A method, device and program product for classifying depression and bipolar disorder based on single-channel electroencephalogram signals

Abstract

The application relates to the field of intelligent medical treatment, in particular to a depression and bipolar classification method, equipment and program product based on single-channel electroencephalogram signals. The single-channel electroencephalogram signal acquisition equipment comprises the following steps: fixing the single-channel electroencephalogram signal acquisition equipment on the frontal lobe of the brain of a person to be measured; acquiring one-dimensional time sequence electroencephalogram signals of the single-channel electroencephalogram signal acquisition equipment when the person to be measured performs a task action; converting the one-dimensional time sequence electroencephalogram signals into two-dimensional electroencephalogram signals based on different time dimensions; and inputting the two-dimensional electroencephalogram signals into a depression and bipolar classification model to obtain a classification result of health, depression and bipolar. The application can extract single-channel task-state electroencephalogram signals to perform classification and judgment of health, depression and bipolar of a tester, can carry out depression screening work anytime and anywhere in a community, a school, a family and the like, solves the use scene limitation of a traditional multi-channel electroencephalogram device, and has good application value.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of intelligent medical treatment, in particular to a depression and bipolar disorder classification method based on single-channel electroencephalogram signals, a device, a program product and a computer readable storage medium. BACKGROUND

[0002] Depression has the characteristics of high prevalence and low recognition rate, which easily leads to missed diagnosis, poor prognosis and other problems, and urgently needs effective objective indicators for auxiliary diagnosis. According to statistics, the number of depression cases is high, but the recognition rate of depression is low. In existing research, more than 50,000 patients have been collected, but less than half of the patients correctly identified depression. At present, the mainstream diagnosis method of depression is through psychiatric interview of patients and their families, supplemented by Diagnostic and Statistical Manual of Mental Disorders (DSM-IV) and Psychiatric Rating Scale, which lacks objective depression evaluation standard, has poor consistency, strong subjectivity, long diagnosis time, and patients may ignore early symptoms or refuse treatment due to fear or honest conversation. In addition, there is some overlap between the symptoms of depression and bipolar disorder, which affects the hippocampus, prefrontal cortex and amygdala, and there is some commonality in brain functional connectivity, which easily leads to misdiagnosis. Previous studies have found that about 40% of bipolar disorder patients are misdiagnosed as unipolar depressive disorder, and the misdiagnosis time can be as long as 5 to 10 years, causing about one-third of bipolar disorder patients to not receive timely and appropriate treatment. How to realize the early and accurate identification of the two is a problem that needs to be solved in current clinical practice. The treatment principles of depression and bipolar disorder are also very different. Depression patients use antidepressants to alleviate depressive mood, and bipolar disorder patients mainly use mood stabilizers for treatment. Therefore, if the disease diagnosis is not accurate, it will cause the wrong treatment plan, and bipolar disorder patients taking inappropriate antidepressant doses may even trigger their manic episodes or exacerbate their condition. Therefore, finding objective auxiliary diagnostic indicators to simultaneously identify depression, bipolar disorder and healthy people can not only optimize clinical decision-making, but also promote scientific research and public health strategies, and ultimately achieve the goal of precision psychiatry.

[0003] Electroencephalogram (EEG) is a non-invasive method for studying brain function, which has been widely used in medical diagnosis, neuroscience research and other fields. EEG can provide information on the functional status of different regions of the brain by recording the potential changes of brain electrical activity, especially in the diagnosis and monitoring of neuropsychiatric diseases. EEG can reflect the electrical activity of the cerebral cortex, including various brain waves related to consciousness, attention, emotion and cognitive function. In the field of mental illness, especially in the study of depressive disorders, EEG has been shown to reveal abnormalities in brain function in patients. By performing spectral analysis, time-domain analysis, nonlinear dynamic analysis and other methods on EEG signals, researchers can identify brain electrical characteristics that are different from normal brain activity. EEG can also be used to differentiate between depressive disorders and bipolar disorders. Studies have found that depressive patients have increased theta power in the left prefrontal cortex, while bipolar patients have higher theta activity in the frontal lobe. EEG changes such as frontal lobe alpha asymmetry and gamma band can also be used to differentiate between the two diseases. Therefore, EEG can provide objective evidence for the diagnosis of mental illness by analyzing the frequency spectrum, time domain and nonlinear characteristics of brain electrical activity to differentiate between depressive disorders, bipolar disorders and healthy individuals. Existing EEG devices mostly rely on multi-channel sensors and usually need to be operated in a laboratory environment, which makes it inconvenient for EEG to be widely used, especially in clinical and home environments. Most current EEG-based studies on depression use 32 channels or more of full-brain EEG signals. However, this process requires professional operation, is complex and relatively expensive, and is not suitable for non-laboratory environments or daily situations. SUMMARY

[0004] Single-channel portable EEG has only one EEG signal acquisition channel, with the characteristics of small size, low power consumption, low cost, simple operation, etc. It can improve the accuracy of machine learning models, accelerate the popularization of EEG technology, and promote the application and development of EEG technology in daily life. To solve the above problems, the present application provides a method for classifying depressive and bipolar disorders based on single-channel EEG signals, which can better help people achieve efficient personal health management, allowing depressive patients to collect EEG signals in their daily lives without being restricted by laboratory space, and assisting large sample populations in the auxiliary diagnosis of depressive disorders and long-term dynamic monitoring in daily environments. Specifically, it includes fixing a single-channel EEG signal acquisition device on the prefrontal cortex of the subject to be tested;

[0005] Obtaining one-dimensional time series EEG signals from the single-channel EEG signal acquisition device when the subject to be tested performs a task action;

[0006] Converting the one-dimensional time series EEG signals into two-dimensional EEG signals based on different time dimensions;

[0007] The two-dimensional brain electrical signal is input into a depressive bipolar classification model to obtain a classification result of health, depression, and bipolar.

[0008] The task actions include an open-eye task, a close-eye task, and a game task.

[0009] Optionally, the subject sequentially performs the open-eye task, the close-eye task, and the game task, and acquires one-dimensional time-series brain electrical signals of performing the task actions, the one-dimensional time-series brain electrical signals of performing the task actions are converted into two-dimensional brain electrical signals based on different time dimensions, and the two-dimensional brain electrical signals are input into a depressive bipolar classification model to perform classification.

[0010] Optionally, the game task is opening a blind box by selection, the box gives a virtual currency income or loss animation, after M rounds, the overall game profit and loss is displayed, and M is a natural number greater than or equal to 50.

