Depression probability determination method, depression probability determination system and related device

Abstract

The embodiment of the invention discloses a depression probability determination method and system and a related device.The method comprises the steps that data sets of N window numbers corresponding to target electroencephalogram signal data of a patient to be detected are obtained through a preset filter, the data sets are input into a trained prediction model, prediction tags are correspondingly obtained, and the prediction tags are used for prediction of the target electroencephalogram signal data of the patient to be detected; according to the N prediction labels, determining the probability of depression of the to-be-detected patient, and taking denoising generation data output by training a preset conditional diffusion model based on the electroencephalogram signal data and the number information of the multiple subjects and the tag information about whether the subjects suffer from depression or not and the initial electroencephalogram signal data of the multiple subjects as a training set of a prediction model. According to the method, the number information and the label information are used as features to be embedded into the diffusion process, so that individual features and depression features of the patient can be better introduced, the condition diffusion model is denoised to generate electroencephalogram signals with more individual differences, and the accuracy of the prediction probability value of the prediction model is improved.

Description

Technical Field

[0001] The present invention relates to the field of model prediction technology, and in particular to a method, system and related device for determining the probability of depression. Background Art

[0002] The existing diagnosis of depression can be carried out through questionnaires, which is cumbersome and time-consuming, and the user's subjective answers are inaccurate, which ultimately affects the accuracy of the diagnosis of depression. Secondly, a classification prediction model can be used for diagnosis, but the existing EEG clinical data on depression is relatively scarce, and the models often have problems with poor generalization ability and low accuracy.

[0003] Therefore, a prediction method is urgently needed to improve the accuracy of depression diagnosis. Summary of the invention

[0004] The main purpose of the present invention is to provide a method, system, device, computer equipment and storage medium for determining the probability of depression, which can solve the problem of low probability prediction accuracy in the prior art.

[0005] To achieve the above object, the present invention provides a method for determining the probability of depression in a first aspect, the method comprising:

[0006] Acquire the EEG signal data of the patient to be tested to obtain target EEG signal data, and use a preset filter to filter the target EEG signal data to obtain a data set of N windows after filtering;

[0007] Inputting N data sets into a preset prediction model respectively, and obtaining N corresponding prediction labels, wherein the prediction model is a model obtained after training with a target training set, and the target training set is a collection of a first training set and a second training set, wherein the first training set includes initial EEG signal data of multiple subjects, and the second training set includes denoised generated data outputted by a conditional diffusion model based on the initial EEG signal data of multiple subjects, first information, and second information, wherein the first information is the subject's number information, and the second information is label information indicating whether the subject suffers from depression;

[0008] The number of target labels in the predicted labels is obtained, and the proportion of the number of target labels in the N predicted labels is determined as the probability that the patient to be tested suffers from depression; wherein the target label is a label that indicates depression.

[0009] In combination with the first aspect, in a possible implementation method, denoised generated data is output based on initial EEG signal data, first information and second information of multiple subjects through a conditional diffusion model, including: obtaining initial EEG signal data of multiple subjects, performing global average reference on the initial EEG signal data to obtain reference EEG signal data, filtering the reference EEG signal data using a preset filter to obtain a filtered initial training set; inputting the initial training set into the conditional diffusion model, forward denoising the initial training set to obtain forward denoised data, and inputting the first information and the second information into the conditional diffusion model to predict the noise of the forward denoised data by training the conditional diffusion model to obtain a trained conditional diffusion model; and inputting random noise, the first information and the second information of the subject into the trained conditional diffusion model, reverse denoising the random noise to obtain denoised generated data.

[0010] In combination with the first aspect, in a possible implementation method, the above-mentioned conditional diffusion model includes a first convolutional layer, a time step embedding layer and a feature embedding layer, and the first information and the second information are input into the conditional diffusion model to train the conditional diffusion model to predict the noise of the forward noisy data to obtain the trained conditional diffusion model, including: inputting the forward noisy data into the first convolutional layer to obtain first noisy data, and using the SiLU activation function to process the first noisy data to obtain second noisy data; inputting the first information and the second information into the feature embedding layer to obtain the corresponding first feature and second feature, and inputting the time step into the time step embedding layer to obtain the corresponding third feature; based on the second noisy data, the first feature, the second feature and the third feature, the conditional diffusion model is trained to predict the noise of the forward noisy data to obtain the trained conditional diffusion model.

[0011] In combination with the first aspect, in a possible implementation, the conditional diffusion model further includes a linear layer and a dilated convolution layer, and the conditional diffusion model is trained based on the second noisy data, the first feature, the second feature, and the third feature to predict the noise of the forward noisy data to obtain the trained conditional diffusion model, including: performing matrix element-by-element addition operation on the first feature, the second feature, and the third feature to obtain a first matrix, inputting the first matrix into the linear layer for processing to obtain a corresponding second matrix, performing matrix element-by-element addition operation on the second matrix and the second noisy data to obtain a third matrix; inputting the third matrix into the dilated convolution layer to obtain a fourth matrix; performing convolution operation on the fourth matrix to obtain a feature matrix and third noisy data; using the third noisy data as the second noisy data, and returning to perform the step of performing matrix element-by-element addition operation on the second matrix and the second noisy data to obtain the third matrix, until the cycle is repeated M times; training the conditional diffusion model based on M feature matrices to predict the noise of the forward noisy data to obtain the trained conditional diffusion model.

