Method, device, program product and electronic device for detecting human brain discharge channels

Abstract

The present disclosure provides a method, device, program product and electronic device for detecting discharge channels of the human brain, and relates to the field of computer technology. The method comprises: obtaining EEG data of a subject; extracting a first feature from the EEG data; obtaining discharge channel codes corresponding to a plurality of preset discharge channels; the plurality of preset discharge channels include the discharge channels corresponding to the EEG data; inputting the first feature and the discharge channel code into a discharge channel detection model to determine the abnormal discharge channels of the subject. The present disclosure realizes the automated detection of abnormal discharge channels of the human brain.

Description

Technical Field

[0001] The embodiments of the present disclosure relate to the field of computer technology. More specifically, the embodiments of the present disclosure relate to a method for detecting a discharge channel in a human brain, a device for detecting a discharge channel in a human brain, a computer program product and an electronic device. Background Art

[0002] This section is intended to provide a background or context to the embodiments of the disclosure that are recited in the claims, and no description herein is admitted to be prior art by inclusion in this section.

[0003] By collecting EEG data and identifying the discharge phenomena in it, it is possible to assess the health of the human body. In particular, if abnormal discharges are detected in the human brain, it indicates that the person may have a specific disease. For example, detecting epileptic-like discharges in an EEG is one of the important criteria for diagnosing epilepsy.

[0004] Faced with the increasing demand for relevant data analysis and detection, the manual method of analyzing data and identifying abnormal discharges is inefficient, and the manual identification results usually have low consistency and are difficult to guarantee reliability. Therefore, the industry has an increasing demand for automated analysis of EEG data.

[0005] In related technologies, artificial intelligence technology is used to detect human brain discharges. For example, by training a neural network for epileptiform discharge recognition, it can automatically process electroencephalogram and other data and output epileptiform discharge detection results. Summary of the invention

[0006] However, the current AI-based brain discharge detection method is difficult to identify abnormal discharge channels. Even if abnormal discharges are detected in the human brain, it is difficult for doctors to make detailed and targeted diagnosis and treatment based on the test results because the abnormal discharge location cannot be accurately located.

[0007] In view of the above problems, embodiments of the present disclosure provide a method for detecting discharge channels in the human brain, a device for detecting discharge channels in the human brain, a computer program product, and an electronic device.

[0008] According to a first aspect of the present disclosure, a method for detecting discharge channels in the human brain is provided, the method comprising: obtaining EEG data of a subject; extracting a first feature from the EEG data; obtaining discharge channel codes corresponding to a plurality of preset discharge channels; the plurality of preset discharge channels including the discharge channels corresponding to the EEG data; and inputting the first feature and the discharge channel code into a discharge channel detection model to determine abnormal discharge channels of the subject.

[0009] In one embodiment, extracting the first feature from the EEG data includes: inputting the EEG data into a feature extraction model; the feature extraction model includes a basic feature layer and multiple parallel deep feature layers, and different depth feature layers have feature extraction units of different scales; processing the EEG data through the basic feature layer to obtain basic features; processing the basic features through multiple deep feature layers respectively to obtain multiple depth features corresponding to the multiple depth feature layers; and obtaining the first feature based on the multiple depth features.

[0010] In one embodiment, the feature extraction unit includes a convolution layer, different deep feature layers have different numbers of convolution layers, and the convolution layers in each deep feature layer are arranged in sequence from small to large in convolution scale; the basic features are processed by multiple deep feature layers respectively to obtain multiple deep features corresponding to the multiple deep feature layers, including: inputting the basic features into each deep feature layer respectively, and processing each convolution layer in each deep feature layer in turn to obtain the depth features corresponding to each deep feature layer.

[0011] In one embodiment, the basic feature layer includes a self-attention layer; the EEG data is processed by the basic feature layer to obtain basic features, including: inputting the EEG data into the self-attention layer, and using the EEG data as query information, key information and value information of the self-attention layer respectively, and obtaining the basic features through processing.

[0012] In one embodiment, the discharge channel detection model includes a feature mapping network, a feature fusion network, and a classification network; the inputting the first feature and the discharge channel code into the discharge channel detection model to determine the abnormal discharge channel of the object under test includes: mapping the first feature and the discharge channel code to a feature space through the feature mapping network to obtain a second feature corresponding to the first feature and a third feature corresponding to the discharge channel code; fusing the second feature and the third feature through the feature fusion network to obtain a fourth feature; and processing the fourth feature through the classification network to determine the abnormal discharge channel of the object under test.

[0013] In one embodiment, the feature fusion network includes a cross-attention network; the fusing of the second feature and the third feature through the feature fusion network to obtain the fourth feature includes: inputting the second feature and the third feature into the cross-attention network, and using the second feature as the query information of the cross-attention network, and the third feature as the key information and value information of the cross-attention network, and obtaining the fourth feature through processing.

[0014] In one embodiment, the feature mapping network includes a first mapping unit and a second mapping unit; mapping the first feature and the discharge channel code to the feature space through the feature mapping network to obtain the second feature corresponding to the first feature and the third feature corresponding to the discharge channel code includes: inputting the first feature into the trained first mapping unit to obtain the second feature; inputting the discharge channel code into the trained second mapping unit to obtain the third feature.

[0015] In one embodiment, the method further includes: obtaining reference data corresponding to the multiple preset discharge channels; extracting a first sample feature from the reference data, inputting the first sample feature into a first mapping unit to be trained, and obtaining a second sample feature; inputting the discharge channel code into a second mapping unit to be trained, and obtaining a third sample feature; matching the second sample feature and the third sample feature, and updating the first mapping unit and the second mapping unit according to the matching result and the correspondence between the reference data and the discharge channel code, so as to obtain a trained first mapping unit and a trained second mapping unit.

[0016] In one embodiment, the reference data corresponding to the multiple preset discharge channels include reference data corresponding to a single preset discharge channel, and the reference data corresponding to the single preset discharge channel includes: EEG data in which abnormal discharge occurs in the single preset discharge channel and no abnormal discharge occurs in other preset discharge channels.

[0017] In one embodiment, the obtaining of discharge channel codes corresponding to the plurality of preset discharge channels comprises at least one of the following steps: encoding a first channel label indicating that abnormal discharge occurs in a single preset discharge channel to obtain the discharge channel code; encoding a second channel label indicating that abnormal discharge occurs in at least two preset discharge channels to obtain the discharge channel code.

[0018] According to a second aspect of the present disclosure, a human brain discharge channel detection device is provided, the device comprising: an EEG data acquisition module, configured to acquire EEG data of a subject; a first feature extraction module, configured to extract a first feature from the EEG data; a discharge channel code acquisition module, configured to acquire discharge channel codes corresponding to a plurality of preset discharge channels; the plurality of preset discharge channels include the discharge channels corresponding to the EEG data; and a model processing module, configured to input the first feature and the discharge channel code into a discharge channel detection model to determine the abnormal discharge channel of the subject.

[0019] In one embodiment, extracting the first feature from the EEG data includes: inputting the EEG data into a feature extraction model; the feature extraction model includes a basic feature layer and multiple parallel deep feature layers, and different depth feature layers have feature extraction units of different scales; processing the EEG data through the basic feature layer to obtain basic features; processing the basic features through multiple deep feature layers respectively to obtain multiple depth features corresponding to the multiple depth feature layers; and obtaining the first feature based on the multiple depth features.