[0011] Optionally, when the subject performs the open-eye task, N-minute open-eye brain electrical signals are acquired, N is a natural number greater than or equal to 5, and the N-minute close-eye brain electrical signals are converted into two-dimensional open-eye brain electrical signals based on different time dimensions.

[0012] Optionally, when the subject performs the close-eye task, N-minute close-eye brain electrical signals are acquired, and the N-minute close-eye brain electrical signals are converted into two-dimensional close-eye brain electrical signals based on different time dimensions.

[0013] Optionally, when the subject performs the game task, 2N-minute game brain electrical signals are acquired, and the 2N-minute game brain electrical signals are converted into two-dimensional game brain electrical signals based on different time dimensions.

[0014] The game brain electrical signals further include signal extraction, extraction of S-second brain electrical signals before opening the box and L-second brain electrical signals after opening the box to obtain extracted game brain electrical signals, and input of the open-eye brain electrical signals, the close-eye brain electrical signals, and the extracted game brain electrical signals into a depressive bipolar classification model to obtain a classification result of health, depression, and bipolar; S and L are natural numbers greater than 1.

[0015] Optionally, the game brain electrical signals extract 3-second brain electrical signals before opening the box and 5-second brain electrical signals after opening the box.

[0016] Optionally, the game signal extraction further includes brain electrical signals after the overall game profit and loss is displayed.

[0017] The method further includes data preprocessing, data preprocessing of the one-dimensional time-series brain electrical signals to obtain preprocessed one-dimensional time-series brain electrical signals, and conversion of the preprocessed one-dimensional time-series brain electrical signals into two-dimensional brain electrical signals based on different time dimensions, and the data preprocessing includes one or more of the following: signal filtering, motion artifact removal, and signal quality screening.

[0018] Optionally, the preprocessing is performed in sequence of signal filtering, motion artifact removal, and signal quality screening.

[0019] The signal quality screening is a quality judgment on the electroencephalogram signal, and the signal with poor quality is removed to obtain the preprocessed electroencephalogram signal.

[0020] Optionally, the quality judgment is performed by a sliding window and quartile statistics.

[0021] Optionally, the quality judgment process is as follows:

[0022] The data amplitude of the task action is counted.

[0023] The mean and variance of the data amplitude are calculated, and the upper and lower limits of the data amplitude are determined based on the mean and variance.

[0024] The data points outside the upper and lower limits of the amplitude are counted in each window by moving the window on the data amplitude, and the number of data points is recorded as signal non-effective value.

[0025] The signal non-effective value is compared with a preset threshold value, and when the signal non-effective value is greater than the preset threshold value, the signal is removed; otherwise, the signal is retained.

[0026] Optionally, the quality judgment process further includes counting the distribution, and the signal non-effective value is counted to obtain a statistical distribution value, and the preset threshold value is obtained based on the statistical distribution value.

[0027] Optionally, the preset threshold value is the upper quartile of the signal non-effective value.

[0028] Optionally, the window width of the moving window is half of the sampling frequency.

[0029] The training process of the depressive bipolar classification model is as follows:

[0030] A one-dimensional time series electroencephalogram signal dataset and labels of open eyes, closed eyes, and games are obtained.

[0031] The one-dimensional time series electroencephalogram signal and the labels are processed in two dimensions to obtain a two-dimensional electroencephalogram signal.

[0032] The two-dimensional electroencephalogram signal is input into a classifier to be trained for training until the loss function is constant, and a depressive bipolar classification model is obtained.

[0033] Optionally, the two-dimensional processing is based on converting different time dimension one-dimensional time series electroencephalogram signals into a two-dimensional matrix to obtain a two-dimensional electroencephalogram signal; wherein the rows or columns of the matrix represent time, and the columns or rows of the matrix represent one-dimensional time series electroencephalogram signals.

[0034] Optionally, the classifier comprises one or more of the following: a convolutional neural network, a residual network, an extreme learning machine, a perception vector machine, a random forest, a decision tree, a support vector machine, AdaBoost, and a Transformer.

[0035] The method further comprises secondary prediction, the depression and bipolar classification model obtains a category probability after classification, obtains basic information of the to-be-tested person, and performs secondary prediction based on the category probability and the basic information to obtain a classification prediction result of health, depression, and bipolar;

[0036] Optionally, the secondary prediction is performed by one or more of the following models: logistic regression, Naive Bayes, AdaBoost, GBM, and LDA.

[0037] Optionally, the basic information comprises one or more of the following: age, gender, medical history, and symptom condition.

[0038] The purpose of the present application is to provide a computer program product comprising a computer program or instructions thereon, which are executed by a processor to implement the above-mentioned depression and bipolar classification method based on single-channel electroencephalogram signals.

[0039] The purpose of the present application is to provide a computer device comprising a memory, a processor, and a computer program or instructions stored on the memory, which are executed by the processor to implement the above-mentioned depression and bipolar classification method based on single-channel electroencephalogram signals.

[0040] The purpose of the present application is to provide a computer-readable storage medium having a computer program or instructions stored thereon, which are executed by a processor to implement the above-mentioned depression and bipolar classification method based on single-channel electroencephalogram signals.

[0041] Advantages of the present application:

[0042] 1. The auxiliary diagnosis of previous depression drug treatment mainly depends on the resting state (open eye / closed eye) EEG paradigm. However, the differential diagnosis of depression needs more comprehensive evaluation. Based on the disease mechanism and biological characteristics of EEG signals generated in a specific task state, specific neural activity characteristics associated with clinical symptoms can be effectively captured. This study constructs a model for auxiliary diagnosis of depression based on the reward and punishment mechanism of cognitive task training strategy, and introduces the time-frequency domain signal information before and after the result display, effectively solving the limitations of traditional methods in data heterogeneity and individual resting state EEG differences, providing a new idea for the auxiliary diagnosis of depression, and providing a reliable data basis and theoretical support for establishing a precise auxiliary diagnosis model.

[0043] Specifically, the application provides a blind box opening game. The blind box is opened by selection. After opening, the yield value and loss value of virtual currency are presented. After multiple rounds of opening, the total income and expenditure result is presented. The electroencephalogram signals before opening and after opening, and the electroencephalogram signals after presenting the total income and expenditure are collected and analyzed, so as to enhance the change of electroencephalogram signals of the tester, extract the change characteristics, and improve the recognition accuracy. The deep learning classification method based on reward and punishment task state slices brain electrical data: different from the time domain signal analysis of the past depression screening research, the application proposes a mental illness classification method based on the combination of prefrontal lobe electroencephalogram signal data of reward and punishment task state short time (8 seconds of data) and multiple frequencies (50 rounds of blind box opening), and a two-dimensional convolutional neural network, which obtains a higher level of classification and recognition accuracy of depression, bipolar disorder and healthy people in the industry, and provides a new technical scheme for depression patient screening.