[0012] In combination with the first aspect, in a possible implementation, the conditional diffusion model further includes a second convolution layer and a third convolution layer, and the convolution processing of the fourth matrix to obtain a feature matrix and third noisy data includes: performing a nonlinear transformation on the fourth matrix through a hyperbolic tangent function and a Sigmoid activation function respectively, to obtain a fifth matrix and a sixth matrix correspondingly; performing a matrix element-by-element multiplication operation on the fifth matrix and the sixth matrix to obtain a seventh matrix, and inputting the seventh matrix into the second convolution layer to obtain a feature matrix; and inputting the seventh matrix into the third convolution layer to obtain third noisy data.

[0013] In combination with the first aspect, in a possible implementation method, the above-mentioned conditional diffusion model also includes a fourth convolutional layer and a fifth convolutional layer, and the conditional diffusion model is trained based on M feature matrices to predict the noise of the forward noisy data to obtain the trained conditional diffusion model, including: performing matrix element-by-element addition operation on the obtained M feature matrices and a preset initial feature matrix to obtain a target feature matrix, and inputting the target feature matrix into the fourth convolutional layer to obtain fourth noisy data; using a SiLU activation function to calculate the fourth noisy data to obtain fifth noisy data, and inputting the fifth noisy data into the fifth convolutional layer to obtain predicted noise data, and training the conditional diffusion model based on the predicted noise data until convergence to obtain the trained conditional diffusion model.

[0014] To achieve the above object, the second aspect of the present invention provides a depression probability determination system, the system comprising an EEG signal acquisition device, a communication module, a host computer and a server;

[0015] The host computer is connected to the EEG signal acquisition device and is used to send a control signal representing the EEG signal of the patient to be tested to the EEG signal acquisition device; the EEG signal acquisition device is used to receive the control signal and collect EEG signal data of the patient to be tested;

[0016] The host computer and the EEG signal acquisition device are respectively connected to a communication module, and the EEG signal acquisition device is also used to send the acquired target EEG signal data to the host computer through the communication module;

[0017] The host computer is also used to receive target EEG signal data and send the target EEG signal data to the server;

[0018] The server is used to receive target EEG signal data, and filter the target EEG signal data with a preset filter to obtain a data set of N windows after filtering; the N data sets are respectively input into a preset prediction model to obtain N corresponding prediction labels, wherein the prediction model is a model obtained after training with a target training set, and the target training set is a collection of a first training set and a second training set, the first training set includes initial EEG signal data of multiple subjects, and the second training set includes denoising generated data output by a conditional diffusion model based on the initial EEG signal data of multiple subjects, first information and second information, the first information is the subject's number information, and the second information is label information indicating whether the subject suffers from depression; the number of target labels in the prediction label is obtained, and the proportion of the number of target labels in the N prediction labels is determined as the probability that the patient to be tested suffers from depression; wherein the target label is a label representing depression.

[0019] To achieve the above object, the present invention provides a third aspect of a depression probability determination device, the device comprising:

[0020] Acquisition module: used to acquire the EEG signal data of the patient to be tested, obtain target EEG signal data, and filter the target EEG signal data using a preset filter to obtain a data set of N windows after filtering;

[0021] Label prediction module: used to input N data sets into preset prediction models respectively, and obtain N corresponding prediction labels, wherein the prediction model is a model obtained after training with a target training set, and the target training set is a collection of a first training set and a second training set, wherein the first training set includes initial EEG signal data of multiple subjects, and the second training set includes denoised generated data output by a conditional diffusion model based on the initial EEG signal data of multiple subjects, first information, and second information, wherein the first information is the subject's number information, and the second information is label information indicating whether the subject suffers from depression;

[0022] Probability prediction module: used to obtain the number of target labels in the predicted labels, and determine the proportion of the number of target labels in N predicted labels as the probability that the patient to be tested suffers from depression; wherein the target label is a label that represents depression.

[0023] To achieve the above object, the fourth aspect of the present invention provides a computer-readable storage medium storing a computer program, wherein when the computer program is executed by a processor, the processor performs the following steps:

[0024] Acquire the EEG signal data of the patient to be tested to obtain target EEG signal data, and use a preset filter to filter the target EEG signal data to obtain a data set of N windows after filtering;

[0025] Inputting N data sets into a preset prediction model respectively, and obtaining N corresponding prediction labels, wherein the prediction model is a model obtained after training with a target training set, and the target training set is a collection of a first training set and a second training set, wherein the first training set includes initial EEG signal data of multiple subjects, and the second training set includes denoised generated data outputted by a conditional diffusion model based on the initial EEG signal data of multiple subjects, first information, and second information, wherein the first information is the subject's number information, and the second information is label information indicating whether the subject suffers from depression;

[0026] The number of target labels in the predicted labels is obtained, and the proportion of the number of target labels in the N predicted labels is determined as the probability that the patient to be tested suffers from depression; wherein the target label is a label that indicates depression.

