged; matching the electrode position and the preset image template to obtain the seizure area of the brain.
  • the detection results of the epileptic seizure area are shown in Figure 10 of the description.
  • two kinds of deep and shallow circular areas correspond to two kinds of electrode sites. According to the depth of the electrodes, it is judged whether it is in the seizure area.
  • the shallower electrodes correspond to the position of the non-seizure area, and the deeper electrodes correspond to the position of the seizure area.
  • the clustering results of the system are compared with the labels given by the hospital, and good results are obtained, which proves the applicability and reliability of the system provided in this embodiment.
  • the brain regions such as the posterior parahippocampal gyrus, temporal lobe, and temporal pole plane are related to the seizure area.
  • the electrodes are corresponding to the brain regions, and the final results are multiple regions such as the posterior parahippocampal gyrus, temporal lobe, temporal pole plane, and inferior frontal gyrus, basically the seizure areas to be removed by the actual surgery Including, it proves the practicality and usability of this system.
  • This embodiment provides an epileptic seizure area positioning system, which organically combines epilepsy-induced data and epilepsy spontaneous data, characterizes feature relationships through epilepsy spontaneous data, epilepsy-induced data represents structural connection relations, and integrates feature relations and structural connection relations to epilepsy.
  • the location of the attack area is judged, which is comprehensive and accurate.
  • signal processing, deep learning, and machine learning are integrated to discover potential features in the data and obtain more accurate detection results.
  • the unsupervised clustering method compared with supervised and semi-supervised learning, it does not require time-consuming training, nor does it need to use a large number of manually labeled samples for pre-training or supervised learning, shortening the detection time and improving detection efficiency.
  • the model is an adaptive graph convolution, it can adaptively judge the number of convolution iterations, so it can carry out targeted detection according to the actual condition of different patients, with strong generalization ability and high detection accuracy.
  • Figure 11 of the description is a schematic structural diagram of a computer device provided by Embodiment 2 of the present invention.
  • the computer device 12 shown in Fig. 11 of the specification is only an example, and should not impose any limitation on the functions and scope of use of the embodiments of the present invention.
  • computer device 12 takes the form of a general-purpose computing device.
  • Components of computer device 12 may include, but are not limited to: one or more processors or processing units 16 , system memory 28 , bus 18 connecting various system components including system memory 28 and processing unit 16 .
  • Computer device 12 typically includes a variety of computer system readable media. These media can be any available media that can be accessed by device computer 12 and include both volatile and nonvolatile media, removable and non-removable media.
  • System memory 28 may include computer system readable media in the form of volatile memory.
  • Computer device 12 may also communicate with one or more external devices 14 (e.g., keyboards, pointing devices, displays, etc.), and with one or more devices that enable a user to interact with
  • the computing device 12 is capable of communicating with any device that communicates with one or more other computing devices.
  • the processing unit 16 executes various functional applications and data processing by running the program stored in the system memory 28, for example, realizing a control method of an epileptic seizure zone positioning system, the method comprising:
  • the data acquisition unit 1 acquires epilepsy spontaneous data, epilepsy-induced data, and epilepsy image data; the epilepsy-induced data includes EEG signals under inducing stimulation.
  • the feature extraction unit 2 acquires graph feature information by performing feature extraction on epilepsy spontaneous data.
  • the structure construction unit 3 acquires the EEG signals of multiple electrode sites under the same evoked stimulus from the epilepsy-induced data, and constructs the graph structure information through the significance test.
  • the graph structure information includes the brain connection relationship under the evoked stimulus Adjacency Matrix Diagram.
  • the seizure area acquisition unit 4 clusters the graph structure information and graph feature information based on the unsupervised clustering process, acquires the electrode sites of the seizure area, and acquires the seizure area according to the electrode sites of the seizure area and the epilepsy image data.
  • step 102 and step 103 may be performed simultaneously or sequentially.
  • a control method of an epileptic seizure area positioning system is applied to a specific computer device, and the method is stored in a memory.
  • the actuator executes the memory, the method will be run to locate the epileptic seizure area, which is quick and easy to use. It is convenient and has a wide range of applications.
  • processor can also implement the technical solution provided by any embodiment of the present invention.
  • Embodiment 3 provides a computer-readable storage medium, on which a computer program is stored, and when the program is executed by a processor, it implements a control method such as a seizure zone positioning system, the method comprising:
  • the data acquisition unit 1 acquires epilepsy spontaneous data, epilepsy-induced data, and epilepsy image data; the epilepsy-induced data includes EEG signals under inducing stimulation.
  • the feature extraction unit 2 acquires graph feature information by performing feature extraction on epilepsy spontaneous data.
  • the structure construction unit 3 acquires the EEG signals of multiple electrode sites under the same evoked stimulus from the epilepsy-induced data, and constructs the graph structure information through the significance test.
  • the graph structure information includes the brain connection relationship under the evoked stimulus Adjacency Matrix Diagram.
  • the seizure area acquisition unit 4 clusters the graph structure information and graph feature information based on the unsupervised clustering process, acquires the electrode sites of the seizure area, and acquires the seizure area according to the electrode sites of the seizure area and the epilepsy image data.
  • step 102 and step 103 may be performed simultaneously or sequentially.
  • the computer storage medium in this embodiment may use any combination of one or more computer-readable media.
  • the computer readable medium may be a computer readable signal medium or a computer readable storage medium.
  • a computer-readable storage medium may be, for example but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device, or device, or any combination thereof. More specific examples (non-exhaustive list) of computer readable storage media include: electrical connections with one or more leads, portable computer disks, hard disks, random access memory (RAM), read only memory (ROM), Erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination of the above.
  • a computer-readable storage medium may be any tangible medium that contains or stores a program that can be used by or in conjunction with an instruction execution system, apparatus, or device.
  • a control method of an epileptic seizure area positioning system is applied to a computer-readable storage medium, on which a computer program is stored, and when the program is executed by a processor, the control of the epileptic seizure area positioning system provided by the present invention is realized.
  • the steps of the method are simple and quick, easy to store and difficult to lose.
  • the present invention provides an epileptic seizure area positioning system, equipment and medium, which organically combines epilepsy-induced data, epilepsy spontaneous data and epilepsy image data, and accurately and quickly obtains the epilepsy data for different patients through signal processing and machine learning methods.
  • the position of the electrode or brain area of the epileptic seizure area can help the doctor locate the epileptic seizure area before the operation, and assist the doctor to make a judgment.
  • the feature relationship is represented by the epileptic spontaneous data
  • the structural connection relationship is represented by the epilepsy-induced data
  • the location and judgment of the epileptic seizure area is performed by integrating the feature relationship and the structural connection relationship.
  • each module or each step of the present invention described above can be realized by a general-purpose computing device, and they can be concentrated on a single computing device, or distributed on a network formed by multiple computing devices.
  • they can be implemented with executable program codes of computer devices, so that they can be stored in storage devices and executed by computing devices, or they can be made into individual integrated circuit modules, or a plurality of modules in them Or the steps are fabricated into a single integrated circuit module to realize.
  • the present invention is not limited to any specific combination of hardware and software.