[0044] 2. Although the existing EEG signal classification technology can distinguish depression patients, bipolar disorder patients and healthy people to some extent, it generally faces the core challenges of low classification accuracy of the classifier, weak generalization ability across subjects and lack of reasonable cognitive task to stimulate depression characteristic brain electrical activity. Therefore, the research innovatively constructs a multi-modal dynamic feature learning framework based on deep learning, realizes adaptive extraction of multi-scale features of EEG signals through design of time-frequency domain fusion feature extraction mechanism, specifically: one-dimensional time series data is converted into two-dimensional data; the one-dimensional time series data of the tester S times (50 times) game task state brain electrical data is converted into two-dimensional data with time dimension as row (column), and single game task state brain electrical one-dimensional time series data as column (row); according to the change of time in the game, the feature extraction of bipolar and depression in the fluctuation change of emotion (brain electrical signal) in the game is carried out, so as to improve the distinction of depression and bipolar patients.

[0045] 3. The feature adopts the signal before and after opening; the data before and after opening is processed in the form of game task state, and there is a difference between the brain electrical signals before and after opening in depression and bipolar patients. The feature extraction of the signal data before and after opening is used to improve the distinction of depression and bipolar patients.

[0046] In addition to the conventional filtering, denoising and other processing before model training, the application innovatively proposes to use statistical method to judge the signal quality of the collected task state single channel brain electrical signal, input the signal with good quality into the training model for parameter adjustment, and ensure the stability and reliability of the model.

[0047] 4. The depression and bipolar classification method based on single-channel electroencephalogram signal provided by the present application is a whole, in order to facilitate the use in clinical and family environment, single-channel signal acquisition and single-channel classification are adopted; in order to improve the reliability of single-channel signal, signal quality evaluation screening, signal denoising filtering and other pretreatments are carried out, secondly, in order to improve the feature representation of the signal, the electroencephalogram signal in the task state of the open blind box game is extracted, and the electroencephalogram signals before and after opening the box are further clarified, the processed signals are input into the trained two-dimensional classification model for classification to obtain the classification result, which is a complete technical solution different from the technical concept of the prior art.

[0048] 5. For the existing EEG equipment, most of which depend on multi-channel sensors and usually need to be operated in a laboratory environment, which brings the problem of inconvenience for the wide application of electroencephalogram, especially in clinical and family environment, the present application provides a single-channel electroencephalogram signal acquisition device to acquire the electroencephalogram signal of the subject to be tested, which can be used in clinical or family environment, reduces the data amount of large model input, improves the calculation efficiency, and is more friendly to large-scale disease screening.

[0049] 6. The current EEG classification technology mostly focuses on analyzing resting state data. However, the diagnosis of mental illness needs more comprehensive evaluation. Based on the electroencephalogram signal generated in the specific task state triggered by the disease mechanism and biological characteristics, the typical data and theoretical support for identifying healthy people, people with depression and people with bipolar disorder can be better provided. Therefore, the present application introduces task state data, acquires the electroencephalogram signal of the subject to be tested when performing the task state, improves the identification of abnormal people through the electroencephalogram signal of the specific task, and improves the identification accuracy. BRIEF DESCRIPTION OF DRAWINGS

[0050] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings needed in the embodiment description will be briefly introduced as follows. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0051] Figure 1 The depression and bipolar classification method based on single-channel electroencephalogram signal provided by the present application is a whole, in order to facilitate the use in clinical and family environment, single-channel signal acquisition and single-channel classification are adopted; in order to improve the reliability of single-channel signal, signal quality evaluation screening, signal denoising filtering and other pretreatments are carried out, secondly, in order to improve the feature representation of the signal, the electroencephalogram signal in the task state of the open blind box game is extracted, and the electroencephalogram signals before and after opening the box are further clarified, the processed signals are input into the trained two-dimensional classification model for classification to obtain the classification result, which is a complete technical solution different from the technical concept of the prior art.

[0052] Figure 2 The depression and bipolar classification method based on single-channel electroencephalogram signal provided by the present application is a whole, in order to facilitate the use in clinical and family environment, single-channel signal acquisition and single-channel classification are adopted; in order to improve the reliability of single-channel signal, signal quality evaluation screening, signal denoising filtering and other pretreatments are carried out, secondly, in order to improve the feature representation of the signal, the electroencephalogram signal in the task state of the open blind box game is extracted, and the electroencephalogram signals before and after opening the box are further clarified, the processed signals are input into the trained two-dimensional classification model for classification to obtain the classification result, which is a complete technical solution different from the technical concept of the prior art.

[0053] Figure 3 The depression and bipolar classification method based on single-channel electroencephalogram signal provided by the present application is a whole, in order to facilitate the use in clinical and family environment, single-channel signal acquisition and single-channel classification are adopted; in order to improve the reliability of single-channel signal, signal quality evaluation screening, signal denoising filtering and other pretreatments are carried out, secondly, in order to improve the feature representation of the signal, the electroencephalogram signal in the task state of the open blind box game is extracted, and the electroencephalogram signals before and after opening the box are further clarified, the processed signals are input into the trained two-dimensional classification model for classification to obtain the classification result, which is a complete technical solution different from the technical concept of the prior art.

[0054] Figure 4 A depressive bipolar classification framework structure provided for an embodiment of the present application;

[0055] Figure 5 An execution task schematic diagram provided for an embodiment of the present application;

[0056] Figure 6 A signal sliding window task state electroencephalogram data acquisition process provided for an embodiment of the present application;

[0057] Figure 7 A model quality evaluation threshold value line provided for an embodiment of the present application;

[0058] Figure 8 A classification model structure diagram provided for an embodiment of the present application;

[0059] Figure 9 A model training number and accuracy rate provided for an embodiment of the present application;

[0060] Figure 10 A model training number and loss rate provided for an embodiment of the present application;

[0061] Figure 11 A classification visualization result provided for an embodiment of the present application. DETAILED DESCRIPTION

[0062] In order to enable persons skilled in the art to better understand the present application scheme, the technical solutions in the embodiments of the present application will be described clearly and completely below in conjunction with the drawings in the embodiments of the present application.

[0063] In some of the descriptions in the specification and claims of the present application and the above-mentioned drawings, a plurality of operations appearing in a specific order are included, but it should be clearly understood that these operations can be executed or in parallel without the order appearing in the text, and the serial numbers of the operations such as S101, S102, etc. are only used to distinguish different operations, and the serial numbers themselves do not represent any execution order. In addition, these processes can include more or fewer operations, and the operations can be executed in sequence or in parallel. It should be noted that the "first", "second", etc. in the text are used to distinguish different messages, devices, modules, etc., and do not represent the order, nor do "first" and "second" represent different types.