[0027] To achieve the above object, a fifth aspect of the present invention provides a computer device, comprising a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the processor performs the following steps:

[0028] Acquire the EEG signal data of the patient to be tested to obtain target EEG signal data, and use a preset filter to filter the target EEG signal data to obtain a data set of N windows after filtering;

[0029] Inputting N data sets into a preset prediction model respectively, and obtaining N corresponding prediction labels, wherein the prediction model is a model obtained after training with a target training set, and the target training set is a collection of a first training set and a second training set, wherein the first training set includes initial EEG signal data of multiple subjects, and the second training set includes denoised generated data outputted by a conditional diffusion model based on the initial EEG signal data of multiple subjects, first information, and second information, wherein the first information is the subject's number information, and the second information is label information indicating whether the subject suffers from depression;

[0030] The number of target labels in the predicted labels is obtained, and the proportion of the number of target labels in the N predicted labels is determined as the probability that the patient to be tested suffers from depression; wherein the target label is a label that indicates depression.

[0031] The embodiments of the present invention have the following beneficial effects:

[0032] The present invention provides a method for determining the probability of depression. By embedding first information and second information of multiple subjects as features into the diffusion process of a conditional diffusion model, individual characteristics of patients and characteristics of depression can be better introduced, so that the conditional diffusion model can denoise and generate electroencephalogram signals with more individual differences. The denoised electroencephalogram signals and a set of initial electroencephalogram signal data of multiple subjects are used as training sets of a prediction model, which can improve the accuracy of the prediction probability value of the prediction model. In addition, by using a data set of N windows after filtering as the input of the prediction model, the overall target brain point signal data can be divided into more detailed data to obtain more accurate prediction labels, thereby further improving the accuracy of the prediction probability value. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0034] in:

[0035] Figure 1 A schematic diagram of a method for determining the probability of depression in an embodiment of the present invention;

[0036] Figure 2 This is a schematic diagram of the conditional diffusion model training process in an embodiment of the present invention;

[0037] Figure 3 Schematic diagram of a reverse denoising process in an embodiment of the present invention;

[0038] Figure 4 A schematic diagram of a prediction model in an embodiment of the present invention;

[0039] Figure 5 This is a structural block diagram of a device for determining the probability of depression in an embodiment of the present invention;

[0040] Figure 6 This is a structural block diagram of a depression probability determination system in an embodiment of the present invention;

[0041] Figure 7 4 is a structural block diagram of a computer device in an embodiment of the present invention. DETAILED DESCRIPTION

[0042] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0043] The embodiment of the present invention provides a method for determining the probability of depression, which is suitable for determining the probability of a patient suffering from depression. Figure 1 , Figure 1 A flowchart of a method for determining the probability of depression provided by an embodiment of the present invention is shown in FIG. Figure 1 As shown, the specific steps of the method include the following:

[0044] Step S101, acquiring EEG signal data of a patient to be tested, obtaining target EEG signal data, and filtering the target EEG signal data using a preset filter to obtain a data set of N windows after filtering.

[0045] Step S102: input N data sets into the preset prediction model respectively, and obtain N corresponding prediction labels.

[0046] Among them, the prediction model is a model obtained after training with a target training set, the target training set is a collection of a first training set and a second training set, the first training set includes initial EEG signal data of multiple subjects, and the second training set includes denoised generated data output by a conditional diffusion model based on initial EEG signal data of multiple subjects, first information, and second information, the first information is the subject's number information, and the second information is label information indicating whether the subject suffers from depression.

[0047] Step S103: Obtain the number of target labels in the predicted labels, and determine the proportion of the number of target labels in the N predicted labels as the probability that the patient to be tested suffers from depression.

[0048] Among them, the target label is a label representing depression.

[0049] In this embodiment, first, the preset conditional diffusion model is trained to obtain a conditional diffusion model for denoising data, wherein the conditional diffusion model is specifically divided into a forward denoising process and a reverse denoising process. In the forward denoising process, for each observed sample, a small amount of noise is continuously added to the sample until the sample is completely denoised to become completely random noise, which belongs to a normal distribution. The reverse denoising process is to randomly sample a noise sample from the normal distribution, and continuously remove a small amount of noise from the sample until the noise sample becomes a generated sample like a real sample.

[0050] In this embodiment, the conditional diffusion model is trained to obtain a trained conditional diffusion model, and denoised generated data is obtained based on the conditional diffusion model. Figure 2 , Figure 2 A schematic diagram of the conditional diffusion model training process provided by an embodiment of the present invention, such as Figure 2 As shown, the steps for training the preset conditional diffusion model are as follows:

[0051] Step S201, obtaining initial EEG signal data of multiple subjects, performing global average reference on the initial EEG signal data to obtain reference EEG signal data, and filtering the reference EEG signal data using a preset filter to obtain a filtered initial training set.