Landscapes

  • Life Sciences & Earth Sciences (AREA)
  • Health & Medical Sciences (AREA)
  • Medical Informatics (AREA)
  • Biophysics (AREA)
  • Pathology (AREA)
  • Engineering & Computer Science (AREA)
  • Biomedical Technology (AREA)
  • Heart & Thoracic Surgery (AREA)
  • Physics & Mathematics (AREA)
  • Molecular Biology (AREA)
  • Surgery (AREA)
  • Animal Behavior & Ethology (AREA)
  • General Health & Medical Sciences (AREA)
  • Public Health (AREA)
  • Veterinary Medicine (AREA)
  • Measurement And Recording Of Electrical Phenomena And Electrical Characteristics Of The Living Body (AREA)

Abstract

一种癫痫发作区定位系统、设备及介质。系统包括，数据获取单元（1）：用于获取癫痫自发数据、癫痫诱发数据和癫痫影像数据；特征提取单元（2）：用于通过对癫痫自发数据进行特征提取，获取图特征信息；结构构建单元（3）：用于在癫痫诱发数据中获取脑电信号，通过显著性检验构建图结构信息；发作区获取单元（4）：用于基于无监督聚类过程对图结构信息和图特征信息进行聚类，获取发作区电极位点，根据发作区电极位点和癫痫影像数据获取癫痫发作区。该系统将癫痫诱发数据、癫痫自发数据和癫痫影像数据有机结合，通过信号处理和机器学习的方法精准、快速地得到病人的癫痫发作区电极或脑区位置，帮助医生手术之前定位癫痫发作区位置，辅助医生做出判断。