[0064] Figure 1 The depressive bipolar classification method based on single-channel electroencephalogram signal provided by the embodiment of the present application is a schematic diagram of a single-channel electroencephalogram signal acquisition device, which specifically comprises:

[0065] S101: Fix the single-channel electroencephalogram signal acquisition device on the frontal lobe position of the brain of the person to be tested;

[0066] In one embodiment, the single-channel electroencephalogram signal device receives the electroencephalogram signal of the testee through one channel, wherein the electrode patch collects the electroencephalogram signal at the frontal lobe of the testee, and the signal is transmitted through a transmission line or a wireless port.

[0067] The depression screening device based on the single-channel portable electroencephalogram head ring device combined with the flat plate blind box task software.

[0068] In one specific embodiment, the present application proposes a three-classification algorithm and device for healthy people, people with depressive disorders, and people with bipolar disorders based on wearable electroencephalogram head rings and reward-punishment tasks, which can effectively solve the complexity and high cost problems of traditional multi-channel devices. By combining the population difference characteristics obtained by statistics in the resting state and the EEG signals in the task state for classification, the discrimination ability of depressive patients and healthy people can also be improved. This technical solution is not only suitable for clinical diagnosis, but also can be widely used in remote monitoring and daily health management of mental illness, and promotes the popularization of portable EEG technology in public health.

[0069] S102: Acquire one-dimensional time sequence electroencephalogram signals of a single-channel electroencephalogram signal acquisition device when the testee performs the task action; convert the one-dimensional time sequence electroencephalogram signals into two-dimensional electroencephalogram signals based on different time dimensions;

[0070] In one embodiment, the task action includes an open-eye task, a closed-eye task, and a game task.

[0071] In one embodiment, the testee performs the open-eye task, the closed-eye task, and the game task in turn, and acquires one-dimensional time sequence electroencephalogram signals of performing each task action. The one-dimensional time sequence electroencephalogram signals of performing each task action are converted into two-dimensional electroencephalogram signals based on different time dimensions. The two-dimensional electroencephalogram signals are input into a depressive bipolar classification model for classification.

[0072] In one embodiment, the game task is to open the box by selecting the blind box. The box gives a virtual currency income or loss animation. After M rounds, the overall game profit and loss is displayed. M is a natural number greater than or equal to 50.

[0073] Optionally, when the testee performs the open-eye task, N minutes of open-eye electroencephalogram signals are acquired, N is a natural number greater than or equal to 5, and N minutes of closed-eye electroencephalogram signals are converted into two-dimensional open-eye electroencephalogram signals based on different time dimensions.

[0074] In one embodiment, when the testee performs the closed-eye task, N minutes of closed-eye electroencephalogram signals are acquired, and N minutes of closed-eye electroencephalogram signals are converted into two-dimensional closed-eye electroencephalogram signals based on different time dimensions.

[0075] In an embodiment, the 2N-minute game EEG signal of the subject performing the game task is converted into a two-dimensional game EEG signal based on different time dimensions.

[0076] The two-dimensional open-eye EEG signal, the two-dimensional closed-eye EEG signal, and the two-dimensional game EEG signal are input into the depressive bipolar classification model for classification.

[0077] In an embodiment, the game EEG signal further includes signal extraction, extraction of S-second pre-opening EEG signal and L-second post-opening EEG signal to obtain the extracted game EEG signal, and input of the open-eye EEG signal, the closed-eye EEG signal, and the extracted game EEG signal into the depressive bipolar classification model to obtain the classification results of health, depression, and bipolar; S and L are natural numbers greater than 1.

[0078] In an embodiment, the game EEG signal extracts 3-second pre-opening EEG signal and 5-second post-opening EEG signal.

[0079] In an embodiment, the game signal extraction further includes EEG signal after showing the overall game profit and loss.

[0080] In an embodiment, the one-dimensional time sequence game EEG signal before opening and the one-dimensional time sequence game EEG signal after opening are extracted, and the one-dimensional time sequence game EEG signal before opening and the one-dimensional time sequence game EEG signal after opening are converted into a two-dimensional game EEG signal based on time dimensions.

[0081] In an embodiment, the method further includes data preprocessing, data preprocessing of the one-dimensional time sequence EEG signal to obtain a preprocessed one-dimensional time sequence EEG signal, and conversion of the preprocessed one-dimensional time sequence EEG signal into a two-dimensional EEG signal based on different time dimensions, wherein the data preprocessing includes one or more of the following: signal filtering, motion artifact removal, and signal quality screening.

[0082] In an embodiment, the preprocessing is performed in the order of signal filtering, motion artifact removal, and signal quality screening.

[0083] In an embodiment, the signal quality screening is quality judgment of the EEG signal, and removal of poor quality signals to obtain a preprocessed EEG signal.

[0084] In an embodiment, the quality judgment is performed by a sliding window and quartile statistics.

[0085] In an embodiment, the quality judgment process is as follows:

[0086] Statistical data amplitude of task action;

[0087] Calculate the mean and variance of the data amplitude, and determine the upper and lower limits of the data amplitude based on the mean and variance;

[0088] By moving the window on the data amplitude, the number of data points falling outside the upper and lower limits of the amplitude in each window is obtained, denoted as signal non-effective value;

[0089] Compare the signal non-effective value with the preset threshold value, and remove when the signal non-effective value is greater than the preset threshold value; otherwise, keep.

[0090] In one embodiment, the quality judgment process further comprises statistical distribution, and the statistical distribution of the signal non-effective value obtains a statistical distribution value, and the preset threshold value is obtained based on the statistical distribution value.

[0091] Optionally, the preset threshold value is the upper quarter of the one-bit value of the statistical distribution of the signal non-effective value;

[0092] In one embodiment, the window width of the moving window is half of the sampling frequency.