[0052] Step S202: input the initial training set into the conditional diffusion model, perform forward noise addition on the initial training set to obtain forward noise added data, input the first information and the second information into the conditional diffusion model, train the conditional diffusion model to predict the noise of the forward noise added data, and obtain the trained conditional diffusion model.

[0053] First, the EEG signal data of multiple subjects are obtained. The subjects are patients known to have depression. The EEG signal data of multiple subjects are subjected to a global average reference (CAR) to obtain reference EEG signal data. The global average reference converts the collected signal from the original reference electrode-based signal to a reference signal based on the average signal.

[0054] Further, the reference EEG signal data is filtered using a preset filter to obtain the initial training set after filtering, wherein the filter can be a 50Hz band-stop FIR filter to filter out the 50Hz power frequency signal with strong energy in the original signal, and then a 1-50HZ bandpass FIR filter is used to retain the signals (Delta, Theta, Alpha, Beta, Gamma) containing the EEG component frequency in the signal, and finally these signals are segmented into 2-second non-overlapping windows, and the data signal that is not enough to form 2 seconds is discarded to form N_trails windows of data to obtain the training set. For example, there is a 3-by-2000 signal with a sampling rate of 250Hz, 3 represents the channel, and 2000 represents the number of data points. If each window is 3-by-500, then N_trails is equal to 4.

[0055] The initial training set is input into the conditional diffusion model, and the initial training set is forward-noised to obtain forward-noised data. The conditional diffusion model is mainly used to amplify samples of EEG data, which has improved the accuracy and generalization of subsequent classification prediction models. Since EEG signals have great heterogeneity in different populations, even under the condition of depression, different patients have different EEG signals. Therefore, in order to better introduce individual patient and depression characteristics, the subject's number information and label information representing whether he or she suffers from depression are added to the diffusion process, so that the conditional diffusion model denoises and generates EEG signals with more individual differences.

[0056] The conditional diffusion model includes a noise adding module, which inputs the initial training set and the random noise z0 extracted from the standard normal distribution into the conditional diffusion model. The noise adding module performs forward noise adding processing on the initial training set to obtain forward noise adding data. Specifically, The original signal (E is the number of channels 3, L is the number of sampling points 500, the initial training set) is denoised. The denoising module calculates the forward denoised data according to formula (2) using the randomly generated time step t and the random noise z0 extracted from the standard normal distribution.

[0057] For initial x0 and x t For example, the following relationship holds:

[0058]

[0059] where α t =1-β t , So we get x t Noised signal at time step t.

[0060] Among them, β min , β maxare the hyperparameters of the preset conditional diffusion model. In a possible implementation, βmin=0.0001, βmax=0.02, and t_step=100.

[0061] In this embodiment, the first information and the second information are input into the conditional diffusion model, and the conditional diffusion model is trained to predict the noise of the forward noisy data to obtain a trained conditional diffusion model.

[0062] The specific steps of training the conditional diffusion model are shown in steps S301 to S303:

[0063] Step S301: input the forward noisy data into the first convolutional layer to obtain first noisy data, and use the SiLU activation function to process the first noisy data to obtain second noisy data.

[0064] Step S302: input the first information and the second information into the feature embedding layer to obtain corresponding first features and second features, and input the time step into the time step embedding layer to obtain corresponding third features.

[0065] Step S303: training a conditional diffusion model based on the second noisy data, the first feature, the second feature and the third feature to predict the noise of the forward noisy data, and obtaining a trained conditional diffusion model.

[0066] The conditional diffusion model also includes the first convolution layer, the time step embedding layer and the feature embedding layer. The forward noisy data set is input into the first convolution layer to obtain the first noisy data. The first noisy data is processed by the SiLU activation function to obtain the second noisy data. The initial noisy data set is convolved by the first convolution layer of size (64,3,1), and then after the SiLU activation function, the feature size of (64,1,500) is obtained. That is, the second noisy data. The second noisy data is represented in a matrix form.

[0067] At the same time, the first information and the second information are input into the feature embedding layer to obtain the corresponding first feature and second feature γ * , and input the time step into the time step embedding layer to obtain the corresponding third feature t * , based on the second noisy data, the first feature, the second feature and the third feature, the conditional diffusion model is trained to predict the noise of the forward noisy data, and the trained conditional diffusion model is obtained. The first feature, the second feature and the third feature are all represented by matrices.

[0068] For the subject ID in γ and the label information representing whether the subject suffers from depression, the following formula can be used for feature embedding:

[0069]

[0070] in Feature input representing subject number information, The feature input represents the subject’s label information, m represents the number of subject ID information, n represents the number of subject label information, and k represents the number of embedding layer feature dimensions.