[0020] In one embodiment, the feature extraction unit includes a convolution layer, different deep feature layers have different numbers of convolution layers, and the convolution layers in each deep feature layer are arranged in sequence from small to large in convolution scale; the basic features are processed by multiple deep feature layers respectively to obtain multiple deep features corresponding to the multiple deep feature layers, including: inputting the basic features into each deep feature layer respectively, and processing each convolution layer in each deep feature layer in turn to obtain the depth features corresponding to each deep feature layer.

[0021] In one embodiment, the basic feature layer includes a self-attention layer; the EEG data is processed by the basic feature layer to obtain basic features, including: inputting the EEG data into the self-attention layer, and using the EEG data as query information, key information and value information of the self-attention layer respectively, and obtaining the basic features through processing.

[0022] In one embodiment, the discharge channel detection model includes a feature mapping network, a feature fusion network, and a classification network; the inputting the first feature and the discharge channel code into the discharge channel detection model to determine the abnormal discharge channel of the object under test includes: mapping the first feature and the discharge channel code to a feature space through the feature mapping network to obtain a second feature corresponding to the first feature and a third feature corresponding to the discharge channel code; fusing the second feature and the third feature through the feature fusion network to obtain a fourth feature; and processing the fourth feature through the classification network to determine the abnormal discharge channel of the object under test.

[0023] In one embodiment, the feature fusion network includes a cross-attention network; the fusing of the second feature and the third feature through the feature fusion network to obtain the fourth feature includes: inputting the second feature and the third feature into the cross-attention network, and using the second feature as the query information of the cross-attention network, and the third feature as the key information and value information of the cross-attention network, and obtaining the fourth feature through processing.

[0024] In one embodiment, the feature mapping network includes a first mapping unit and a second mapping unit; mapping the first feature and the discharge channel code to the feature space through the feature mapping network to obtain the second feature corresponding to the first feature and the third feature corresponding to the discharge channel code includes: inputting the first feature into the trained first mapping unit to obtain the second feature; inputting the discharge channel code into the trained second mapping unit to obtain the third feature.

[0025] In one embodiment, the device also includes a model training module, which is configured to: obtain reference data corresponding to the multiple preset discharge channels; extract a first sample feature from the reference data, input the first sample feature into a first mapping unit to be trained, and obtain a second sample feature; input the discharge channel code into a second mapping unit to be trained, and obtain a third sample feature; match the second sample feature and the third sample feature, and update the first mapping unit and the second mapping unit according to the matching result and the correspondence between the reference data and the discharge channel code to obtain a trained first mapping unit and a trained second mapping unit.

[0026] In one embodiment, the reference data corresponding to the multiple preset discharge channels include reference data corresponding to a single preset discharge channel, and the reference data corresponding to the single preset discharge channel includes: EEG data in which abnormal discharge occurs in the single preset discharge channel and no abnormal discharge occurs in other preset discharge channels.

[0027] In one embodiment, the obtaining of discharge channel codes corresponding to the plurality of preset discharge channels comprises at least one of the following steps: encoding a first channel label indicating that abnormal discharge occurs in a single preset discharge channel to obtain the discharge channel code; encoding a second channel label indicating that abnormal discharge occurs in at least two preset discharge channels to obtain the discharge channel code.

[0028] According to a third aspect of the present disclosure, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the method of the first aspect and possible implementation methods thereof are implemented.

[0029] According to a fourth aspect of the present disclosure, an electronic device is provided, comprising: a processor; and a memory for storing executable instructions of the processor; wherein the processor is configured to execute the method of the above-mentioned first aspect and its possible implementation methods by executing the executable instructions.

[0030] The embodiments of the present disclosure have the following technical effects:

[0031] The EEG data of the subject is obtained, the first feature is extracted from the EEG data, the discharge channel codes corresponding to multiple preset discharge channels are obtained, and the first feature and the discharge channel code are input into the discharge channel detection model to determine the abnormal discharge channel of the subject. The automatic detection of abnormal discharge channels of the human brain is realized. The discharge channel code can guide the processing process of the discharge channel detection model, improve the model's ability to distinguish abnormal discharge channels, and improve the accuracy of the test results. The test results provide comprehensive and accurate auxiliary diagnostic information, which can help doctors locate the abnormal discharge position of the subject, and help doctors to make detailed and targeted diagnosis and treatment according to the test results. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] Figure 1A A schematic diagram showing a system architecture in an embodiment of the present disclosure.

[0033] Figure 1B A schematic diagram showing another system architecture in an embodiment of the present disclosure.

[0034] Figure 2 A flow chart of a method for detecting discharge channels in the human brain in an embodiment of the present disclosure is shown.

[0035] Figure 3 A schematic diagram of an electroencephalogram in an embodiment of the present disclosure is shown.

[0036] Figure 4 A flowchart for extracting a first feature in an embodiment of the present disclosure is shown.

[0037] Figure 5 A schematic diagram of extracting a first feature in an embodiment of the present disclosure is shown.

[0038] Figure 6 A flow chart for determining an abnormal discharge channel in an embodiment of the present disclosure is shown.

[0039] Figure 7 A flowchart of training a first mapping unit and a second mapping unit in an embodiment of the present disclosure is shown.

[0040] Figure 8 A schematic diagram of processing training sample data in an embodiment of the present disclosure is shown.

[0041] Fig.9A A schematic diagram of marking an abnormal discharge channel in an embodiment of the present disclosure is shown.

[0042] Fig. 9B Another schematic diagram of marking abnormal discharge channels in an embodiment of the present disclosure is shown.

[0043] Fig.10A schematic structural diagram of a human brain discharge channel detection device in an embodiment of the present disclosure is shown.

[0044] Fig.11 A schematic structural diagram of an electronic device in an embodiment of the present disclosure is shown.

[0045] In the drawings, the same or corresponding reference numerals represent the same or corresponding parts. DETAILED DESCRIPTION

[0046] The following is a detailed description of the principles and spirit of the present disclosure with reference to several representative embodiments of the present disclosure. It should be understood that these embodiments are provided only to enable those skilled in the art to better understand and implement the present disclosure, and are not intended to limit the scope of the present disclosure in any way. On the contrary, these embodiments are provided to make the present disclosure more thorough and complete, and to fully convey the scope of the present disclosure to those skilled in the art.

[0047] The embodiments of the present disclosure may be implemented as a system, device, apparatus, method or computer program product. Therefore, the present disclosure may be specifically implemented in the following forms, namely: complete hardware, complete software (including firmware, resident software, microcode, etc.), or a combination of hardware and software. SUMMARY OF THE INVENTION

[0049] In the related technology, the human brain discharge detection method based on artificial intelligence is difficult to identify abnormal discharge channels. Even if abnormal discharge is detected in the human brain, it is difficult for doctors to make detailed and targeted diagnosis and treatment based on the test results because the abnormal discharge position cannot be accurately located.

[0050] In view of the above, the present disclosure provides a method, device, program product and electronic device for detecting discharge channels in the human brain. Obtain the EEG data of the subject, extract the first feature from the EEG data, obtain the discharge channel codes corresponding to multiple preset discharge channels, input the first feature and the discharge channel codes into the discharge channel detection model to determine the abnormal discharge channels of the subject. Automated detection of abnormal discharge channels in the human brain is achieved. The discharge channel codes can guide the processing process of the discharge channel detection model, improve the model's ability to distinguish abnormal discharge channels, and improve the accuracy of the detection results. The detection results provide comprehensive and accurate auxiliary diagnostic information, which can help doctors locate the abnormal discharge position of the subject, and facilitate doctors to perform detailed and targeted diagnosis and treatment based on the detection results.