[0093] In a specific embodiment, the depression and bipolar classification screening mainly includes five modules, namely data acquisition module, data processing module, classification model training module, depression identification module and report output module (as shown in Figure 4 Firstly, the system will give a voice prompt and display a gaze cross on the screen to help the user concentrate. In the open-eye state of the user, the system will collect 5 minutes of resting EEG data. Then, the user will hear a closed-eye prompt sound, and the resting EEG data in the closed-eye state will be collected, also for 5 minutes. Finally, the user will perform about 10 minutes of EEG data collection in the task interaction state. Since low reward sensitivity or abnormal reward sensitivity is a sign of abnormal patients (depression), the EEG task designed by the present application adopts the experimental paradigm of opening a blind box to obtain rewards or punishments, aiming to identify abnormal patients. The user completes 50 rounds of blind box game through an interactive tablet. In each round, the user selects one of the two boxes appearing on the screen, and then an opening animation is played and feedback (income or loss) is given to the user. Usually, there is no specific requirement for the setting of reward amount and loss amount. During the 50 rounds of selection, the reward and punishment ratio is set to 50%, and the opening result of each round is randomly set. After 50 rounds, the user is shown the overall game profit and loss. Before the experiment, ensure that the portable single-channel EEG head ring device is operating normally and test the signal quality. During the experiment, the EEG signal of the user when completing the task is recorded in real time, and the experiment duration is controlled to be about 20 minutes.

[0094] The data processing module includes signal filtering, signal noise reduction, signal quality screening, signal analysis and model input processing. By including high-quality task-state electroencephalogram signals and tagging the user population that obtains the data, including healthy people, depression patients and bipolar disorder patients, the model training data and model verification data are arranged in a certain proportion.

[0095] In one embodiment, the data acquisition process is as shown in Figure 5 Resting state with eyes open data acquisition: In the resting state with eyes open data acquisition, the user is required to keep quiet and look straight at the fixed point on the screen or maintain natural vision. At this time, the environment should minimize interference, including shielding external noise and reducing the impact of light changes, to ensure that the acquired electroencephalogram signals are stable and reliable. The portable single-channel electroencephalogram head ring is correctly worn on the head to ensure good contact between the electrode and the skin, and the signal quality is confirmed by the device software to meet the acquisition standard. After the experiment starts, the device will continuously record the user's electroencephalogram signals, and the acquisition time is 5 minutes, which is enough to capture stable resting state signals.

[0096] Resting state with eyes closed data acquisition: In the resting state with eyes closed data acquisition, the user is required to close his eyes, keep relaxed and minimize facial and body muscle activity to acquire pure closed-eye electroencephalogram signals. The acquisition environment should be quiet and the light should be moderate to avoid external sound and light stimuli affecting the user's relaxed state. After the experiment starts, the device begins to record the user's electroencephalogram signals when the eyes are closed, and the acquisition time is 3-5 minutes.

[0097] Task state (reward and punishment task) data acquisition: The user uses an interactive tablet to complete 50 rounds of blind box opening game, and ensures that the portable single-channel electroencephalogram head ring device is running normally before the experiment and tests the signal quality. Real-time recording of the user's electroencephalogram signals before each round of blind box opening for 3 seconds and the 5-second data obtained after opening the blind box is performed during the experiment, focusing on the neural activity before and after decision-making, and the experiment duration is controlled at about 10 minutes. The user's behavioral data is recorded for combined analysis with the electroencephalogram data.

[0098] After completing the resting state and task state data acquisition, all electroencephalogram signals are arranged and stored as standardized data files. The data files use common electroencephalogram data formats (such as EDF, CSV or MATLAB format) for subsequent analysis and processing. Each file contains complete time series signals, experiment condition labels and user information (such as number and group category). Each electroencephalogram data is labeled with experiment state labels, including open eyes, closed eyes, task state, etc.; experiment time, device number and environmental parameters are recorded as metadata to ensure data traceability.

[0099] In one embodiment, the data processing module is an important step to ensure the effectiveness and reliability of the EEG signal, which mainly includes the following contents:

[0100] Signal filtering: 50Hz power frequency interference is removed by using a notch filter, and a band-pass filter (usually 0.5-40 Hz) is used to remove high-frequency noise (such as electromyographic interference) and low-frequency drift (such as electrode noise), and to retain the signal frequency band related to cognitive activity.

[0101] y[n]=F -1 (F(x[n])·H(f))

[0102] Where F is the Fourier transform, H(f) is the frequency response of the filter, x[n] is the original signal, and y[n] is the filtered signal.

[0103] Remove motion artifacts: remove continuous motion artifact data by time domain threshold, mark the continuous data area greater than the threshold as false data, and remove false data in the filtered data. In this case, the EEG threshold is -200 μV ~ +200 μV.

[0104] Because single-channel EEG signals cannot achieve correlation noise reduction similar to double-channel signals, we use a sliding window combined with quartile statistical methods to judge the quality of the collected EEG signals in the signal quality screening, and exclude user data with poor signal quality in modeling to avoid the interference of eye movement, body movement and other noise signals on modeling.

[0105] Statistical resting state, task state all data segment amplitude, get the mean μ and variance std of the amplitude, and define the upper and lower limits as [μ-1.5*std, μ+1.5*std]. A specific moving window width (typical value is half of the sampling frequency) and a specific sliding step (typical value is 1) are used to count the number of data points outside the amplitude upper and lower limits in each window. These points are defined as signal non-effective values (such as Figure 6 The number of non-effective values is statistically analyzed according to the natural distribution, and the upper quartile based on the distribution is set as the quality evaluation threshold Ratingthred (such as Figure 7 The number of non-effective values on the left side of the red threshold line is considered as acceptable signal quality and is included in the classification training model.

[0106] In one embodiment, two-dimensional processing is to take single-person single-time task-state EEG signals as columns or rows, a total of S times, to form a two-dimensional matrix (two-dimensional EEG signals), S is a natural number greater than or equal to 50, and the time dimension is taken as a row or a column. By converting one-dimensional time series signal data into two-dimensional data for depression and bipolar classification model classification, the classification accuracy of the model can be improved.

[0107] In one specific embodiment, the benefits of one-dimensional to two-dimensional are:

[0108] 1) Improve statistical power: By multiple measurements, the amount of data can be increased, the random error can be reduced, and the sensitivity and reliability of statistical analysis can be improved.

[0109] 2) Evaluate intra-individual changes: The dynamic changes of the same subject at different time points can be tracked, and the interference of inter-individual differences can be reduced.

[0110] 3) Reduce sample size requirements: Compared with cross-sectional studies, repeated measurements can use multi-time point data of the same subjects, reducing the dependence on large-scale samples.

[0111] 4) Enhance the robustness of the results: Through multiple observations, the stability of the results can be verified, and accidental deviations of single measurement can be avoided.

[0112] S103: input the two-dimensional electroencephalogram signal into the depression and bipolar classification model for classification to obtain the classification results of health, depression, and bipolar.