Description

一种癫痫发作区定位系统、设备及介质 技术领域
本发明涉及医学图像处理领域，具体而言，涉及一种癫痫发作区定位系统、设备及介质。
背景技术
癫痫疾病是大脑神经元突发性异常放电，导致短暂的大脑功能障碍的一种慢性疾病，属于神经系统疾病。当前治疗癫痫的方法一般为定位癫痫发作区域，通过手术的方式将癫痫发作区去除。
对于癫痫发作区位置的检测，相关研究较少。有研究者使用相位锁相值等方法将电极位点进行分类来实现，但这样做需要电极位点标签且没有考虑全脑连接关系。
癫痫疾病的相关数据包括癫痫自发数据和癫痫诱发数据，医生在手术前可以得到如下数据：
1.使用立体脑电图在病人颅内放置侵入式电极，用来记录颅内立体脑电图(SEEG)。
2.自发癫痫数据；根据立体脑电图可以记录病人的自发数据，得到癫痫发作、癫痫间期(癫痫未发作)下睡眠和清醒状态的电极脑电信号(EEG)。
3.癫痫诱发数据：通过主动给电极刺激可以得到皮层-皮质间诱发电位(Cortico-cortical evoked potintial)数据，通过刺激颅内一对电极位点以记录全脑电极的反应，测量电极或脑区之间的有效性连接。
在手术之前，医生往往根据自发癫痫数据结合MRI和CT影像等观测影像资料确定癫痫发作区的位置，通过确定的发作区电极位置找出需要切除 或热凝的脑区，之后进行手术。而癫痫诱发数据在定位癫痫发作区中往往占比不大。由此，现有技术普遍存在如下缺陷：
1.现有技术多只通过自发癫痫数据进行判断而不考虑诱发癫痫数据，忽略了有效性连接关系信息对癫痫的影响。
2.现有技术对自发癫痫脑电数据处理比较简单，大多只用简单的信号处理手段进行处理，往往会漏掉潜在的特征信息。
3.现有技术基本采用自监督方法，需要人工标注电极位点的标签或需要人工判断，费时费力，处理效率，由于临床专业人士之间的经验差异，诊断变得难以评估。
4.现有技术泛化能力弱，需要大量已有样本才能进行检测，且不能根据每个病人的实际病情进行针对性检测。
因此，需要一种高效、准确的癫痫发作区定位系统，能够解决上述问题。
发明内容
基于现有技术存在的问题，本发明提供了一种癫痫发作区定位系统、设备及介质。具体方案如下：
一种癫痫发作区定位系统，包括，
数据获取单元：用于获取癫痫自发数据、癫痫诱发数据和癫痫影像数据，所述癫痫诱发数据包括电极位点处于诱发刺激下的脑电信号；
特征提取单元：用于对所述癫痫自发数据进行特征提取，获取图特征信息；
结构构建单元：用于在所述癫痫诱发数据中获取多个电极位点处于同一诱发刺激下的脑电信号，通过显著性检验构建图结构信息，所述图结构信息包括处于诱发刺激下的脑连接关系的邻接矩阵图；
发作区获取单元：用于基于无监督聚类过程对所述图结构信息和所述图特征信息进行聚类，获取发作区电极位点，根据所述发作区电极位点和所述癫痫影像数据获取癫痫发作区。
在一个具体实施例中，所述特征提取单元具体包括：
预处理模块：用于在所述癫痫自发数据中获取自发脑电信号，对所述自发脑电信号进行预处理；
特征提取模块：用于对预处理后的自发脑电信号进行特征提取，获取时频特征；
信息获取模块：用于基于深度学习网络处理所述时频特征得到降维后的时频特征，将降维后的时频特征作为所述图特征信息。
在一个具体实施例中，所述发作区获取单元设置有聚类模块；
聚类模块：用于根据所述图结构信息构建图滤波器，根据所述图特征信息构建图表示特征；
通过图卷积操作进特征融合，获取融合特征，每次所述图卷积包括将所述图滤波器与所述图表示特征进行矩阵相乘，所述图卷积的次数根据簇内距离进行自适应判断；
对所述融合特征进行聚类得到多个簇，筛选得到发作区簇，所述发作区簇仅包括所述发作区电极位点。
在一个具体实施例中，所述癫痫影像数据包括CT影像和MRI影像；
所述发作区获取单元还设置有定位模块；
定位模块：用于对所述CT影像和所述MRI影像进行图像配准获取第一影像数据；根据所述第一影像数据判断所述发作区电极位点在影像中的电极位置；匹配所述电极位置和预设影像模板获取脑部的癫痫发作区。
在一个具体实施例中，对所述融合特征进行两类聚类，得到发作区簇和非发作区簇，所述发作区簇仅包括所述发作区电极位点，所述非发作区簇仅包括非发作区电极位点；
选取电极位点较少的簇作为所述发作区簇。
在一个具体实施例中，所述自发数据包括癫痫病人在癫痫发作时、癫痫发作期间的自发脑电信号；
所述诱发数据包括在刺激一对电极位点时，颅内所有电极位点在该刺激下的脑电信号。
在一个具体实施例中，所述预处理包括去除伪差、去除工频干扰、滤波处理、降采样处理、去势处理；
在一个具体实施例中，所述特征提取模块通过傅里叶变换或小波变换对预处理后的自发脑电信号进行特征提取，获取时频特征。
在一个具体实施例中，所述图滤波器的构建过程包括：
从所述图结构信息中的获取邻接矩阵；
对所述邻接矩阵进行拉普拉斯变换，构造出归一化的拉普拉斯矩阵；
根据所述拉普拉斯矩阵构造图滤波器。
在一个具体实施例中，所述拉普拉斯矩阵的表达式为
L s＝D -1/2AD -1/2
其中，L s表示拉普拉斯矩阵，D表示度矩阵，A表示邻接矩阵；
所述图滤波器的表达式为：
Figure PCTCN2021094805-appb-000001
其中，G表示图滤波器，I表示单位矩阵，L s表示拉普拉斯矩阵。
在一个具体实施例中，其特征在于，所述深度学习网络包括自动编码器模型，所述信息获取模块基于所述自动编码器模型处理所述时频特征。
在一个具体实施例中，所述图特征信息的获取流程具体包括：
根据所述时频特征构建出特征向量；
通过所述自动编码器模型中的编码部分进行降维和特征学习，得到一个中间状态的低维隐藏特征表示；
通过所述自动编码器模型中的解码部分将所述低维隐藏特征表示转换为所述特征向量；
基于损失函数调节网络参数；
迭代上述步骤，直至满足迭代停止条件，将满足迭代停止条件的低维特征表示作为图特征信息。
一种计算机设备，所述计算机设备包括：
一个或多个处理器；
存储器，用于存储一个或多个程序；
当所述一个或多个程序被所述一个或多个处理器执行，使得所述一个或多个处理器实现如下处理：
数据获取单元获取癫痫自发数据、癫痫诱发数据和癫痫影像数据；
特征提取单元对所述癫痫自发数据进行特征提取，获取图特征信息；
结构构建单元在所述癫痫诱发数据中获取多个电极位点处于同一诱发刺激下的脑电信号，通过显著性检验构建图结构信息，所述图结构信息包括处于诱发刺激下的脑连接关系的邻接矩阵图；
发作区获取单元基于无监督聚类过程对所述图结构信息和所述图特征信息进行聚类，获取发作区电极位点，根据所述发作区电极位点和所述癫痫影像数据获取癫痫发作区。
一种计算机可读存储介质，其上存储有计算机程序，该程序被处理器执行时实现如下处理：
数据获取单元获取癫痫自发数据、癫痫诱发数据和癫痫影像数据；
特征提取单元对所述癫痫自发数据进行特征提取，获取图特征信息；
结构构建单元在所述癫痫诱发数据中获取多个电极位点处于同一诱发刺激下的脑电信号，通过显著性检验构建图结构信息，所述图结构信息包括处于诱发刺激下的脑连接关系的邻接矩阵图；
发作区获取单元基于无监督聚类过程对所述图结构信息和所述图特征 信息进行聚类，获取发作区电极位点，根据所述发作区电极位点和所述癫痫影像数据获取癫痫发作区。
本发明具有如下有益效果：