[0051] After introducing the basic principles of the present disclosure, various non-limiting embodiments of the present disclosure are described in detail below.

[0052] Application Scenario Overview

[0053] It should be noted that the following application scenarios are only shown to facilitate understanding of the spirit and principles of the present disclosure, and the embodiments of the present disclosure are not limited in this regard and can be applied to any applicable scenarios.

[0054] The embodiments of the present disclosure can be applied to the detection of suspected patients and related scenarios of auxiliary diagnosis and treatment. For example, if a patient is suspected of having epilepsy, the patient's EEG data can be detected, and the abnormal discharge channels of the patient can be detected by the human brain discharge channel detection method disclosed in the present disclosure. The detection results can assist doctors in diagnosing and treating patients. The application scenarios are described in detail below in combination with the system architecture.

[0055] Figure 1A A schematic diagram of a system architecture for detecting discharge channels in the human brain is shown. The system architecture includes a test object 101, a biomedical monitoring device 102, and a processing device 103. When it is necessary to perform human brain discharge channel detection on the test object 101, the biomedical monitoring device 102 can be set for the test object 101 to collect electrical signal data, and the data is processed by the processing device 103.

[0056] The biomedical monitoring device 102 may include detection electrodes 1021, 1022, 1023 and a main device 1024. The detection electrodes 1021, 1022, 1023 may contact different parts of the subject 101 to collect electrical signals. The main device 1024 aggregates and processes the electrical signals to obtain original or pre-processed EEG data. Exemplarily, the biomedical monitoring device 102 may be an EEG monitoring device (such as an EEG machine). The detection electrodes 1021, 1022, 1023 may be fixed at different positions on the head of the subject 101 to collect multi-channel EEG signals. Of course, Figure 1A The biomedical monitoring device 102 shown is merely exemplary and may also include a helmet, a hat, bedding, packaged portable detection electrodes, and other components, which are not limited in the present disclosure.

[0057] The biomedical monitoring device 102 can be connected to the processing device 103 in communication, such as through a wired or wireless communication link, so as to send the collected data to the processing device 103. The processing device 103 can execute the human brain discharge channel detection method in this exemplary embodiment, process the acquired EEG data, and obtain the discharge channel detection result of the subject 101. In one embodiment, the processing device 103 can include a display, which can display the detection result of the discharge channel and the EEG data.

[0058] In one embodiment, the biomedical monitoring device 102 and the processing device 103 may be integrated into the same device. Exemplarily, the main device 1024 of the biomedical monitoring device 102 may serve as the processing device 103 .

[0059] Figure 1B A schematic diagram of another system architecture for detecting discharge channels in the human brain is shown. The system architecture includes a subject 101, a biomedical monitoring device 102, a data transceiver device 104, and a server 105. The biomedical monitoring device 102 can be connected to the data transceiver device 104 for communication, and the data transceiver device 104 is connected to the server 105 for communication. Thus, after the data transceiver device 104 obtains the original or pre-processed EEG data, it sends it to the server 105. The server 105 may include any form of data processing server such as a cloud server and a distributed server. After obtaining the data sent by the data transceiver device 104, the server 105 obtains the discharge channel detection result of the subject 101 by executing the human brain discharge channel detection method in the embodiment of the present disclosure. In one embodiment, the server 105 can return the discharge channel detection result to the data transceiver device 104 to display it on the data transceiver device 104 or the biomedical monitoring device 102.

[0060] Figure 1B The system architecture shown is suitable for portable scenarios. For example, the subject 101 can use the portable biomedical monitoring device 102 to collect data at home, office, etc., and the data transceiver device 104 sends the data to the server 105 for human brain discharge channel detection. In this way, the user can understand his own abnormal discharge without leaving home. When encountering problems such as suspected epilepsy, the obtained test results can be sent to the doctor to help the doctor judge the condition.

[0061] Exemplary Methods

[0062] The following is an exemplary description of the method for detecting discharge channels in the human brain in an embodiment of the present disclosure. Figure 2 The flow chart of the method for detecting the discharge channel of the human brain is shown, which may include steps S210 to S230. Figure 2 Provide detailed instructions for each step.

[0063] refer to Figure 2 In step S210, the EEG data of the subject is obtained.

[0064] Among them, the test object refers to a person who needs to be tested for abnormal discharge channels, such as a possible epilepsy patient. Electrical signal data is data related to electrical activity detected from the human body by medical detection methods, which can reflect the physiological state or physical signs of the person. EEG data can be an electroencephalogram collected using a multi-channel electroencephalogram machine. The electroencephalogram (EEG) is a graph obtained by amplifying and recording the spontaneous biopotential of the cerebral cortex from the scalp through precision instruments. It is the spontaneous and rhythmic electrical activity of a group of brain cells recorded by electrodes. This electrical activity is a plane diagram of the relationship between the potential and time, with the potential as the vertical axis and time as the horizontal axis. The frequency (period), amplitude and phase of brain waves constitute the basic characteristics of the electroencephalogram. Of course, EEG data can also be other forms of data. For example, the above-mentioned electroencephalogram is usually data in the time domain, and EEG data in the frequency domain and the joint time-frequency domain can also be obtained.

[0065] In one embodiment, in addition to EEG data, other biomedical monitoring data of the subject can also be obtained to provide more comprehensive information and help achieve more accurate detection of human brain discharge channels. Other biomedical monitoring data include but are not limited to:

[0066] ECG data can be an electrocardiogram collected by a multi-lead electrocardiogram machine. Electrocardiogram (ECG) refers to the graph of various forms of potential changes drawn from the body surface by an electrocardiograph during each cardiac cycle, with the pacemaker, atrium, and ventricle being excited successively, accompanied by changes in the electrocardiogram bioelectricity. The electrocardiogram is an objective indicator of the occurrence, propagation, and recovery process of cardiac excitement.

[0067] Myoelectric data (such as electromyography (EMG)) is the temporal and spatial superposition of action potentials of motor units in many muscle fibers, which can be obtained by sticking electromyographic sensors on the skin.

[0068] Electrooculogram (EOG) data, such as electrooculogram (EOG), is a bioelectric signal caused by the potential difference between the cornea and retina of the eye. It is very easy to collect and can be completed with a small number of electrodes.

[0069] Gastrointestinal data (such as electrogastrogram (EGG)) are electrical signals generated by the contraction of stomach muscles and can be collected on the surface of the abdominal skin using electrodes.

[0070] Non-electrical signal data, such as blood pressure, blood oxygen saturation, etc.

[0071] In one embodiment, the above-mentioned acquisition of EEG data of the subject may include the following steps:

[0072] Preprocessing the original detection data of the subject to obtain EEG data;

[0073] Preprocessing includes one or more of the following: resampling, filtering, data enhancement, noise data removal, numerical standardization, and time-frequency transformation.

[0074] The original detection data is the original data collected by the biomedical monitoring equipment. For example, the biomedical monitoring equipment includes an electroencephalograph, and the original detection data includes the original EEG data or electroencephalogram collected by the electroencephalogram. The electroencephalogram is essentially a graph drawn by the EEG signals at different times, and the EEG signal can be considered to be equivalent to the electroencephalogram. Figure 3 A schematic diagram of multi-channel EEG data is shown. The original EEG data is plotted as a graph to obtain Figure 3 The waveform shown is the electroencephalogram.

[0075] To facilitate subsequent processing, the original detection data may be preprocessed. The specific method of preprocessing is exemplified below.