[0113] In one embodiment, the training process of the depression and bipolar classification model is:

[0114] Obtain a one-dimensional time series electroencephalogram signal dataset and labels of open eyes, closed eyes, and games;

[0115] The one-dimensional time series electroencephalogram signal and the label are two-dimensionally processed to obtain a two-dimensional electroencephalogram signal;

[0116] The two-dimensional electroencephalogram signal is input into the classifier to be trained for training until the loss function is constant, and a depression and bipolar classification model is obtained;

[0117] Optionally, the two-dimensional processing is based on converting different time dimension one-dimensional time series electroencephalogram signals into a two-dimensional matrix to obtain a two-dimensional electroencephalogram signal; wherein the rows or columns of the matrix represent time, and the columns or rows of the matrix represent one-dimensional time series electroencephalogram signals.

[0118] In one embodiment, the classifier includes one or more of the following: convolutional neural network, residual network, extreme learning machine, perception vector machine, random forest, decision tree, support vector machine, AdaBoost, and Transformer.

[0119] In one embodiment, the method further comprises secondary prediction, the depression and bipolar classification model obtains a category probability after classification, basic information of the subject to be measured is obtained, and the category probability and the basic information are used for secondary prediction to obtain the classification prediction results of health, depression, and bipolar.

[0120] In one embodiment, the secondary prediction is performed by one or more of the following models: logistic regression, naive Bayes, AdaBoost, GBM, and LDA.

[0121] In one embodiment, the basic information includes one or more of the following: age, gender, medical history, symptom condition.

[0122] In one specific embodiment, in the EEG study of differentiating patients with depression, bipolar disorder and healthy people in this study, repeated measurements can identify the stability or dynamic changes of EEG characteristics (such as theta, alpha, gamma bands), which can assist in differential diagnosis.

[0123] In one specific embodiment, a deep learning algorithm based on a two-dimensional convolutional neural network is used for modeling, and a trained model file is generated to support the subsequent prediction stage. In the model prediction process, the new EEG data is first preprocessed, and then the preprocessed data is input into the trained model for prediction. After the model outputs the classification result, a comprehensive decision is made with the patient's basic information, and the final recognition result is obtained to help judge the mental illness status of the individual. The system outputs the test report of the user to assist the clinic in early depression screening and help patients receive timely treatment.

[0124] In one specific embodiment, the algorithm modeling uses a two-dimensional convolutional neural network to process the time series EEG signals before and after 50 rounds of unboxing, achieving accurate identification of healthy people and people with depression. The model structure is as shown in Figure 8 The model structure is as shown in

[0125] Input layer: The input layer accepts two-dimensional EEG signal data, and the data consists of 8-second length EEG data during the 50 rounds of unblinded box experiments of the subjects. The head ring data sampling rate is 160 Hz, so the amount of input data is 50* (8*160) or 50*1280 data points.

[0126] Convolution layer: The convolution layer is the core of the convolutional neural network, which extracts features from the input data through convolution operations. Each convolution layer contains multiple convolution kernels (filters), which slide (i.e., convolution) through different parts of the input data to capture local patterns. The size of the convolution kernel is usually k x 1, such as 3 or 5 (indicating that 3 or 5 time steps are covered by each convolution operation). The convolution layer outputs a feature map, usually a combination of multiple feature maps, each corresponding to a convolution kernel. The convolution operation formula is as follows:

[0127]

[0128] Where: is the input signal sequence; is the convolution kernel; is the convolution output.

[0129] Activation Layer: Activation layer is usually followed by the convolution layer, and the commonly used activation function is ReLU (Rectified Linear Unit). ReLU can effectively introduce nonlinearity, so that the network can learn complex patterns. The ReLU function is as follows:

[0130]

[0131] where negative values will be suppressed to zero, and positive values remain unchanged.

[0132] Pooling Layer: Pooling layer is a kind of operation in convolutional neural network used to reduce the size of feature map, reduce the computational complexity, avoid overfitting and extract the most important features. Pooling layer is usually located after the convolution layer, which simplifies the model by sampling the features in the local area. There are many ways of pooling, the most common of which are max pooling and average pooling. Max pooling layer, in each pooling window, max pooling selects the maximum value in the window as the output of the window. Max pooling can preserve the most significant feature information, and is especially suitable for tasks that contain strong features. Average pooling layer, the output is obtained by calculating the average of all values in the pooling window. Average pooling is more smooth, which helps to extract the overall trend in the region rather than focusing on the local strongest features.

[0133] Flattening Layer: The output of convolutional and pooling layers is usually multi-dimensional, while fully connected layers require one-dimensional input. Therefore, the output of convolutional and pooling layers needs to be flattened into a one-dimensional vector.

[0134] Fully Connected Layer: Fully connected layer weights and sums the features extracted before, which is usually used to integrate the extracted features and finally generate the output. The neurons of fully connected layer are connected to all neurons of the previous layer. The fully connected formula is as follows:

[0135]

[0136] where, is the weight matrix, is the input vector, is the bias.

[0137] Output Layer: The output layer is the last layer of the network, mainly used to generate the prediction result. For classification tasks, the output layer activation function is used for classification, which converts the model output into class probability.

[0138] Softmax activation function (multi-classification):

[0139]

[0140] In multiple logistic regression and linear discriminant analysis, the input of the function is the result obtained from K different linear functions, and the probability of the sample vector x belonging to the jth category is P(y = j).

[0141] After the model is defined, the model needs to be compiled, which includes selecting the loss function, optimizer and evaluation metric. For classification tasks, the cross-entropy loss is used, the optimizer can use Adam, and the evaluation metric is accuracy.

[0142] After the model is compiled, the model is trained, which is a process of optimizing the model by inputting data and labels. During the training process, the network updates the weights through backpropagation, so that the model can better fit the data. Training includes specifying training data, batch size, training rounds, and validation data parameters.

[0143] By training the accuracy (as shown in Figure 9 ) and loss rate (as shown in Figure 10 ), the optimal parameter model for classifying healthy people and people with depression is determined.

[0144] In addition, whether the model has obtained better results in learning the brain electrical data characteristics of healthy people, bipolar disorder patients and depression patients in the reward and punishment task state, the application also uses the T-SNE method to visualize and verify the feature quantity (as shown in Figure 11 ). From the vector of the feature space, the features of depression patients, bipolar disorder patients and healthy people have been well distinguished, indicating the effectiveness of the model.

[0145] In one embodiment, the application uses 587 sets of data (287 healthy people, 242 depression patients, and 58 bipolar disorder patients) as model training data, and then tests and verifies 254 sets of data (261 healthy people, 590 depression patients, and 159 bipolar disorder patients). It is found that the overall recognition accuracy of the model is 96.7%, of which the recognition accuracy of healthy people is 97.0%, the recognition accuracy of depression patients is 94.6%, and the recognition accuracy of bipolar disorder patients is 98.5%, reaching the best level of depression screening based on EEG signals in a large population reported so far.