[0071] W represents a feature search table. For example, the subject number information is represented by a feature vector. Assuming that the dimension of a feature vector is 1*k, the feature vectors corresponding to the subject number information are sequentially combined into an m*k matrix according to the subject number, that is, According to the subject number, Get the feature vector representing the subject number corresponding to the subject Similarly, for the label information of the subject, assuming that the dimension of a feature vector representing the label information is 1*k, then the feature vectors corresponding to the subject label information are sequentially combined into an n*k matrix according to the subject number, that is, According to the subject number, you can Get the feature vector representing the label information corresponding to the subject

[0072] The embedding layer at time step t is obtained by the following formula, where N is 512:

[0073] θ={X|x=2n,n∈N) (5)

[0074]

[0075] θ represents an even-numbered queue, t represents the current time step, from 0 to 99, and Te represents a 2*N matrix.

[0076] The steps of training the conditional diffusion model based on the second noisy data, the first feature, the second feature and the third feature to predict the noise of the forward noisy data and obtaining the trained conditional diffusion model are shown in steps S401 and S402:

[0077] Step S401, perform a matrix element-by-element addition operation on the first feature, the second feature, and the third feature to obtain a first matrix, input the first matrix to a linear layer for processing to obtain a corresponding second matrix, perform a matrix element-by-element addition operation on the second matrix and the second noisy data to obtain a third matrix; input the third matrix to a dilated convolution layer to obtain a fourth matrix.

[0078] Step S402, performing convolution processing on the fourth matrix to obtain a feature matrix and third noisy data; using the third noisy data as the second noisy data, returning to perform a step of performing a matrix element-by-element addition operation on the second matrix and the second noisy data to obtain a third matrix, until the cycle is repeated M times; training a conditional diffusion model based on the M feature matrices to predict the noise of the forward noisy data, to obtain a trained conditional diffusion model.

[0079] The conditional diffusion model also includes a residual feature extraction module, which includes a linear layer and a hole convolution layer. * And the third feature t * Perform a matrix element-by-element addition operation to obtain a first matrix, input the first matrix to a linear layer to obtain a corresponding second matrix, perform a matrix element-by-element addition operation on the second matrix and the second noisy data to obtain a third matrix, input the third matrix to a dilated convolution layer to obtain a fourth matrix. In a possible implementation, the first matrix E is obtained by formula (7), the target matrix is ​​input to a linear layer to obtain a corresponding first matrix, the first matrix and the second noisy data are added element-by-element to obtain a third matrix, and the third matrix is ​​input to a dilated convolution layer to obtain a fourth matrix.

[0080] E=T e +E s +E l (7)

[0081] The fourth matrix is ​​convolved to obtain the feature matrix and the third noisy data. The third noisy data is used as the second noisy data. Return to perform the step of performing the matrix element-by-element addition operation on the second matrix and the second noisy data to obtain the third matrix, until the cycle is repeated M times to obtain M feature matrices. The conditional diffusion model is trained based on the M feature matrices to predict the noise of the forward noisy data, and the trained conditional diffusion model is obtained. According to formula (3-6), t and γ are obtained to obtain the corresponding features after the embedding layer. and Will t * , γ * As input parameters, they are input into the residual feature extraction module. * and γ * The features are processed by linear layers and finally Perform element-by-element addition on the matrix to obtain the embedded third matrix, and then pass it through a dilated convolution layer to obtain the fourth matrix of (128, 1, 500).

[0082] The steps of performing convolution processing on the fourth matrix to obtain the feature matrix and the third noisy data are shown in step S501-step S502:

[0083] Step S501: After performing nonlinear transformation on the fourth matrix using a hyperbolic tangent function and a Sigmoid activation function respectively, a fifth matrix and a sixth matrix are obtained correspondingly.

[0084] Step S502: perform a matrix element-by-element multiplication operation on the fifth matrix and the sixth matrix to obtain a seventh matrix, input the seventh matrix into the second convolution layer to obtain a feature matrix; input the seventh matrix into the third convolution layer to obtain third noisy data.

[0085] The residual feature extraction module also includes a second convolution layer and a third convolution layer. After performing nonlinear transformation on the fourth matrix through the hyperbolic tangent function and the Sigmoid activation function, respectively, a fifth matrix and a sixth matrix are obtained correspondingly. The fifth matrix and the sixth matrix are multiplied element by element to obtain a seventh matrix. The seventh matrix is ​​input into the second convolution layer to obtain a feature matrix. The seventh matrix is ​​input into the third convolution layer to obtain third noisy data.

[0086] The third noisy data is used as the second noisy data, and the step of performing element-by-element addition operation on the second matrix and the second noisy data to obtain the third matrix is ​​returned to the step, until the loop is repeated M times, and the conditional diffusion model is trained based on the M feature matrices to predict the noise of the forward noisy data, so as to obtain the trained conditional diffusion model.

[0087] In this embodiment, the conditional diffusion model is trained based on M feature matrices to predict the noise of the forward noisy data, and the steps of obtaining the trained conditional diffusion model are shown in steps S601 and S602:

[0088] Step S601: perform a matrix element-by-element addition operation on the obtained M feature matrices and a preset initial feature matrix to obtain a target feature matrix, and input the target feature matrix into a fourth convolutional layer to obtain fourth noisy data.