[0076] ① Resampling. Resampling can change the sampling rate of the signal and can convert non-uniformly sampled signals into uniformly sampled signals. For example, the EEG data can be resampled at 500 Hz.

[0077] ② Filtering. For example, filtering can be performed by using notch filtering or the like. For example, all discharge channels of the EEG data are subjected to 50 Hz notch filtering to reduce interference from AC signals, and bandpass filtering (such as using a 0.1-70 Hz frequency band) is performed to reduce signal interference in non-EEG signal frequency bands.

[0078] ③ Data enhancement. For example, the original lead, average lead, left hemisphere lead, and right hemisphere lead data of the EEG can be obtained, and all the lead data are spliced ​​to obtain enhanced EEG data.

[0079] ④ Eliminate noise data. Noise data may be caused by factors such as poor contact (such as poor contact between the biomedical monitoring equipment and the parts of the object being tested, poor contact between the equipment cables, etc.). Eliminating noise data can improve data quality and the accuracy of test results. For example, data in channels with more noise can be eliminated.

[0080] ⑤ Numerical standardization processing. Numerical standardization processing can map data of different modes and types into the same appropriate numerical range for unified processing. For example, the original detection data can be dimensioned first and then normalized into an appropriate numerical range, such as by multiplying the corresponding coefficients for numerical mapping to complete the numerical standardization processing.

[0081] ⑥ Time-frequency transformation. The original EEG data is usually data in the time domain, expressing the changes of electrical signals over time, and the information is relatively simple. Through time-frequency transformation, the original detection data can be converted to the frequency domain or the joint time-frequency domain to obtain frequency domain data or time-frequency data, which can provide the distribution information of the signal in the frequency domain, or the joint distribution information in the time domain and the frequency domain. The time-frequency transformation methods include but are not limited to short-time Fourier transform, wavelet transform, etc. The present disclosure does not limit which specific method is adopted.

[0082] According to actual needs and the characteristics of the original detection data, any one or more of the above preprocessing methods can be used, and different preprocessing methods can be used for different original detection data. For example, the original EEG data can be processed by filtering, data enhancement, and time-frequency transformation in sequence to obtain preprocessed EEG data.

[0083] Continue to refer Figure 2 , in step S220, a first feature is extracted from the EEG data.

[0084] The first feature is an abstract feature learned by the computer from the EEG data. For example, the EEG data can be encoded to obtain the first feature. In one embodiment, the first feature may include an electrical signal feature and a sleep feature. The electrical signal feature is a feature extracted based on the distribution law of the electrical signal data, and the sleep feature is a feature related to the sleep information of the subject. Exemplarily, the distribution of the EEG data can be statistically analyzed, such as sampling or segmenting the EEG data according to a certain window, and the maximum value, minimum value, average value, median, standard deviation and other indicator values ​​in each window are statistically analyzed, and the statistical results are used as electrical signal features. And detect the sleep physiological waves in the EEG data, extract the characteristics of the sleep physiological waves, such as the position, width, peak height, etc. of the peak, or perform statistics on the data of the sleep physiological waves to obtain sleep characteristics.

[0085] In one implementation, if EEG data and other biomedical monitoring data of the subject are obtained in step S210, the first feature may be extracted from the EEG data and other biomedical monitoring data.

[0086] In one embodiment, reference Figure 4 As shown, the above-mentioned extraction of the first feature from the EEG data may include the following steps S410 to S440:

[0087] Step S410, inputting the EEG data into a feature extraction model; the feature extraction model includes a basic feature layer and a plurality of parallel deep feature layers, and different deep feature layers have feature extraction units of different scales.

[0088] The feature extraction model is a model for extracting the first feature, and can be any type of machine learning model, such as a convolutional neural network. In addition, the feature extraction model can be a pre-trained neural network, or a neural network that has been trained (such as fine-tuning) on ​​a specific data set.

[0089] In the feature extraction model, the basic feature layer is the part that extracts basic and shallow features, and the deep feature layer is the part that further extracts deep features. Multiple deep feature layers are arranged in parallel, and deep features of different scales can be extracted in parallel. It can be seen that after the EEG data is input into the feature extraction model, the two stages of processing, basic feature extraction and deep feature extraction, are mainly performed. The following is explained through steps S420 and S430 respectively.

[0090] Step S420, processing the EEG data through the basic feature layer to obtain basic features.

[0091] For example, the basic feature layer may be an embedding layer, which obtains basic features by embedding EEG data.

[0092] In one embodiment, the basic feature layer includes a self-attention layer. The above-mentioned processing of EEG data through the basic feature layer to obtain basic features may include the following steps:

[0093] The EEG data is input into the self-attention layer, and the EEG data is used as the query information, key information and value information of the self-attention layer respectively, and the basic features are obtained through processing.

[0094] That is to say, in the self-attention layer, the query information (query, referred to as Q), key information (key, referred to as K), and value information (value, referred to as V) all come from the EEG data. The self-attention layer can mine the associations between different parts of the EEG data, such as the associations between data from different discharge channels, and thus extract basic features.

[0095] Step S430: Process the basic features respectively through multiple depth feature layers to obtain multiple depth features corresponding to the multiple depth feature layers.

[0096] Among them, multiple deep feature layers can be connected to the basic feature layer in parallel to form multiple branches after the basic feature layer, and extract deep features from the basic features respectively.

[0097] In one embodiment, the feature extraction unit includes convolutional layers, different depth feature layers have different numbers of convolutional layers, and the convolutional layers in each depth feature layer are arranged in order from small to large convolution scales. For example, Figure 5A schematic diagram of extracting a first feature through a feature extraction model is shown. The feature extraction model includes at least three deep feature layers. Deep feature layer 1 has a convolution layer k11; deep feature layer 2 has two convolution layers k21 and k22, and the convolution scale of k21 (such as the convolution kernel scale) is smaller than k22; deep feature layer 3 has three convolution layers k31, k32, and k33, and the convolution scale of k31 is smaller than k32, and the convolution scale of k32 is smaller than k33.

[0098] The above-mentioned processing of the basic features through multiple depth feature layers respectively to obtain multiple depth features corresponding to the multiple depth feature layers may include the following steps:

[0099] The basic features are input into each deep feature layer respectively, and processed by each convolution layer in each deep feature layer in turn to obtain the deep features corresponding to each deep feature layer.

[0100] For example, the basic feature F0 is input into the depth feature layer 1, and convolution processing is performed through the convolution layer k11 to obtain the depth feature F1 corresponding to the depth feature layer 1; the basic feature F0 is input into the depth feature layer 2, and convolution processing is performed through the convolution layer k21. The processed intermediate data is then convolution processing is performed through the convolution layer k22 to obtain the depth feature F2 corresponding to the depth feature layer 2; the basic feature F0 is input into the depth feature layer 3, and convolution processing is performed through the convolution layers k31, k32, and k33 in sequence to obtain the depth feature F3 corresponding to the depth feature layer 3.

[0101] In this way, through different deep feature layers, feature mining of different scales and levels is achieved, which is conducive to obtaining more comprehensive deep features.

[0102] Step S440: obtaining a first feature according to a plurality of depth features.

[0103] Multiple deep features may be fused to obtain the first feature. The fusion method includes but is not limited to any one or more of concatenation, addition, averaging, and weighted averaging.

[0104] based on Figure 4 The method ensures the validity and comprehensiveness of the information of the extracted first feature, which is conducive to the subsequent accurate detection of human brain discharge channels.