[0146] In one embodiment, after the model training is completed, the optimal parameters obtained by training are saved as a model file, which mainly includes the following contents:

[0147] File format: Use standardized model storage format (such as HDF5 or ONNX) for cross-platform application.

[0148] Storage content: Network structure description; trained weight and bias parameters; input feature standardization parameters.

[0149] Version management: Store the model file by version, and record the hyperparameter settings and performance indicators of each training, which is convenient for subsequent improvement and reproduction.

[0150] In one specific embodiment, after the signal is preprocessed, each frame of the preprocessed signal is input into the trained deep learning model for prediction. The specific steps are as follows:

[0151] Input data: The framed EEG signal data is used as input, and the data size is the same as the data size used during training, which is a 50*1280 matrix.

[0152] Output result: The result output by the model gives the probability of belonging to a certain category (healthy population, bipolar disorder population, depression population), combined with the patient's basic information, including age, gender, medical history, and symptom condition, using a logistic regression enhancement model to form the final health or disease state prediction result.

[0153] Patient basic information (BI: Basic Information): Age, gender, medical history, and symptom condition;

[0154] Model diagnosis probability (P m : Probability of the model): Disease probability from model prediction;

[0155] log(P final ) = + + β•log(P m )

[0156] Where: is the weight coefficient of the patient's basic information, and β is the adjustment coefficient of the model classification probability.

[0157] The final probability is obtained using the sigmoid function:

[0158] P final =

[0159] In one specific embodiment, the report output module: the final diagnosis result is presented in text form, or displayed to the doctor, patient or other relevant personnel through a visual interface, the report can be printed, stored, and the report content includes the opening of the blind box task description, the resting state EEG analysis situation, the task state EEG analysis situation, and gives the analysis logic of the system diagnosis as healthy, bipolar disorder or depression.

[0160] In one embodiment, the present application obtains single-channel electroencephalogram signals through a single-channel electroencephalogram signal device, the electroencephalogram signals are electroencephalogram signals obtained during the execution of open-eye, closed-eye and open-box game, the electroencephalogram signals are input into a trained depression and bipolar classification model to obtain classification results of depression, bipolar and health, wherein the electroencephalogram signals before opening, after opening and / or after the total profit and loss of the game are extracted for classification, in addition, for single signal channel, signal quality judgment is proposed for signal screening to improve signal quality, the present application provides a whole scheme, from single-channel signal to the final classification result, there is data interaction between each module, and the health or abnormality (depression or bipolar) of the to-be-tested person is detected.

[0161] The present application discloses an embodiment of a computer program product or system, comprising a computer program, which, when executed by a processor, implements the above-mentioned depression and bipolar classification method based on single-channel electroencephalogram signals.

[0162] Figure 2 The present application provides a depression and bipolar classification system based on single-channel electroencephalogram signals, which specifically comprises:

[0163] The fixed unit fixes the single-channel electroencephalogram signal acquisition device on the frontal lobe of the to-be-tested person;

[0164] The acquisition unit acquires the electroencephalogram signals of the single-channel electroencephalogram signal acquisition device when the to-be-tested person performs the task action; and converts the one-dimensional time sequence electroencephalogram signals into two-dimensional electroencephalogram signals based on different time dimensions;

[0165] The classification unit inputs the electroencephalogram signals into a depression and bipolar classification model to obtain classification results of health, depression and bipolar.

[0166] In one embodiment, the system further comprises a data processing unit for data preprocessing of the electroencephalogram signals to obtain preprocessed electroencephalogram signals.

[0167] Figure 3 The present application provides a depression and bipolar classification device based on single-channel electroencephalogram signals, which specifically comprises:

[0168] The memory is used for storing program instructions, and the processor is used for calling the program instructions, when the program instructions are executed, any one of the above-mentioned depression and bipolar classification methods based on single-channel electroencephalogram signals.

[0169] The present application discloses an embodiment of a computer readable storage medium, which stores a computer program, when the computer program is executed by a processor, any one of the above-mentioned depression and bipolar classification methods based on single-channel electroencephalogram signals.

[0170] The verification result of the verification embodiment shows that assigning inherent weights to the indications can improve the performance of the method compared with the default setting. It can be clearly understood by those skilled in the art that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the foregoing method embodiments, which will not be described here. In the several embodiments provided by the present application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. The division of the units is only a logical function division. There can be another division manner in actual implementation. For example, a plurality of units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the coupling or direct coupling or communication connection between the units shown or discussed can be indirect coupling or communication connection through some interfaces, devices or units, and can be electrical, mechanical or other forms. The units described as separate components can or can not be physically separate, and the components displayed as units can or can not be physical units, i.e. can be located in one place, or can be distributed on a plurality of network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the embodiment scheme. In addition, each functional unit in the various embodiments of the present application can be integrated in one processing unit, or each unit can be a physically independent unit, or two or more units can be integrated in one unit. The integrated unit can be implemented in the form of hardware, or in the form of software functional units. Those skilled in the art can understand that all or part of the steps of the various methods in the above embodiments can be completed by programs instructing relevant hardware, and the programs can be stored in a computer readable storage medium, which can include read only memory (ROM), random access memory (RAM), magnetic disk or optical disk, etc.

[0171] Those skilled in the art can understand that all or part of the steps of the above-mentioned embodiment methods can be completed by programs instructing relevant hardware, and the programs can be stored in a computer readable storage medium, and the above-mentioned medium storage can be read only memory, magnetic disk or optical disk, etc.

[0172] The computer device provided by the present application has been described in detail above. For those skilled in the art, according to the idea of the embodiment of the present application, there will be changes in specific implementation and application range. In view of the above, the content of the specification should not be understood as a limitation of the present application.