本发明针对现有技术，提供了一种癫痫发作区定位系统、设备及介质，将癫痫诱发数据、癫痫自发数据和癫痫影像数据有机结合，通过信号处理和机器学习的方法精准快速得到针对不同病人的癫痫发作区电极或脑区位置，帮助医生手术之前定位癫痫发作区位置，辅助医生做出判断。
通过癫痫自发数据表征特征关系，癫痫诱发数据表征结构连接关系，综合特征关系和结构连接关系对癫痫发作区进行定位判断，相比于现有技术，考虑了全脑连接关系对癫痫的影响，综合性强、准确度高。
通过特征提取和表征手段，将信号处理、深度学习、机器学习进行整合，能够发现数据中潜在的特征，进而获取更准确的检测效果。
使用无监督聚类方法，相比于监督和半监督学习，无需耗时训练，也无需使用大量人工标注的样本进行预训练或监督学习，缩短检测时间，提高检测效率。
由于模型为自适应图卷积，可自适应的判断卷积迭代次数，因而能够根据不同病人的实际病情进行针对性检测，泛化能力强，检测准确度高。
为使本发明的上述目的、特征和优点能更明显易懂，下文特举较佳实施例，并配合所附附图，作详细说明如下。
附图说明
为了更清楚地说明本发明实施例的技术方案，下面将对实施例中所需要使用的附图作简单地介绍，应当理解，以下附图仅示出了本发明的某些实施例，因此不应被看作是对范围的限定，对于本领域普通技术人员来讲，在不付出创造性劳动的前提下，还可以根据这些附图获得其他相关的附图。
图1是本发明实施例1的癫痫发作区定位系统结构图；
图2是本发明实施例1的癫痫发作区定位系统原理示意图；
图3是本发明实施例1的癫痫诱发数据电极示意图；
图4是本发明实施例1的癫痫诱发数据电极诱发示意图；
图5是本发明实施例1的特征提取流程图；
图6是本发明实施例1的自动编码器模型的原理图；
图7是本发明实施例1的电极位点邻接矩阵图；
图8是本发明实施例1的聚类流程图；
图9是本发明实施例1的自适应图卷积流程图；
图10是本发明实施例1的癫痫发作区定位系统实验结果图；
图11是本发明实施例2的计算机设备的结构示意图。
附图标记：
1-数据获取单元；2-特征提取单元；3-结构构建单元；4-发作区获取单元；21-预处理模块；22-特征提取模块；23-信息获取模块；41-聚类模块；42-定位模块；12-计算机设备；14-外部设备；16-处理单元；18-总线；28-系统存储器。
具体实施方式
下面将结合本发明实施例中的附图，对本发明实施例中的技术方案进行清楚、完整地描述，显然，所描述的实施例仅是本发明的一部分实施例，而不是全部的实施例。基于本发明中的实施例，本领域普通技术人员在没有做出创造性劳动前提下所获得的所有其他实施例，都属于本发明保护的范围。
本发明通过CCEP诱发点位来研究癫痫网络的连接关系，构建图结构 信息。通过提取癫痫自发数据中的脑电特征信息得到图特征信息。将图特征信息和图结构信息结合，利用基于机器学习的自适应图卷积方法进行无监督聚类，得到癫痫发作区和癫痫非发作区两个类别，帮助医生手术之前定位癫痫发作区位置，辅助医生做出判断。
相比于传统的聚类算法(如Kmeans，谱聚类)只用节点特征进行聚类，本发明提供的自适应无监督聚类是一种新型的聚类算法，适用于属性图，并利用图结构中的边连接关系作为一种结构特征，通过图卷积操作聚合每个节点及其k阶邻居特征来得到更好的特征表征。
本发明提供了一种癫痫发作区定位系统、设备及介质。通过将癫痫自发数据和癫痫诱发数据有机结合，进行癫痫发作区定位，用自发数据表征特征关系，用诱发数据表征结构连接关系，充分考虑到脑连接关系对癫痫疾病的影响。通过信号处理方法与深度学习方法提取特征，可以发现潜在的特征，取得更好的检测效果。使用自适应图卷积方法，不需要用人工标注的样本进行预训练或监督学习，也无需对大量已有样本进行训练学习，大大减少了人力成本，缩短了检测时间，提高了检测效率。由于模型为自适应图卷积，可自适应的判断卷积迭代次数，因而能够根据不同病人进行针对性检测，泛化能力强。
实施例1
本实施例提出了一种癫痫发作区定位系统，系统的模块图如说明书附图1所示，原理图如说明书附图2所示。具体方案如下:
一种癫痫发作区定位系统，包括数据获取单元1、特征提取单元2、结构构建单元3和发作区获取单元4。数据获取单元1分别连接特征提取单元2和结构构建单元3，发作区获取单元4分别连接特征提取单元2和结构构建单元3。
数据获取单元1：用于获取癫痫自发数据、癫痫诱发数据和癫痫影像数 据，癫痫诱发数据包括电极位点处于诱发刺激下的脑电信号；
特征提取单元2：用于通过对癫痫自发数据进行特征提取，获取图特征信息；
结构构建单元3：用于在癫痫诱发数据中获取多个电极位点处于同一诱发刺激下的脑电信号，通过显著性检验构建图结构信息，图结构信息包括处于诱发刺激下的脑连接关系的邻接矩阵图；
发作区获取单元4：用于基于无监督聚类过程对图结构信息和图特征信息进行聚类，获取发作区电极位点，根据发作区电极位点和癫痫影像数据获取癫痫发作区。
具体地，数据获取单元1：用于获取癫痫影像数据、癫痫自发数据和癫痫诱发数据。其中，癫痫影像数据包括CT影像、MRI影像，癫痫自发数据为医院采集到的病人在癫痫发作时、癫痫发作期间(包括清醒和睡眠状态下)的自发脑电信号数据，其中癫痫发作期间数据最为重要。CCEP(Cortico-cortical evoked potential)称为皮层-皮质间诱发点位，通过刺激一对电极位点，得到颅内所有电极位点在此诱发刺激下的脑电信号。对所有电极对进行刺激后即可得到一种诱发的脑连接关系，可以探索有效和功能性连接。脑部电极成对设置，通过刺激一对电极进而获取所有电极位点的脑电反应。如说明书附图3所示，当刺激附图3中的一对电极位点时，可得到说明书附图4的所有电极位点的脑电反应。
特征提取单元2：用于通过对癫痫自发数据进行特征提取，获取图特征信息。其中，特征提取单元2具体包括预处理模块21、特征提取模块22和信息获取模块23。
预处理模块21：用于在癫痫自发数据中获取自发脑电信号，对自发脑电信号进行预处理；
特征提取模块22：用于对预处理后的自发脑电信号进行特征提取，获取时频特征；
信息获取模块23：用于基于深度学习网络处理时频特征得到降维后的时频特征，将降维后的时频特征作为图特征信息。
具体的特征提取过程如说明书附图5所示。首先需要预处理模块对原始自发脑电信号进行预处理，预处理包括去除伪差、去除工频干扰、滤波处理、降采样处理、去势等信号处理手段；得到预处理后信号后，特征提取模块通过傅里叶变换或小波变换等手段提取信号特征，获取时频特征，例如，使用小波功率谱可以得到不同频段下的各个功率特征；信息获取模块对时频特征使用基于深度学习的自动编码器模型等手段，通过无监督方法得到降维之后的时频特征，将该时频特征作为图特征信息。
图特征信息的获取流程具体包括：根据时频特征构建出特征向量；通过自动编码器模型中的编码部分进行降维和特征学习，得到一个中间状态的低维隐藏特征表示；通过自动编码器模型中的解码部分将低维隐藏特征表示转换为特征向量；基于损失函数调节网络参数；迭代上述步骤，直至满足迭代停止条件，将满足迭代停止条件的低维特征表示作为图特征信息。具体地，在原始记录到的自发信号基础上，进行预处理操作后能得到时域特征，或者使用傅里叶变换和小波变换得到时频特征。根据这些特征构建出能代表原始波形的特征向量。之后通过自动编码器模型中的编码部分进行降维和特征学习，编码过程结束会得到一个中间状态的低维隐藏特征表示，再通过解码过程将这个低维特征表示转回原始特征向量，通过损失函数调节网络参数，最后通过多轮迭代之后得到的输入特征和输出特征接近，即可使用中间状态的低维特征表示作为图特征信息。