[0089] Step S602: Use the SiLU activation function to calculate the fourth noisy data to obtain fifth noisy data, input the fifth noisy data into the fifth convolutional layer to obtain predicted noise data, train the conditional diffusion model based on the predicted noise data until convergence, and obtain the trained conditional diffusion model.

[0090] The conditional diffusion model also includes a noise extraction module, which also includes a fourth convolutional layer and a fifth convolutional layer. The obtained M feature matrices are subjected to matrix element-by-element addition operations to obtain a target feature matrix, which is input into the fourth convolutional layer to obtain fourth noisy data, and the fourth noisy data is calculated and processed using a SiLU activation function to obtain fifth noisy data, which is input into the fifth convolutional layer to obtain predicted noise data.

[0091] The conditional diffusion model is trained based on the predicted noise data and the random noise extracted from the standard normal distribution until convergence, and the trained conditional diffusion model is obtained. Specifically, the predicted noise data and the random noise are input into the MSE loss function for calculation, and the MSE loss function is continuously iterated to converge, and the conditional diffusion model is obtained when the calculated value of the MSE loss function reaches a stable state, so as to complete the training of the conditional diffusion model and obtain the trained target model.

[0092] After the nonlinear transformation of the tanh function (hyperbolic tangent function) and the sigmoid function, it finally passes through two different convolutional layers and outputs two matrices of the same size (64, 1, 500), one of which is used as the input of the next residual feature extraction module, and the other is added to the feature matrix. After calculation by the residual network extraction module of M (for example, M = 32) layers, M feature matrices are obtained. The obtained M feature matrices and the preset initial feature matrix are added element by element to obtain a target feature matrix. After the target feature matrix is ​​sent to the noise extraction module, the noise data of the same size as the original input is obtained, and the MSE loss is obtained according to formula (8), and then the network ∈ is trained. θ .

[0093]

[0094] Training model ∈ θ The essence of is that at each time step t, according to the input label γ of the sample and the current predicted noise data To predict the random noise ∈ added in the current step, so that formula (8) is minimized, where is the mean square error (MSE) loss function,

[0095] When the conditional diffusion model training is completed, step S203 is executed to input the random noise, the first information and the second information of the subject into the trained conditional diffusion model, and reverse denoising is performed on the random noise to obtain denoised generated data.

[0096] The random noise extracted from the standard normal distribution and the first information and the second information of the subject are input into the trained conditional diffusion model, and the random noise is reversely denoised to obtain the denoised generated data.

[0097] Reference Figure 3 , Figure 3 A flowchart of a reverse denoising process provided by an embodiment of the present invention is shown in FIG. Figure 3 As shown, in the reverse denoising process, the subject's ID information, label information, time step t and Gaussian signal x t (random noise), sampled according to the following formula:

[0098]

[0099] In this embodiment, the time step t is from 99 to 0, a total of 100 time steps, and the random noise x t , time step t and label γ, are input to the network ∈ θ According to formula (8), the denoised signal x is obtained t-1 , and loop in sequence until x0 is generated. By sampling each subject, this method obtains K times the number of original samples. These generated data are merged with the original EEG signal data to obtain the final data enhancement sample.

[0100] Specifically, the initial EEG signal data of multiple subjects and the denoised generated data output by the conditional diffusion model are combined to obtain a target training set, and the prediction model is obtained by training a preset model with the target training set. For example, the preset model is a neural network model.

[0101] The EEG signal data of the patient to be tested is obtained to obtain target EEG signal data, a global average reference is performed on the target EEG signal data to obtain target reference EEG signal data, and the target reference EEG signal data is filtered using a preset filter to obtain a data set of N windows after filtering.

[0102] Input the data sets of N windows into the trained prediction model respectively, and obtain N corresponding prediction labels. Obtain the number of target labels in the prediction labels, and determine the proportion of the number of target labels in the N prediction labels as the probability that the patient to be tested suffers from depression. Among them, the target label is a label that represents depression. For example, if N is 4, if the number of target labels in the prediction label is 3, then the probability that the patient to be tested suffers from depression is

[0103] In one possible implementation, the preset prediction model calculates the corresponding prediction probability values ​​based on the input N target noise data, and the output label corresponding to the prediction probability value greater than the preset probability threshold is the target label, and the output label corresponding to the prediction probability value not greater than the preset probability threshold is the non-target label, and the non-target label indicates that the patient to be tested does not suffer from depression. That is, when the prediction probability value corresponding to the target noise data is greater than the preset probability threshold, the output label is the target label, and when the prediction probability value corresponding to the target noise data is not greater than the preset probability threshold, the output label is the non-target label. In another possible implementation, the N target noise data are respectively input into the preset prediction model, and N corresponding prediction probability values ​​are obtained. The number of prediction probability values ​​greater than the preset probability threshold is obtained to obtain the target number, and the proportion of the target number in the total number of prediction probability values ​​is determined as the probability that the patient to be tested suffers from depression.