[0105] refer to Figure 5 As shown in the figure, after the EEG data is input into the feature extraction model, shallow features are first extracted through the self-attention layer to obtain basic features; then deep features of different levels are extracted through multiple parallel deep feature layers to obtain deep features of different scales; finally, the deep features of different scales are spliced ​​to obtain the first feature.

[0106] Continue to refer Figure 2 In step S230, discharge channel codes corresponding to multiple preset discharge channels are obtained; the multiple preset discharge channels include discharge channels corresponding to the EEG data.

[0107] The preset discharge channel may refer to a channel where abnormal discharge may occur, such as a discharge channel included in the EEG data. For example, if the EEG data acquired in step S210 includes 23 discharge channels, then in step S230, the discharge channel codes corresponding to the 23 discharge channels may be acquired accordingly. The discharge channel code is an abstract numerical representation of the entity object of the discharge channel. For example, the preset discharge channel may be encoded in an embedded manner to obtain the discharge channel code.

[0108] In one embodiment, the above-mentioned obtaining the discharge channel codes corresponding to the plurality of preset discharge channels may include at least one of the following steps:

[0109] The first channel label indicating that a single preset discharge channel has abnormal discharge is encoded to obtain a discharge channel code. For example, the preset discharge channel includes 23 discharge channels, and the corresponding 23 first channel labels can be obtained, each of which indicates that a single preset discharge channel has abnormal discharge. The first channel label can be encoded, such as by using a one-hot encoding method, encoding the dimension value corresponding to the preset discharge channel where abnormal discharge occurs as 1, and encoding the remaining dimension values ​​as 0, to obtain the discharge channel code. In addition, other encoding methods can also be used, such as further obtaining the discharge channel code by embedding based on the one-hot encoding.

[0110] The second channel labels indicating abnormal discharges in at least two preset discharge channels are encoded to obtain discharge channel codes. Each second channel label indicates that abnormal discharges have occurred in multiple preset discharge channels. The dimension values ​​corresponding to the preset discharge channels where abnormal discharges have occurred can be encoded as 1, and the remaining dimension values ​​can be encoded as 0 to obtain the discharge channel codes. Other encoding methods can also be used, such as further obtaining the discharge channel codes by embedding.

[0111] Continue to refer Figure 2 In step S240, the first feature and the discharge channel code are input into the discharge channel detection model to determine the abnormal discharge channel of the object under test.

[0112] The discharge channel detection model is a model for detecting abnormal discharge channels, and can be any type of machine learning model, such as a neural network. In addition, the discharge channel detection model can be a pre-trained neural network, or a neural network that has been trained (such as fine-tuned) on a specific data set. The discharge channel detection model is used to process the first feature and the discharge channel encoding, and a discharge channel detection result can be output, which can characterize which discharge channel of the object under test is an abnormal discharge channel.

[0113] In one embodiment, the discharge channel detection model includes a feature mapping network, a feature fusion network, and a classification network. Figure 6 As shown, the above-mentioned inputting the first feature and the discharge channel code into the discharge channel detection model to determine the abnormal discharge channel of the detected object may include the following steps S610 to S630:

[0114] Step S610, mapping the first feature and the discharge channel code to the feature space through a feature mapping network to obtain a second feature corresponding to the first feature and a third feature corresponding to the discharge channel code.

[0115] Among them, the first feature and the discharge channel code belong to information of different modes, which are mapped to a unified feature space for subsequent fusion. The feature mapping network can realize the mapping between different spaces, such as the mapping between the space where the first feature is located and the feature space, and the mapping between the space where the discharge channel code is located and the feature space. Exemplarily, the feature mapping network includes a feature matrix for realizing the mapping between different spaces. The first feature and the discharge channel code are input into the feature mapping network, and the information of the two different modes can be aligned to obtain the second feature and the third feature in the feature space.

[0116] In one embodiment, the feature mapping network can implement feature mapping in an embedding manner. For example, the sparse discharge channel coding is embedded through the feature mapping network to map it into a dense third feature.

[0117] In one embodiment, the first feature can be used as the second feature, that is, the first feature does not need to be mapped. For example, the discharge channel code is mapped to a feature space through a feature mapping network, and the feature space is the space where the first feature is located, thereby obtaining a third feature corresponding to the discharge channel code.

[0118] In one implementation, the discharge channel code is represented by an initialization vector to obtain an initialization vector corresponding to the discharge channel code, and the initialization vector is embedded through a feature mapping network to map it to the feature space where the first feature is located to obtain a third feature. The first feature is used as the second feature. The second feature and the third feature can be processed later.

[0119] In one embodiment, the feature mapping network includes a first mapping unit and a second mapping unit. The above-mentioned mapping of the first feature and the discharge channel code to the feature space through the feature mapping network to obtain the second feature corresponding to the first feature and the third feature corresponding to the discharge channel code may include the following steps:

[0120] Inputting the first feature into the trained first mapping unit to obtain a second feature;

[0121] The discharge channel code is input into the trained second mapping unit to obtain the third feature.

[0122] In this way, in the feature mapping network, the first feature and the discharge channel encoding can be processed separately, which is conducive to the accurate realization of information alignment and unification of feature space.

[0123] Step S620, fusing the second feature and the third feature through a feature fusion network to obtain a fourth feature.

[0124] The fusion method includes but is not limited to concatenation, addition, averaging, weighted averaging, etc. In one embodiment, the feature fusion network includes a cross attention network. The above-mentioned fusion of the second feature and the third feature through the feature fusion network to obtain the fourth feature may include the following steps:

[0125] The second feature and the third feature are input into the cross-attention network, and the second feature is used as the query information of the cross-attention network, and the third feature is used as the key information and value information of the cross-attention network, and the fourth feature is obtained through processing.

[0126] In the cross-attention network, the query information (Q) comes from the second feature, and the key information (K) and value information (V) come from the third feature. The cross-attention network can learn the interaction between the second and third features, especially the association between different parts of the data in the second feature and different parts of the data in the third feature. In this way, the second and third features are fused to obtain the fourth feature.

[0127] Step S630: Process the fourth feature through a classification network to determine the abnormal discharge channel of the object under test.

[0128] Exemplarily, the classification network may include one or more fully connected layers, and output the discharge channel detection result by performing fully connected processing on the fourth feature.

[0129] In one embodiment, reference Figure 7 As shown, the method for detecting discharge channels in the human brain may further include the following steps S710 to S740:

[0130] Step S710, obtaining reference data corresponding to a plurality of preset discharge channels.

[0131] Among them, the reference data corresponding to the preset discharge channel may be the EEG data of abnormal discharge in the preset discharge channel, which provides a reference for EEG characteristics when abnormal discharge occurs in the preset discharge channel. In one embodiment, the reference data corresponding to multiple preset discharge channels may include the reference data corresponding to a single preset discharge channel, and the reference data corresponding to a single preset discharge channel includes: EEG data of abnormal discharge in a single preset discharge channel and no abnormal discharge in other preset discharge channels. For example, the EEG data of each preset discharge channel corresponding to its abnormal discharge (no abnormal discharge in other preset discharge channels) may be obtained separately as reference data.

[0132] Step S720: extract the first sample feature from the reference data, input the first sample feature into the first mapping unit to be trained, and obtain the second sample feature.