特征提取过程的最后中用到了基于深度学习的自动编码器模型。该模型是无监督的特征提取手段，契合于之后聚类过程中的无监督，使得整个系统整体都是无监督的学习过程。以自动编码器模型为例，介绍无监督特征提取过程：自动编码器模型(Autoencoder)是一种神经网络模型，目标是将输入特征复制到输出。通过将输入特征压缩到低维的隐藏空间特征表示， 随后对这个低维特征表示进行重构升维，使得重构出来的输出特征与输入特征较为接近，则认为低维的隐藏空间特征表示即可代表原始高维的特征。
网络结构如说明书附图6所示，每一层可用全连接层进行连接，中间加上ReLu和Sigmoid等激活函数实现非线性变换，损失函数可以选择均方误差等；训练过程是通过设置损失函数计算输入特征与输出特征的距离实现。一般使用常用的深度学习框架如TensorFlow或Pytorch等都可以实现。
使用自动编码器模型或其变种模型，如可变自动编码器模型，都可以作为特征提取手段。
结构构建单元3：用于在癫痫诱发数据中获取多个电极位点处于同一诱发刺激下的脑电信号，通过显著性检验，构建图结构信息，图结构信息包括处于诱发刺激下的脑连接关系的邻接矩阵图。
基于癫痫诱发数据的获取原理，记录的诱发脑电信号一般是重复记录。癫痫诱发数据包括刺激一对电极位点时得到的颅内所有电极位点在此诱发刺激下的脑电信号。在重复记录的脑电信号中，筛选出在同一诱发刺激下的脑电信号，即在相同电极对刺激下的脑电信号。结构构建单元3针对筛选出的脑电信号，选取基线和脑电反应进行显著性检验，将显著性水平符合预设条件的部分挑选出来即可得到诱发刺激下的脑连接关系的邻接矩阵图，迎接矩阵图即为图结构信息。电极位点邻接矩阵图如说明书附图7所示，其中，纵轴表示诱发刺激，即给予刺激的电极对，横轴表示所有的反应电极，不同灰度表示不同的显著性水平。
显著性检验(significance test)就是事先对总体(随机变量)的参数或总体分布形式做出一个假设，然后利用样本信息来判断假设(备择假设)是否合理，即判断总体的真实情况与原假设是否有显著性差异。或者说，显著性检验要判断样本与假设之间的差异是纯属机会变异，还是假设与总体真实情况之间不一致所引起的。显著性检验是针对总体所做的假设做检验，其原理就是“小概率事件实际不可能性原理”来接受或否定假设。
发作区获取单元4：基于无监督聚类过程对图结构信息和图特征信息进行聚类，获取发作区电极位点，根据发作区电极位点和预设影像数据获取癫痫发作区。在无监督聚类过程使用自适应图卷积手段，结合图结构信号和图特征信息进行处理，相比于自监督学习和半监督学习，不需要用大量医生标注的样本进行预训练或监督学习。发作区获取单元4包括聚类模块41和定位模块42。
聚类模块41具体包括：将图结构信息通过拉普拉斯变换构建图滤波器，根据图特征信息构建图表示特征；通过图卷积操作进特征融合，获取融合特征，每次图卷积包括将图滤波器与图表示特征进行矩阵相乘，图卷积的次数根据簇内距离进行自适应判断；根据融合特征进行聚类得到多个簇，筛选得到发作区电极位点。
图结构信息由癫痫诱发数据获取，包含大量电极位点的相关信息。将每个电极位点看作图结构中的一个节点，每个节点有自己的一个特征，特征与图特征信息相关联，所有节点的特征信息共同构成完整的图特征信息。通过癫痫诱发数据得到的邻接矩阵反应出所有节点的连接关系。
本实施例提供的自适应无监督聚类是一种新型的聚类算法，适用于属性图，并利用图结构中的边连接关系作为一种结构特征，通过图卷积操作聚合每个节点及其k阶邻居特征来得到更好的特征表征。
给定属性图G(V，E，X)，其中V＝{V 1，V 2，…，V n}为节点集合，E为边的集合，构成邻接矩阵A＝{aij}∈R n*n，代表V i与V j之间有无向边的连接，图中所有节点构成的属性特征为X＝{X 1，X 2，…，X n} T，代表每个节点的属性。目标是将所有的节点聚类到m个簇里，簇的集合为C＝{C1，……，Cm}。
首先需要将邻接矩阵A构造成归一化的拉普拉斯矩阵L s＝D -1/2AD -1/2，继而构造图滤波器
Figure PCTCN2021094805-appb-000002
其中，L s表示拉普拉 斯矩阵，D表示度矩阵，A表示邻接矩阵，I表示单位矩阵，L s表示拉普拉斯矩阵。一次图卷积的过程即是一次矩阵相乘的过程:X^＝GX，k阶图卷积的过程即
Figure PCTCN2021094805-appb-000003
流程如说明书附图8所示。
在自适应图卷积模型中，将图结构信息以拉普拉斯变换的方式构建出图滤波器，根据图特征信息构建出图表示特征。图滤波器与图表示特征两个矩阵相乘视作一次图卷积的过程，进行自适应的图卷积操作揉合两种特征得到融合特征，其中卷积的次数是根据簇内距离来自适应判断。根据融合特征进行聚类得到多个簇，每个簇中包含多个节点。从多个簇中筛选出发作区簇，发作区簇中仅包括发作区电极位点，根据发作区电极位点获取对应的癫痫发作区。需要说明的是，由于聚类模型的不同，可能会产生多个簇，发作区簇也包括由多个子簇构成。由于模型为自适应图卷积，可自适应的判断卷积迭代次数，因而能够根据不同病人的病情进行针对性检测，具有极强的泛化性。
自适应图卷积的过程如说明书附图9所示，定义最大迭代次数max_iter，定义t作为循环次数，intra(C (t))即是所有簇的簇内距离之和，自适应卷积的过程是通过比较每次迭代中簇内距离大小，决定是否跳出循环，跳出循环时即认为当前结果最优，同时得到的k的值是图卷积的卷积次数。最终得到k阶图信号G kX，再通过谱聚类得到聚类结果。根据簇内距离变化进行迭代，每次迭代都进行一次卷积，获得节点k阶邻居的特征
在两类聚类模式下，聚类得到两个簇，分别为发作区簇和非发作区簇。根据融合特征进行聚类只能得到两个簇，避免了筛选的过程。其中，发作区簇仅包括发作区电极位点，非发作区簇仅包括非发作区电极位点，而发作区电极位点在影像中的位置就对应着癫痫发作区。由于每个簇包含多个节点，每个节点对应一个电极位点，因此，此处的发作区簇和非发作区簇都是电极位点集合，集合中包含多个电极位点，发作区簇包含一个或多个发作区电极位点，非发作区包含一个或多个非发作区电极位点。
在多类聚类模式下，聚类得到两个以上的簇。如将类别分为癫痫起始区、快速散播区、激惹区、非致痫区等多个区域，类别选择具体可以根据实际情况进行选择。
获取发作区电极位点之后，还需进行电极定位，定位模块42结合癫痫影像数据和发作区电极位点，获取发作区电极位点的位置，进而结合预设影像模板获取脑部的癫痫发作区。定位模块具体包括：获取癫痫病人的CT影像和MRI影像，对CT影像和MRI影像进行图像配准获取第一影像数据。根据第一影像数据判断发作区电极位点在影像中的电极位置；匹配电极位置和预设影像模板获取脑部的癫痫发作区。