[0104] Reference Figure 4 , Figure 4 A schematic diagram of a prediction model provided by an embodiment of the present invention, such as Figure 4 As shown, in the prediction model, for the input data set with N windows The overall calculation is performed using three convolutional layers and one fully connected layer. After each convolutional layer, BatchNormalized normalization, ReLU activation function, average pooling, and Dropout strategies are used. These four steps become the general processing module after the convolutional layer. For the first convolutional layer (time domain convolution module), a convolution kernel with a convolution kernel size of (1, 60) is first used to perform convolution operations on each channel signal to extract the features of each time domain signal. Here, the method of padding both ends with 0 is adopted. This method makes the size of the final extracted features consistent with the original size, and then it is sent to the general processing module output, and then the second convolutional layer (through The channel domain convolution module) uses a channel domain convolution kernel of (3, 1) to merge the channels of the three features to obtain a one-dimensional feature vector, which is sent to the general processing module for output. The third convolution layer uses a convolution kernel of (1, 10) and (1, 1) to extract the one-dimensional feature vector successively, which is output through the general processing module. Finally, a fully connected layer is used to convert the size of the feature vector to 2. The final prediction model will output the probability values ​​corresponding to N_trails windows, count the number of N_trails probabilities greater than the probability threshold Q, and finally calculate the probability of depression based on (Q / N_trails). The above is the reasoning process of the model. The training set for training the prediction model is the original EEG signals collected clinically and the EEG data generated by the subsequent conditional diffusion model. The training process uses cross entropy as the loss function and Adam optimizer, with a learning rate of 0.0001, a regularization parameter of 0.01, a BatchSize of 15, a Dropout of 0.2, and a maximum Epoch number of 100. The training process uses the Pytorch framework, the GPU is NVIDIAA40 graphics card, and the system is Ubuntu20.04 LST.

[0105] Based on the above method, the first information and the second information are embedded into the diffusion process as features, which can better introduce the individual characteristics of patients and depression characteristics, so that the conditional diffusion model denoising generates EEG signals with more individual differences. In addition, using the filtered N window data set as the input of the prediction model, the overall target brain point signal data can be divided into more detailed data to obtain more accurate prediction labels and improve the accuracy of the prediction probability value.

[0106] In order to better implement the above method, the embodiment of the present invention provides a depression probability determination device, referring to Figure 5 , Figure 5 A structural block diagram of a depression probability determination device provided by an embodiment of the present invention, such as Figure 5 As shown, the device 50 specifically includes:

[0107] Acquisition module 501: used to acquire EEG signal data of a patient to be tested, obtain target EEG signal data, and filter the target EEG signal data using a preset filter to obtain a data set of N windows after filtering;

[0108] The label prediction module 502 is used to input N data sets into a preset prediction model respectively, and obtain N corresponding prediction labels, wherein the prediction model is a model obtained after training with a target training set, and the target training set is a collection of a first training set and a second training set, wherein the first training set includes initial EEG signal data of multiple subjects, and the second training set includes denoised generated data outputted by a conditional diffusion model based on the initial EEG signal data of multiple subjects, first information, and second information, wherein the first information is the subject's number information, and the second information is label information indicating whether the subject suffers from depression;

[0109] Probability prediction module 503: used to obtain the number of target labels in the predicted labels, and determine the proportion of the number of target labels in N predicted labels as the probability that the patient to be tested suffers from depression; wherein the target label is a label that represents depression.

[0110] In one possible design, the acquisition module 501 is specifically used to: acquire initial EEG signal data of multiple subjects, perform global average reference on the initial EEG signal data to obtain reference EEG signal data, filter the reference EEG signal data using a preset filter, and obtain a filtered initial training set.

[0111] In a possible design, the device 50 also includes a conditional diffusion model data generation module 504, which is specifically used to: input the initial training set into the conditional diffusion model, forward-noise the initial training set to obtain forward-noised data, and input the first information and the second information into the conditional diffusion model to train the conditional diffusion model to predict the noise of the forward-noised data to obtain a trained conditional diffusion model; and input random noise, the first information and the second information of the subject into the trained conditional diffusion model, reversely denoise the random noise, and obtain denoised generated data.

[0112] In a possible design, the label prediction module 502 is specifically used to: train a preset model through a target training set to obtain a trained prediction model.

[0113] In a possible design, the conditional diffusion model includes a first convolutional layer, a time step embedding layer and a feature embedding layer, and the above-mentioned conditional diffusion model data generation module 504 is specifically used to: input the forward noisy data into the first convolutional layer to obtain first noisy data, and use the SiLU activation function to process the first noisy data to obtain second noisy data; input the first information and the second information into the feature embedding layer to obtain the corresponding first feature and second feature, and input the time step into the time step embedding layer to obtain the corresponding third feature; based on the second noisy data, the first feature, the second feature and the third feature, train the conditional diffusion model to predict the noise of the forward noisy data to obtain the trained conditional diffusion model.