[0133] The first sample feature can be extracted from the reference data in the manner of step S220. For example, the reference data is input into the above-mentioned feature extraction model to obtain the first sample feature. The first sample feature is input into the first mapping unit to be trained, and the first mapping unit can map the first sample feature to the feature space (it should be noted that since the first mapping unit is in a state to be trained at this time, the mapping process performed by it may not be accurate, that is, it may not be accurately mapped to the feature space), and the second sample feature is obtained.

[0134] Step S730: input the discharge channel code into the second mapping unit to be trained to obtain a third sample feature.

[0135] The discharge channel code is input into the second mapping unit to be trained, and the second mapping unit can map the discharge channel code to the feature space (it should be noted that since the second mapping unit is in a state to be trained at this time, the mapping processing it performs may not be accurate, that is, it may not be accurately mapped to the feature space), and the third sample feature is obtained.

[0136] Step S740, matching the second sample feature and the third sample feature, and updating the first mapping unit and the second mapping unit according to the matching result and the correspondence between the reference data and the discharge channel code to obtain a trained first mapping unit and a trained second mapping unit.

[0137] According to the corresponding relationship between the reference data and the discharge channel code, it can be determined whether the second sample feature and the third sample feature are derived from the same preset discharge channel. If the second sample feature and the third sample feature are derived from the same preset discharge channel, the second sample feature and the third sample feature should have a matching relationship. For example, the reference data corresponding to the preset discharge channel T9 is obtained, in which abnormal discharge occurs in T9 and abnormal discharge does not occur in other preset discharge channels. The first sample feature is extracted from the reference data and mapped to the second sample feature through the first mapping unit. The discharge channel code of the preset discharge channel T9 is obtained, and the discharge channel code indicates that abnormal discharge occurs in T9 and abnormal discharge does not occur in other preset discharge channels. The discharge channel code is mapped to the third sample feature through the second mapping unit. The second sample feature and the third sample feature both represent the information that abnormal discharge occurs alone in T9, and the two should match. The similarity between the second sample feature and the third sample feature can be calculated. If the similarity reaches a preset similarity threshold (which can be determined based on experience or specific circumstances), the two match. The second sample feature and the third sample feature derived from the same preset discharge channel should match, and the second sample feature and the third sample feature derived from different preset discharge channels should not match. If the actual matching result deviates, it means that the first mapping unit or the second mapping unit needs to be trained. The first mapping unit and the second mapping unit can be updated accordingly to realize the training of the first mapping unit and the second mapping unit. After training, the first mapping unit and the second mapping unit can learn the correlation between the characteristics of the EEG data and the two different modal data of the discharge channel encoding, and find the feature space for unifying the two modal features. In this way, in the actual discharge channel detection task, the information of the discharge channel encoding can be accurately embedded in the detection process of the discharge channel detection model, thereby improving the accuracy of the detection results.

[0138] In one implementation, the feature extraction model may be updated according to the matching result of the second sample feature and the third sample feature, and the corresponding relationship between the reference data and the discharge channel coding, so as to implement the training of the feature extraction model.

[0139] In one implementation, the loss function can be constructed as follows:

[0140] (1)

[0141] Wherein, x represents the reference data corresponding to the preset discharge channel, y represents the discharge channel code corresponding to the preset discharge channel, FS2(x) represents the second sample feature extracted based on the reference data x, and FS3(y) represents the third sample feature extracted based on the discharge channel code y. Q(FS2(x), FS3(y)) represents the matching result of the second sample feature and the third sample feature, and the value is 1 (indicating that the second sample feature and the third sample feature match) or 0 (indicating that the second sample feature and the third sample feature do not match). P(x,y) represents the corresponding relationship between the reference data x and the discharge channel code y. If the reference data x and the discharge channel code y correspond to the same preset discharge channel, then P(x,y)=1. If the reference data x and the discharge channel code y correspond to different preset discharge channels, then P(x,y)=0. It can be seen that the above loss function characterizes the consistency between the matching result of the second sample feature and the third sample feature and the corresponding relationship between the reference data and the discharge channel code. If the matching result and the corresponding relationship are inconsistent, the loss function value is large. Thus, the first mapping unit and the second mapping unit are updated, and the feature extraction model can also be updated. By training these models, feature alignment of two different modalities of data, EEG data and discharge channel encoding, can be achieved.

[0142] In one embodiment, reference Figure 8 As shown in FIG. 1 , a multi-channel EEG can be collected by an EEG machine as training sample data. The data is recorded in segments, non-abnormal discharge segments are manually filtered, and specific discharge channels of abnormal discharge segments are manually labeled. For example, Fig.9A and Fig. 9B The multi-channel EEG shown can be manually framed to select the part where abnormal discharge occurs. This will obtain the labeled data of the abnormal discharge channel. The discharge channel detection model is trained using the training sample data, discharge channel encoding, and labeled data. This enables the discharge channel detection model to accurately detect the abnormal discharge channel in the EEG data.

[0143] Exemplary Devices

[0144] The following is a description of the human brain discharge channel detection device in the embodiment of the present disclosure. Fig.10 As shown, the human brain discharge channel detection device 1000 may include the following modules:

[0145] The EEG data acquisition module 1010 is configured to acquire EEG data of the subject;

[0146] A first feature extraction module 1020 is configured to extract a first feature from the EEG data;

[0147] The discharge channel code acquisition module 1030 is configured to acquire discharge channel codes corresponding to a plurality of preset discharge channels; the plurality of preset discharge channels include discharge channels corresponding to the EEG data;

[0148] The model processing module 1040 is configured to input the first feature and the discharge channel code into the discharge channel detection model to determine the abnormal discharge channel of the object under test.

[0149] The above modules can be program modules or hardware modules.

[0150] In one embodiment, extracting a first feature from EEG data includes: inputting the EEG data into a feature extraction model; the feature extraction model includes a basic feature layer and multiple parallel deep feature layers, and different deep feature layers have feature extraction units of different scales; processing the EEG data through the basic feature layer to obtain basic features; processing the basic features through multiple deep feature layers respectively to obtain multiple depth features corresponding to the multiple depth feature layers; and obtaining the first feature based on the multiple depth features.

[0151] In one embodiment, the feature extraction unit includes a convolution layer, different deep feature layers have different numbers of convolution layers, and the convolution layers in each deep feature layer are arranged in sequence from small to large in convolution scale; basic features are processed respectively by multiple deep feature layers to obtain multiple deep features corresponding to the multiple deep feature layers, including: inputting the basic features into each deep feature layer respectively, and processing each convolution layer in each deep feature layer in turn to obtain the depth features corresponding to each deep feature layer.

[0152] In one embodiment, the basic feature layer includes a self-attention layer; the EEG data is processed through the basic feature layer to obtain basic features, including: inputting the EEG data into the self-attention layer, and using the EEG data as query information, key information and value information of the self-attention layer respectively, and obtaining the basic features through processing.

[0153] In one embodiment, a discharge channel detection model includes a feature mapping network, a feature fusion network, and a classification network; a first feature and a discharge channel code are input into the discharge channel detection model to determine an abnormal discharge channel of the object under test, including: mapping the first feature and the discharge channel code to a feature space through a feature mapping network to obtain a second feature corresponding to the first feature and a third feature corresponding to the discharge channel code; fusing the second feature and the third feature through a feature fusion network to obtain a fourth feature; and processing the fourth feature through a classification network to determine the abnormal discharge channel of the object under test.

[0154] In one embodiment, the feature fusion network includes a cross-attention network; the second feature and the third feature are fused through the feature fusion network to obtain a fourth feature, including: inputting the second feature and the third feature into the cross-attention network, and using the second feature as query information of the cross-attention network, and the third feature as key information and value information of the cross-attention network, and obtaining the fourth feature through processing.