癫痫发作区检测结果如说明书附图10所示。其中，两种深浅的圆形区域对应两种电极位点，根据电极的深浅判断是否处于发作区，较浅的电极对应非发作区的位置，较深的电极对应发作区的位置。经过实验验证，将系统聚类结果和医院给出的标注进行比较，得到了较好的结果，证明了本实施例提供的系统的适用性和可靠性。根据医生标注的电极脑区对应得到海马旁回后部、颞叶、颞极平面等脑区是发作区相关的脑区。根据系统得到的结果将电极和脑区进行对应，得到的最终结果则是海马旁回后部、颞叶、颞极平面、下额额回等多个区域，基本将真正手术要去除的发作区包括在内，证明了本系统的实用性与可用性。
本实施例提供了一种癫痫发作区定位系统，将癫痫诱发数据和癫痫自发数据进行有机结合，通过癫痫自发数据表征特征关系，癫痫诱发数据表征结构连接关系，综合特征关系和结构连接关系对癫痫发作区进行定位判断，综合性强、准确度高。通过特征提取和表征手段，将信号处理、深度学习、机器学习进行整合，能够发现数据中潜在的特征，进而获取更准确的检测效果。使用无监督聚类方法，相比于监督和半监督学习，无需耗时训练，也无需使用大量人工标注的样本进行预训练或监督学习，缩短检测时间，提高检测效率。由于模型为自适应图卷积，可自适应的判断卷积迭 代次数，因而能够根据不同病人的实际病情进行针对性检测，泛化能力强，检测准确度高。
实施例2
说明书附图11为本发明实施例2提供的一种计算机设备的结构示意图。说明书附图11显示的计算机设备12仅仅是一个示例，不应对本发明实施例的功能和使用范围带来任何限制。
如说明书附图11所示，计算机设备12以通用计算设备的形式表现。计算机设备12的组件可以包括但不限于：一个或者多个处理器或者处理单元16，系统存储器28，连接不同系统组件(包括系统存储器28和处理单元16)的总线18。计算机设备12典型地包括多种计算机系统可读介质。这些介质可以是任何能够被设备计算机12访问的可用介质，包括易失性和非易失性介质，可移动的和不可移动的介质。系统存储器28可以包括易失性存储器形式的计算机系统可读介质。
计算机设备12也可以与一个或多个外部设备14(例如键盘、指向设备、显示器等)通信，还可与一个或者多个使得用户能与该计算机设备12交互的设备通信，和/或与使得该计算机设备12能与一个或多个其它计算设备进行通信的任何设备通信。
处理单元16通过运行存储在系统存储器28中的程序，从而执行各种功能应用以及数据处理，例如实现一种癫痫发作区定位系统的控制方法，该方法包括：
101、数据获取单元1获取癫痫自发数据、癫痫诱发数据和癫痫影像数据；癫痫诱发数据包括处于诱发刺激下的脑电信号。
102、特征提取单元2通过对癫痫自发数据进行特征提取，获取图特征信息。
103、结构构建单元3在癫痫诱发数据中获取多个电极位点处于同一诱 发刺激下的脑电信号，通过显著性检验，构建图结构信息，图结构信息包括处于诱发刺激下的脑连接关系的邻接矩阵图。
104、发作区获取单元4基于无监督聚类过程对图结构信息和图特征信息进行聚类，获取发作区电极位点，根据发作区电极位点和癫痫影像数据获取癫痫发作区。
其中，步骤102和步骤103可同时进行，也可依次进行。
本实施例将一种癫痫发作区定位系统的控制方法应用到具体的计算机设备中，将该方法存储到存储器中，当执行器执行该存储器时，会运行该方法进行癫痫发作区定位，使用快捷方便，适用范围广。
当然，本领域技术人员可以理解，处理器还可以实现本发明任意实施例所提供的技术方案。
实施例3
本实施例3提供了一种计算机可读存储介质，其上存储有计算机程序，该程序被处理器执行时实现如一种癫痫发作区定位系统的控制方法，该方法包括：
101、数据获取单元1获取癫痫自发数据、癫痫诱发数据和癫痫影像数据；癫痫诱发数据包括处于诱发刺激下的脑电信号。
102、特征提取单元2通过对癫痫自发数据进行特征提取，获取图特征信息。
103、结构构建单元3在癫痫诱发数据中获取多个电极位点处于同一诱发刺激下的脑电信号，通过显著性检验，构建图结构信息，图结构信息包括处于诱发刺激下的脑连接关系的邻接矩阵图。
104、发作区获取单元4基于无监督聚类过程对图结构信息和图特征信息进行聚类，获取发作区电极位点，根据发作区电极位点和癫痫影像数据获取癫痫发作区。
其中，步骤102和步骤103可同时进行，也可依次进行。
本实施例的计算机存储介质，可以采用一个或多个计算机可读的介质的任意组合。计算机可读介质可以是计算机可读信号介质或者计算机可读存储介质。计算机可读存储介质例如可以是但不限于：电、磁、光、电磁、红外线、或半导体的系统、装置或器件，或者任意以上的组合。计算机可读存储介质的更具体的例子(非穷举的列表)包括：具有一个或多个导线的电连接、便携式计算机磁盘、硬盘、随机存取存储器(RAM)、只读存储器(ROM)、可擦式可编程只读存储器(EPROM或闪存)、光纤、便携式紧凑磁盘只读存储器(CD-ROM)、光存储器件、磁存储器件、或者上述的任意合适的组合。在本文件中，计算机可读存储介质可以是任何包含或存储程序的有形介质，该程序可以被指令执行系统、装置或者器件使用或者与其结合使用。
本实施例将一种癫痫发作区定位系统的控制方法应用到一种计算机可读存储介质，其上存储有计算机程序，该程序被处理器执行时实现本发明提供的癫痫发作区定位系统的控制方法的步骤，简便快捷，易于存储，不易丢失。
本发明针对现有技术，提供了一种癫痫发作区定位系统、设备及介质，将癫痫诱发数据、癫痫自发数据和癫痫影像数据有机结合，通过信号处理和机器学习的方法精准快速得到针对不同病人的癫痫发作区电极或脑区位置，帮助医生手术之前定位癫痫发作区位置，辅助医生做出判断。通过癫痫自发数据表征特征关系，癫痫诱发数据表征结构连接关系，综合特征关系和结构连接关系对癫痫发作区进行定位判断，相比于现有技术，考虑了全脑连接关系对癫痫的影响，综合性强、准确度高。通过特征提取和表征手段，将信号处理、深度学习、机器学习进行整合，能够发现数据中潜在的特征，进而获取更准确的检测效果。使用无监督聚类方法，相比于监督 和半监督学习，无需耗时训练，也无需使用大量人工标注的样本进行预训练或监督学习，缩短检测时间，提高检测效率。由于模型为自适应图卷积，可自适应的判断卷积迭代次数，因而能够根据不同病人的实际病情进行针对性检测，泛化能力强，检测准确度高。
本领域普通技术人员应该明白，上述的本发明的各模块或各步骤可以用通用的计算装置来实现，它们可以集中在单个计算装置上，或者分布在多个计算装置所组成的网络上，可选地，他们可以用计算机装置可执行的程序代码来实现，从而可以将它们存储在存储装置中由计算装置来执行，或者将它们分别制作成各个集成电路模块，或者将它们中的多个模块或步骤制作成单个集成电路模块来实现。这样，本发明不限制于任何特定的硬件和软件的结合。
注意，上述仅为本发明的较佳实施例及所运用技术原理。本领域技术人员会理解，本发明不限于这里的特定实施例，对本领域技术人员来说能够进行各种明显的变化、重新调整和替代而不会脱离本发明的保护范围。因此，虽然通过以上实施例对本发明进行了较为详细的说明，但是本发明不仅仅限于以上实施例，在不脱离本发明构思的情况下，还可以包括更多其他等效实施例，而本发明的范围由所附的权利要求范围决定。
以上公开的仅为本发明的几个具体实施场景，但是，本发明并非局限于此，任何本领域的技术人员能思之的变化都应落入本发明的保护范围。