Claims

1. A method for classifying depression and bipolar based on single-channel electroencephalogram signals, characterized in that, A single-channel electroencephalogram signal acquisition device, comprising: Fixing the single-channel electroencephalogram signal acquisition device on the frontal lobe of the brain of the subject to be tested; Obtaining one-dimensional time sequence electroencephalogram signals of the single-channel electroencephalogram signal acquisition device when the subject to be tested performs a task action; Converting the one-dimensional time sequence electroencephalogram signals into two-dimensional electroencephalogram signals based on different time dimensions; the method further comprises data preprocessing, obtaining preprocessed one-dimensional time sequence electroencephalogram signals by preprocessing the one-dimensional time sequence electroencephalogram signals, and then converting the preprocessed one-dimensional time sequence electroencephalogram signals into two-dimensional electroencephalogram signals based on different time dimensions; the data preprocessing comprises signal quality screening; the signal quality screening is a quality judgment of the electroencephalogram signals, and the preprocessed one-dimensional time sequence electroencephalogram signals are obtained by removing signals with poor quality; Converting the task state one-dimensional time sequence electroencephalogram signals of the tester s times into two-dimensional matrices based on different time dimensions to obtain two-dimensional electroencephalogram signals; wherein the rows or lists of the matrix represent time, and the columns or rows of the matrix represent the execution of single task state one-dimensional time sequence electroencephalogram signals; the task action comprises an open-eye task, a closed-eye task, and a game task; the game task is to open a blind box by selection, the box gives a virtual currency income or loss animation, after M rounds, the overall game profit and loss is displayed, and M is a natural number greater than or equal to 50; The subject to be tested performs the open-eye task, the closed-eye task, and the game task in turn and obtains one-dimensional time sequence electroencephalogram signals of each task action, and converts the one-dimensional time sequence electroencephalogram signals of each task action into two-dimensional electroencephalogram signals based on different time dimensions; The game electroencephalogram signal further comprises signal extraction, extraction of electroencephalogram signals before opening the box for S seconds and electroencephalogram signals after opening the box for L seconds to obtain extracted game electroencephalogram signals, and input of the open-eye electroencephalogram signals, the closed-eye electroencephalogram signals, and the extracted game electroencephalogram signals into a depression and bipolar classification model to obtain a classification result of health, depression, and bipolar; S and L are natural numbers greater than 1.

2. The single-channel electroencephalogram signal-based depression bipolar classification method according to claim 1, characterized in that, When the subject to be tested performs the open-eye task, open-eye electroencephalogram signals for N minutes are obtained, N is a natural number greater than or equal to 5, and the N-minute closed-eye electroencephalogram signals are converted into two-dimensional open-eye electroencephalogram signals based on different time dimensions. 3.The method of claim 1, wherein, When the subject to be tested performs the closed-eye task, closed-eye electroencephalogram signals for N minutes are obtained, and the N-minute closed-eye electroencephalogram signals are converted into two-dimensional closed-eye electroencephalogram signals based on different time dimensions. 4.The method of claim 1, wherein, When the subject to be tested performs the game task, game electroencephalogram signals for 2N minutes are obtained, and the 2N-minute game electroencephalogram signals are converted into two-dimensional game electroencephalogram signals based on different time dimensions.

5. The single-channel electroencephalogram signal-based depression bipolar classification method according to claim 1, characterized in that, The game electroencephalogram signal extracts electroencephalogram signals for 3 seconds before opening the box and electroencephalogram signals for 5 seconds after opening the box.

6. The single-channel electroencephalogram signal-based depression bipolar classification method according to claim 1, characterized in that, The game electroencephalogram signal extraction further comprises electroencephalogram signals after the overall game profit and loss is displayed.

7. The single-channel electroencephalogram signal-based depression bipolar classification method according to claim 1, characterized in that, The data preprocessing comprises one or more of the following: signal filtering, removing motion artifacts, and signal quality screening. 8.The method of claim 1, wherein, The preprocessing is performed in the order of signal filtering, removing motion artifacts, and signal quality screening. 9.The method of claim 1, wherein, The quality judgment is performed by a sliding window and quartile statistics.

10. The single-channel electroencephalogram signal-based depression bipolar classification method according to claim 1, characterized in that, The quality judgment process is as follows: Statistical data amplitude of the task action; Calculating the mean and variance of the data amplitude and formulating the upper and lower limits of the data amplitude based on the mean and variance; The signal non-effective value is compared with a preset threshold value, and when the signal non-effective value is greater than the preset threshold value, the signal non-effective value is removed; otherwise, the signal non-effective value is retained. The quality judgment process further includes statistical distribution, and a statistical distribution value is obtained by performing statistical distribution on the signal non-effective value, and the preset threshold value is obtained based on the statistical distribution value.

11. The single-channel electroencephalogram signal based depression bipolar classification method according to claim 10, characterized in that, The preset threshold value is an upper quarter of a one-bit value of the signal non-effective value after statistical distribution.

12. The single-channel electroencephalogram signal-based depression bipolar classification method according to claim 11, characterized in that, The window width of the moving window is half of the sampling frequency.

13. The single-channel electroencephalogram signal based depression bipolar classification method according to claim 10, wherein, The training process of the depression and bipolar classification model is as follows:

14. The single-channel electroencephalogram signal-based depression bipolar classification method according to claim 1, characterized in that, A one-dimensional time series electroencephalogram signal data set and labels of open eyes, closed eyes, and games are obtained. The one-dimensional time series electroencephalogram signal and the labels are processed in two dimensions to obtain a two-dimensional electroencephalogram signal. The two-dimensional electroencephalogram signal is input into a classifier to be trained for training until the loss function is constant, and a depression and bipolar classification model is obtained. The classifier includes one or more of the following: convolutional neural network, residual network, extreme learning machine, perception vector machine, random forest, decision tree, support vector machine, AdaBoost, and Transformer.

15. The single-channel electroencephalogram signal based depression bipolar classification method according to claim 14, characterized in that, The method further includes secondary prediction, and a class probability is obtained after classification by the depression and bipolar classification model, basic information of a subject to be measured is obtained, and a health, depression, and bipolar classification prediction result is obtained based on the class probability and the basic information.

16. The single-channel electroencephalogram signal-based depression bipolar classification method according to claim 1, wherein, The secondary prediction is performed by one or more of the following models: logistic regression, naive Bayes, AdaBoost, GBM, and LDA.

17. The single-channel electroencephalogram signal based depression bipolar classification method according to claim 16, characterized in that, The basic information includes one or more of the following: age, gender, medical history, and symptom condition.

18. The single-channel electroencephalogram signal based depression bipolar classification method according to claim 16, wherein, The computer program or instructions are executed by a processor to implement the depression and bipolar classification method based on single-channel electroencephalogram signals according to any one of claims 1-18.

19. A computer program product comprising a computer program or instructions embodied therein, characterized in that, The computer program or instructions are executed by a processor to implement the depression and bipolar classification method based on single-channel electroencephalogram signals according to any one of claims 1-18.

20. A computer device comprising a memory, a processor, and a computer program or instructions stored on the memory, wherein, The computer program or instructions are executed by a processor to implement the depression and bipolar classification method based on single-channel electroencephalogram signals according to any one of claims 1-18.

21. A computer readable storage medium having stored thereon a computer program or instructions, characterized in that, ​

Images