[0155] In one embodiment, the feature mapping network includes a first mapping unit and a second mapping unit; the first feature and the discharge channel code are mapped to the feature space through the feature mapping network to obtain the second feature corresponding to the first feature and the third feature corresponding to the discharge channel code, including: inputting the first feature into the trained first mapping unit to obtain the second feature; inputting the discharge channel code into the trained second mapping unit to obtain the third feature.

[0156] In one embodiment, the human brain discharge channel detection device 1000 also includes a model training module, which is configured to: obtain reference data corresponding to multiple preset discharge channels; extract a first sample feature from the reference data, input the first sample feature into a first mapping unit to be trained, and obtain a second sample feature; input the discharge channel code into a second mapping unit to be trained, and obtain a third sample feature; match the second sample feature and the third sample feature, and update the first mapping unit and the second mapping unit according to the matching result and the correspondence between the reference data and the discharge channel code to obtain a trained first mapping unit and a trained second mapping unit.

[0157] In one embodiment, the reference data corresponding to multiple preset discharge channels include reference data corresponding to a single preset discharge channel, and the reference data corresponding to the single preset discharge channel includes: EEG data in which the single preset discharge channel has abnormal discharge and other preset discharge channels do not have abnormal discharge.

[0158] In one embodiment, obtaining discharge channel codes corresponding to multiple preset discharge channels includes at least one of the following steps: encoding a first channel label indicating that abnormal discharge occurs in a single preset discharge channel to obtain a discharge channel code; encoding a second channel label indicating that abnormal discharge occurs in at least two preset discharge channels to obtain a discharge channel code.

[0159] In addition, other specific details of the embodiments of the present disclosure have been described in detail in the embodiments of the above method and will not be repeated here.

[0160] Exemplary Program Products

[0161] The computer program product in the embodiment of the present disclosure is described below. The computer program product includes a computer program, and when the computer program is executed by a processor, the above method of the present disclosure is implemented.

[0162] In one embodiment, the computer program product may be a tangible product, such as a computer-readable storage medium storing a computer program. The readable storage medium may be based on electrical, magnetic, optical, electromagnetic, infrared, and other signals, including but not limited to: random access memory (RAM), read-only memory (ROM), magnetic tape, floppy disk, flash memory (Flash), mechanical hard disk (HDD), solid state drive (SSD), and the like. Exemplarily, the computer program product may be a non-volatile storage medium storing a computer program, such as a read-only memory, a NAND flash memory (Nand Flash), and the like.

[0163] In one embodiment, the computer program product may be an intangible product. For example, the computer program product may be a virtual digital product, such as an executable file or installation package containing a computer program.

[0164] The code of the computer program can be written in one or more programming languages. Programming languages ​​include, but are not limited to, C, Java, C++, etc. The program code can be executed entirely on the user computing device, or partially on the user computing device, or as a separate software package, or partially on the user computing device and partially on a remote computing device, or entirely on a remote computing device or server. In the case of a remote computing device, the remote computing device can be connected to the user computing device through any type of network, such as a local area network (LAN), a wide area network (WAN), etc., or can be connected to an external computing device (e.g., an Internet connection provided by an operator).

[0165] Computer programs can be carried or transmitted through electrical, magnetic, optical, electromagnetic, infrared and other signals. Electronic devices can convert signals carrying computer programs into digital signals, and then run the computer programs. When a computer program runs on an electronic device, its code is used to enable the electronic device to execute (more specifically, it can enable the processor of the electronic device to execute) the method steps of various embodiments of the present disclosure, for example: step S210, obtaining the EEG data of the subject; step S220, extracting the first feature from the EEG data; step S230, obtaining the discharge channel codes corresponding to multiple preset discharge channels; multiple preset discharge channels include the discharge channels corresponding to the EEG data; step S240, inputting the first feature and the discharge channel code into the discharge channel detection model to determine the abnormal discharge channel of the subject.

[0166] The above steps are performed by a computer program to obtain the EEG data of the subject, extract the first feature from the EEG data, obtain the discharge channel codes corresponding to multiple preset discharge channels, and input the first feature and the discharge channel code into the discharge channel detection model to determine the abnormal discharge channel of the subject. Automated detection of abnormal discharge channels in the human brain is achieved. The discharge channel code can guide the processing process of the discharge channel detection model, improve the model's ability to distinguish abnormal discharge channels, and improve the accuracy of the test results. The test results provide comprehensive and accurate auxiliary diagnostic information, which can help doctors locate the abnormal discharge position of the subject, and help doctors perform detailed and targeted diagnosis and treatment based on the test results.

[0167] Exemplary Electronic Devices

[0168] The electronic device in the embodiment of the present disclosure is described below. The electronic device may be Figure 1A or Figure 1B Any device in the present invention. The electronic device includes a processor and a memory, and the memory is used to store executable instructions of the processor. The processor is configured to execute the above method of the present invention by executing the executable instructions.

[0169] refer to Fig.11 An electronic device according to an embodiment of the present disclosure is exemplified. Fig.11 The electronic device 1100 shown is merely an example and should not bring any limitation to the functions and scope of use of the embodiments of the present disclosure.

[0170] like Fig.11 As shown, the electronic device 1100 is presented in the form of a general-purpose computing device. The components of the electronic device 1100 may include, but are not limited to: a processor 1110, a memory 1120, a bus 1130 connecting different system components (including the memory 1120 and the processor 1110), an I / O (input / output) interface 1140, and a network adapter 1150.

[0171] The memory 1120 stores program codes, and the program codes can be executed by the processor 1110, so that the processor 1110 executes the method steps of the embodiment of the present disclosure. For example, the processor 1110 can execute the following steps: step S210, obtaining the EEG data of the subject; step S220, extracting the first feature from the EEG data; step S230, obtaining the discharge channel codes corresponding to the multiple preset discharge channels; the multiple preset discharge channels include the discharge channels corresponding to the EEG data; step S240, inputting the first feature and the discharge channel code into the discharge channel detection model to determine the abnormal discharge channel of the subject.

[0172] The processor 1110 performs the above steps to obtain the EEG data of the subject, extract the first feature from the EEG data, obtain the discharge channel codes corresponding to multiple preset discharge channels, and input the first feature and the discharge channel code into the discharge channel detection model to determine the abnormal discharge channel of the subject. Automated detection of abnormal discharge channels in the human brain is achieved. The discharge channel code can guide the processing process of the discharge channel detection model, improve the model's ability to distinguish abnormal discharge channels, and improve the accuracy of the test results. The test results provide comprehensive and accurate auxiliary diagnostic information, which can help doctors locate the abnormal discharge position of the subject, and help doctors perform detailed and targeted diagnosis and treatment based on the test results.

[0173] The memory 1120 may include a volatile memory, such as a random access memory (RAM) 1121 and / or a cache unit 1122, and may also include a non-volatile memory, such as a read-only memory unit (ROM) 1123. The memory 1120 may also include one or more program modules 1124, such program modules 1124 include but are not limited to: an operating system, one or more application programs, other program modules, and program data, each of which or a certain combination may include the implementation of a network environment. For example, the program module 1124 may include each module in the above-mentioned device.

[0174] The processor 1110 may include processing units such as an AP (Application Processor), a modem processor, a GPU (Graphics Processing Unit), an ISP (Image Signal Processor), a controller, an encoder, a decoder, a DSP (Digital Signal Processor), a baseband processor and / or an NPU (Neural-Network Processing Unit).