Claims (14)

  1. 一种癫痫发作区定位系统，其特征在于，包括，
    数据获取单元：用于获取癫痫自发数据、癫痫诱发数据和癫痫影像数据，所述癫痫诱发数据包括电极位点处于诱发刺激下的脑电信号；
    特征提取单元：用于对所述癫痫自发数据进行特征提取，获取图特征信息；
    结构构建单元：用于在所述癫痫诱发数据中获取多个电极位点处于同一诱发刺激下的脑电信号，通过显著性检验构建图结构信息，所述图结构信息包括处于诱发刺激下的脑连接关系的邻接矩阵图；
    发作区获取单元：用于基于无监督聚类过程对所述图结构信息和所述图特征信息进行聚类，获取发作区电极位点，根据所述发作区电极位点和所述癫痫影像数据获取癫痫发作区。
  2. 根据权利要求1所述的癫痫发作区定位系统，其特征在于，所述特征提取单元具体包括：
    预处理模块：用于在所述癫痫自发数据中获取自发脑电信号，对所述自发脑电信号进行预处理；
    特征提取模块：用于预处理后的自发脑电信号进行特征提取，获取时频特征；
    信息获取模块：用于基于深度学习网络处理所述时频特征得到降维后的时频特征，将降维后的时频特征作为所述图特征信息。
  3. 根据权利要求1或2所述的癫痫发作区定位系统，其特征在于，所述发作区获取单元设置有聚类模块；
    聚类模块：用于根据所述图结构信息构建图滤波器，根据所述图特征信息构建图表示特征；
    通过图卷积操作进特征融合，获取融合特征，每次所述图卷积包括将 所述图滤波器与所述图表示特征进行矩阵相乘，所述图卷积的次数根据簇内距离进行自适应判断；
    对所述融合特征进行聚类得到多个簇，筛选得到发作区簇，所述发作区簇仅包括所述发作区电极位点。
  4. 根据权利要求3所述的癫痫发作区定位系统，其特征在于，所述癫痫影像数据包括CT影像和MRI影像；
    所述发作区获取单元还设置有定位模块；
    定位模块：用于对所述CT影像和所述MRI影像进行图像配准获取第一影像数据；根据所述第一影像数据判断所述发作区电极位点在影像中的电极位置；匹配所述电极位置和预设影像模板获取脑部的癫痫发作区。
  5. 根据权利要求3所述的癫痫发作区定位系统，其特征在于，对所述融合特征进行两类聚类，得到发作区簇和非发作区簇，所述发作区簇仅包括所述发作区电极位点，所述非发作区簇仅包括非发作区电极位点；
    选取电极位点较少的簇作为所述发作区簇。
  6. 根据权利要求1所述的癫痫发作区定位系统，其特征在于，所述自发数据包括癫痫病人在癫痫发作时、癫痫发作期间的自发脑电信号；
    所述诱发数据包括在刺激一对电极位点时，颅内所有电极位点在该刺激下的脑电信号。
  7. 根据权利要求2所述的癫痫发作区定位系统，其特征在于，所述预处理包括去除伪差、去除工频干扰、滤波处理、降采样处理、去势处理；
  8. 根据权利要求7所述的癫痫发作区定位系统，其特征在于，所述特征提取模块通过傅里叶变换或小波变换对预处理后的自发脑电信号进行特征提取，获取时频特征。
  9. 根据权利要求3所述的癫痫发作区定位系统，其特征在于，所述图滤波器的构建过程包括：
    从所述图结构信息中的获取邻接矩阵；
    对所述邻接矩阵进行拉普拉斯变换，构造出归一化的拉普拉斯矩阵；
    根据所述拉普拉斯矩阵构造图滤波器。
  10. 根据权利要求9所述的癫痫发作区定位系统，其特征在于，所述拉普拉斯矩阵的表达式为
    L s＝D -1/2AD -1/2
    其中，L s表示拉普拉斯矩阵，D表示度矩阵，A表示邻接矩阵；
    所述图滤波器的表达式为：
    Figure PCTCN2021094805-appb-100001
    其中，G表示图滤波器，I表示单位矩阵，L s表示拉普拉斯矩阵。
  11. 根据权利要求2所述的癫痫发作区定位系统，其特征在于，所述深度学习网络包括自动编码器模型，所述信息获取模块基于所述自动编码器模型处理所述时频特征。
  12. 根据权利要求11所述的癫痫发作区定位系统，其特征在于，所述图特征信息的获取流程具体包括：
    根据所述时频特征构建出特征向量；
    通过所述自动编码器模型中的编码部分进行降维和特征学习，得到一个中间状态的低维隐藏特征表示；
    通过所述自动编码器模型中的解码部分将所述低维隐藏特征表示转换为所述特征向量；
    基于损失函数调节网络参数；
    迭代上述步骤，直至满足迭代停止条件，将满足迭代停止条件的低维特征表示作为图特征信息。
  13. 一种计算机设备，其特征在于，所述计算机设备包括：
    一个或多个处理器；
    存储器，用于存储一个或多个程序；
    当所述一个或多个程序被所述一个或多个处理器执行，使得所述一个或多个处理器实现如下处理：
    数据获取单元获取癫痫自发数据、癫痫诱发数据和癫痫影像数据；
    特征提取单元对所述癫痫自发数据进行特征提取，获取图特征信息；
    结构构建单元在所述癫痫诱发数据中获取多个电极位点处于同一诱发刺激下的脑电信号，通过显著性检验构建图结构信息，所述图结构信息包括处于诱发刺激下的脑连接关系的邻接矩阵图；
    发作区获取单元基于无监督聚类过程对所述图结构信息和所述图特征信息进行聚类，获取发作区电极位点，根据所述发作区电极位点和所述癫痫影像数据获取癫痫发作区。
  14. 一种计算机可读存储介质，其上存储有计算机程序，其特征在于，该程序被处理器执行时实现如下处理：
    数据获取单元获取癫痫自发数据、癫痫诱发数据和癫痫影像数据；
    特征提取单元对所述癫痫自发数据进行特征提取，获取图特征信息；
    结构构建单元在所述癫痫诱发数据中获取多个电极位点处于同一诱发刺激下的脑电信号，通过显著性检验构建图结构信息，所述图结构信息包括处于诱发刺激下的脑连接关系的邻接矩阵图；
    发作区获取单元基于无监督聚类过程对所述图结构信息和所述图特征信息进行聚类，获取发作区电极位点，根据所述发作区电极位点和所述癫痫影像数据获取癫痫发作区。
PCT/CN2021/094805 2021-05-20 2021-05-20 一种癫痫发作区定位系统、设备及介质 WO2022241710A1 (zh)