[0175] The bus 1130 is used to connect different parts of the electronic device 1100 and may include a data bus, an address bus, a control bus, and the like.

[0176] The electronic device 1100 may also communicate with one or more external devices 1200 (eg, a keyboard, a pointing device, a Bluetooth device, etc.), and such communication may be performed through the I / O interface 1140 .

[0177] The electronic device 1100 can also communicate with one or more networks through the network adapter 1150. For example, the network adapter 1150 can provide mobile communication solutions such as 3G / 4G / 5G, or wireless communication solutions such as wireless LAN, Bluetooth, near field communication, etc. The network adapter 1150 can communicate with other modules of the electronic device 1100 through the bus 1130.

[0178] Although not shown in the figure, other hardware and / or software modules may be provided in the electronic device 1100, including but not limited to: a display, a microcode, a device driver, a redundant processing unit, an external disk drive array, a RAID (Redundant Arrays of Independent Disks) system, a tape drive, and a data backup storage system.

[0179] It should be noted that, although several modules or submodules of the device are mentioned in the above detailed description, such division is merely exemplary and not mandatory. In fact, according to the embodiments of the present disclosure, the features and functions of two or more units / modules described above can be embodied in one unit / module. Conversely, the features and functions of one unit / module described above can be further divided into multiple units / modules to be embodied.

[0180] In addition, although the operations of the disclosed method are described in a specific order in the drawings, this does not require or imply that the operations must be performed in this specific order, or that all the operations shown must be performed to achieve the desired results. Additionally or alternatively, some steps may be omitted, multiple steps may be combined into one step, and / or one step may be decomposed into multiple steps.

[0181] Although the spirit and principle of the present disclosure have been described with reference to several specific embodiments, it should be understood that the present disclosure is not limited to the disclosed specific embodiments, and the division of various aspects does not mean that the features in these aspects cannot be combined to benefit, and such division is only for the convenience of expression. The present disclosure is intended to cover various modifications and equivalent arrangements included in the spirit and scope of the attached claims.

Claims

1. A method for detecting discharge channels in the human brain, characterized in that: The method comprises: Obtaining EEG data of the subject; Extracting a first feature from the EEG data; Acquire discharge channel codes corresponding to a plurality of preset discharge channels; the plurality of preset discharge channels include the discharge channels corresponding to the EEG data; Inputting the first feature and the discharge channel code into a discharge channel detection model to determine the abnormal discharge channel of the object under test; Wherein, the step of obtaining discharge channel codes corresponding to a plurality of preset discharge channels comprises at least one of the following steps: Encoding a first channel label indicating that abnormal discharge occurs in a single preset discharge channel to obtain the discharge channel code; The second channel labels indicating that abnormal discharge occurs in at least two preset discharge channels are encoded to obtain the discharge channel codes.

2. The method according to claim 1, characterized in that The extracting a first feature from the EEG data comprises: Inputting the EEG data into a feature extraction model; the feature extraction model comprises a basic feature layer and a plurality of parallel deep feature layers, and different deep feature layers have feature extraction units of different scales; Processing the EEG data through the basic feature layer to obtain basic features; The basic features are processed respectively through multiple depth feature layers to obtain multiple depth features corresponding to the multiple depth feature layers; The first feature is obtained according to the multiple depth features.

3. The method according to claim 2, characterized in that The feature extraction unit includes a convolution layer, different depth feature layers have different numbers of convolution layers, and the convolution layers in each depth feature layer are arranged in order from small to large convolution scales; The step of processing the basic features respectively through a plurality of depth feature layers to obtain a plurality of depth features corresponding to the plurality of depth feature layers includes: The basic features are input into each deep feature layer respectively, and are processed by each convolution layer in each deep feature layer in turn to obtain the deep features corresponding to each deep feature layer.

4. The method according to claim 2, characterized in that: The basic feature layer includes a self-attention layer; the EEG data is processed by the basic feature layer to obtain basic features, including: The EEG data is input into the self-attention layer, and the EEG data is used as the query information, key information and value information of the self-attention layer respectively, and the basic features are obtained through processing.

5. The method according to claim 1, characterized in that: The discharge channel detection model includes a feature mapping network, a feature fusion network, and a classification network; the step of inputting the first feature and the discharge channel encoding into the discharge channel detection model to determine the abnormal discharge channel of the object under test includes: Mapping the first feature and the discharge channel code to a feature space through the feature mapping network to obtain a second feature corresponding to the first feature and a third feature corresponding to the discharge channel code; fusing the second feature and the third feature through the feature fusion network to obtain a fourth feature; The fourth feature is processed by the classification network to determine an abnormal discharge channel of the object under test.

6. The method according to claim 5, characterized in that The feature fusion network includes a cross attention network; and fusing the second feature and the third feature through the feature fusion network to obtain a fourth feature includes: The second feature and the third feature are input into the cross-attention network, and the second feature is used as the query information of the cross-attention network, and the third feature is used as the key information and value information of the cross-attention network, and the fourth feature is obtained through processing.

7. The method according to claim 5, characterized in that The feature mapping network includes a first mapping unit and a second mapping unit; the first feature and the discharge channel code are mapped to a feature space through the feature mapping network to obtain a second feature corresponding to the first feature and a third feature corresponding to the discharge channel code, including: Inputting the first feature into a trained first mapping unit to obtain the second feature; The discharge channel code is input into the trained second mapping unit to obtain the third feature.

8. The method according to claim 7, characterized in that The method further comprises: Acquire reference data corresponding to the plurality of preset discharge channels; Extracting a first sample feature from the reference data, and inputting the first sample feature into a first mapping unit to be trained to obtain a second sample feature; Inputting the discharge channel code into a second mapping unit to be trained to obtain a third sample feature; The second sample feature and the third sample feature are matched, and the first mapping unit and the second mapping unit are updated according to the matching result and the corresponding relationship between the reference data and the discharge channel code to obtain a trained first mapping unit and a trained second mapping unit.

9. The method according to claim 7, characterized in that: The reference data corresponding to the multiple preset discharge channels include reference data corresponding to a single preset discharge channel, and the reference data corresponding to the single preset discharge channel includes: EEG data in which abnormal discharge occurs in the single preset discharge channel and abnormal discharge does not occur in other preset discharge channels.

10. A human brain discharge channel detection device, characterized in that: The device comprises: An EEG data acquisition module is configured to acquire EEG data of a subject; A first feature extraction module is configured to extract a first feature from the EEG data; A discharge channel code acquisition module is configured to acquire discharge channel codes corresponding to a plurality of preset discharge channels; the plurality of preset discharge channels include the discharge channels corresponding to the EEG data; A model processing module is configured to input the first feature and the discharge channel code into a discharge channel detection model to determine an abnormal discharge channel of the object under test; Wherein, the step of obtaining discharge channel codes corresponding to a plurality of preset discharge channels comprises at least one of the following steps: Encoding a first channel label indicating that abnormal discharge occurs in a single preset discharge channel to obtain the discharge channel code; The second channel labels indicating that abnormal discharge occurs in at least two preset discharge channels are encoded to obtain the discharge channel codes.

11. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the method according to any one of claims 1 to 9 is implemented.

12. An electronic device, characterized in that: include: processor; as well as A memory, configured to store executable instructions of the processor; The processor is configured to perform the method of any one of claims 1 to 9 by executing the executable instructions.

Images