Priority Applications (1)

Application Number Priority Date Filing Date Title
PCT/CN2021/094805 WO2022241710A1 (zh) 2021-05-20 2021-05-20 一种癫痫发作区定位系统、设备及介质

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
PCT/CN2021/094805 WO2022241710A1 (zh) 2021-05-20 2021-05-20 一种癫痫发作区定位系统、设备及介质

Publications (1)

Publication Number Publication Date
WO2022241710A1 true WO2022241710A1 (zh) 2022-11-24

Family

ID=84140109

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2021/094805 WO2022241710A1 (zh) 2021-05-20 2021-05-20 一种癫痫发作区定位系统、设备及介质

Country Status (1)

Country Link
WO (1) WO2022241710A1 (zh)

Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20080027346A1 (en) * 2006-05-22 2008-01-31 The Trustees Of The University Of Pennsylvania Method and device for the recording, localization and stimulation-based mapping of epileptic seizures and brain function utilizing the intracranial and extracranial cerebral vasculature and/or central and/or peripheral nervous system
WO2016200952A1 (en) * 2015-06-09 2016-12-15 Children's Medical Center Corporation Method and system for locating seizure focus from interictal data
CN108113669A (zh) * 2017-12-06 2018-06-05 中南大学 一种癫痫病灶定位方法及系统
CN108852350A (zh) * 2018-05-18 2018-11-23 中山大学 一种基于深度学习算法的头皮脑电图致痫区的识别与定位方法
US20200178832A1 (en) * 2017-05-25 2020-06-11 Brent M. Berry Seizure onset zone localization
CN111951228A (zh) * 2020-07-22 2020-11-17 电子科技大学 一种融合梯度激活映射和深度学习模型的癫痫灶定位系统
CN112348785A (zh) * 2020-10-29 2021-02-09 深圳大学 一种癫痫病灶定位方法及系统

Patent Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20080027346A1 (en) * 2006-05-22 2008-01-31 The Trustees Of The University Of Pennsylvania Method and device for the recording, localization and stimulation-based mapping of epileptic seizures and brain function utilizing the intracranial and extracranial cerebral vasculature and/or central and/or peripheral nervous system
WO2016200952A1 (en) * 2015-06-09 2016-12-15 Children's Medical Center Corporation Method and system for locating seizure focus from interictal data
US20200178832A1 (en) * 2017-05-25 2020-06-11 Brent M. Berry Seizure onset zone localization
CN108113669A (zh) * 2017-12-06 2018-06-05 中南大学 一种癫痫病灶定位方法及系统
CN108852350A (zh) * 2018-05-18 2018-11-23 中山大学 一种基于深度学习算法的头皮脑电图致痫区的识别与定位方法
CN111951228A (zh) * 2020-07-22 2020-11-17 电子科技大学 一种融合梯度激活映射和深度学习模型的癫痫灶定位系统
CN112348785A (zh) * 2020-10-29 2021-02-09 深圳大学 一种癫痫病灶定位方法及系统

Similar Documents

Publication Publication Date Title
Guo et al. Inter-patient ECG classification with convolutional and recurrent neural networks
Covert et al. Temporal graph convolutional networks for automatic seizure detection
Sadiq et al. Toward the development of versatile brain–computer interfaces
Li et al. Multi-modal bioelectrical signal fusion analysis based on different acquisition devices and scene settings: Overview, challenges, and novel orientation
Dissanayake et al. Geometric deep learning for subject independent epileptic seizure prediction using scalp EEG signals
WO2020103683A1 (zh) 基于脑功能图谱的猴-人跨物种迁移进行精神疾病的个体化预测方法和系统
US11333730B2 (en) Systems and methods for mapping neuronal circuitry and clinical applications thereof
CN110392549B (zh) 确定引起期望行为的大脑刺激的系统、方法和介质
US10420937B2 (en) Method and apparatus to determine optimal brain stimulation to induce desired behavior
CN113274037B (zh) 一种动态脑功能网络的生成方法、系统及设备
Miller et al. Basis profile curve identification to understand electrical stimulation effects in human brain networks
Liu et al. Network community structure detection for directional neural networks inferred from multichannel multisubject EEG data
WO2024098449A1 (zh) 基于深度学习和电生理信号的癫痫病灶区定位系统及方法
CN114129147A (zh) 基于脑功能网络的帕金森患者dbs术后效果预测系统及方法
CN115363523A (zh) 一种癫痫发作区定位系统、设备及介质
CN116831594A (zh) 一种基于迭代图卷积神经网络的癫痫脑电分类方法
Hosseini Brain-computer interface for analyzing epileptic big data
CN107256408B (zh) 一种大脑功能网络的关键路径搜索方法
Liu et al. Automated Machine Learning for Epileptic Seizure Detection Based on EEG Signals.
O'Driscoll et al. SOM and MCODE methods of defining functional clusters in MRI of the brain
WO2022241710A1 (zh) 一种癫痫发作区定位系统、设备及介质
Nelson et al. Deep-learning-based intelligent neonatal seizure identification using spatial and spectral GNN optimized with the Aquila algorithm
Antonakakis et al. Comparison of brain network models using cross-frequency coupling and attack strategies
Duan et al. Classification of epilepsy period based on combination feature extraction methods and spiking swarm intelligent optimization algorithm
Paillard et al. GREEN: a lightweight architecture using learnable wavelets and Riemannian geometry for biomarker exploration

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 21940166

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 21940166

Country of ref document: EP

Kind code of ref document: A1

122 Ep: pct application non-entry in european phase

Ref document number: 21940166

Country of ref document: EP

Kind code of ref document: A1