[0114] In a possible design, the conditional diffusion model also includes a linear layer and a hole convolution layer, and the above-mentioned conditional diffusion model data generation module 504 is specifically used to: input the first feature, the second feature and the third feature into the linear layer for processing to obtain the corresponding first matrix, perform matrix element-by-element addition operation on the first matrix and the second noisy data to obtain a third matrix; input the third matrix into the hole convolution layer to obtain a fourth matrix; perform convolution processing on the fourth matrix to obtain a feature matrix and third noisy data; use the third noisy data as the second noisy data, return to execute the step of performing matrix element-by-element addition operation on the first matrix and the second noisy data to obtain the third matrix, until the cycle is repeated M times; train the conditional diffusion model based on M feature matrices to predict the noise of the forward noisy data to obtain the trained conditional diffusion model.

[0115] In a possible design, the conditional diffusion model also includes a second convolution layer and a third convolution layer, and the conditional diffusion model data generation module 504 is specifically used to: perform nonlinear transformation on the fourth matrix through a hyperbolic tangent function and a Sigmoid activation function, respectively, to obtain a fifth matrix and a sixth matrix correspondingly; perform a matrix element-by-element multiplication operation on the fifth matrix and the sixth matrix to obtain a seventh matrix, and input the seventh matrix into the second convolution layer to obtain a feature matrix; input the seventh matrix into the third convolution layer to obtain third noisy data.

[0116] In a possible design, the conditional diffusion model also includes a fourth convolutional layer and a fifth convolutional layer, and the above-mentioned conditional diffusion model data generation module 504 is specifically used to: perform matrix element-by-element addition operation on the obtained M feature matrices and the preset initial feature matrix to obtain a target feature matrix, and input the target feature matrix into the fourth convolutional layer to obtain fourth noisy data; use the SiLU activation function to calculate the fourth noisy data to obtain fifth noisy data, input the fifth noisy data into the fifth convolutional layer to obtain predicted noise data, and train the conditional diffusion model based on the predicted noise data until convergence to obtain the trained conditional diffusion model.

[0117] Based on the above device, a more accurate probability of depression can be predicted.

[0118] The embodiment of the present invention provides a depression probability determination system, referring to Figure 6 , Figure 6 A structural block diagram of a depression probability determination system provided by an embodiment of the present invention, such as Figure 6As shown, the system 60 includes an EEG signal acquisition device 601, a communication module 602, a host computer 603 and a server 604. The EEG signal acquisition device is connected to the host computer, and the host computer and the EEG signal acquisition device are respectively connected to the communication module, wherein the host computer is used to send a control signal representing the EEG signal of the patient to be tested to the EEG signal acquisition device, and the EEG signal acquisition device is used to receive the control signal, collect the EEG signal data of the patient to be tested, and send the acquired target EEG signal data to the host computer through the communication module. The communication module may include Bluetooth, WIFI, USB, etc. The host computer is also used to receive the target EEG signal data and send the target EEG signal data to the server. The server is used to receive the target EEG signal data and use a preset filter to filter the target EEG signal data to obtain N filtered data. The data set of the number of windows; inputting N data sets into the preset prediction model respectively, and obtaining N prediction labels correspondingly, wherein the prediction model is a model obtained after training with the target training set, and the target training set is a collection of the first training set and the second training set, the first training set includes the initial EEG signal data of multiple subjects, and the second training set includes the denoising generated data output by the conditional diffusion model based on the initial EEG signal data of multiple subjects, the first information and the second information, the first information is the subject's number information, and the second information is the label information indicating whether the patient suffers from depression; obtaining the number of target labels in the prediction label, and determining the proportion of the number of target labels in the N prediction labels as the probability of the patient to be tested suffering from depression; wherein the target label is a label characterizing depression. The server is also used to output the calculated probability of the patient to be tested suffering from depression to the host computer for display. The server is also used to output the calculated probability of the patient to be tested suffering from depression to the host computer for display.

[0119] The EEG signal acquisition equipment can be controlled by a single-chip microcomputer and a 24-bit high-precision ADC acquisition circuit. The sampling chip can use an 8-channel synchronous sampling ADS1299 chip, which can simultaneously acquire microvolt-level EEG signals from three leads, namely Fp1, Fz, and Fp2 leads. Their positions are determined according to the 10-20 international standard lead system, and the sampling rate is 250Hz. The EEG signal can be transmitted to the host computer software end for processing via USB, Bluetooth, and WIFI. When the acquisition is completed, the software end will perform persistent processing on the recorded EEG signal. The upper computer software transmits and returns the persistently stored EEG signal data to and from the server. Based on the TCP\IP protocol of byte stream transmission, the Socket application programming interface is used to realize the data signal transmission and return from the software to the server. At this time, the software acts as a client and transmits and returns data to and from the server as a server. The encoding and decoding methods of the client and server are both performed in little-endian mode. The specific steps are as follows: First, the server establishes a fixed IP and port number, and the client connects according to the server's IP and port number. Then the client sends the persistent EEG data file to the server. Read and calculate the number of bytes N of all EEG data to be transmitted, and transmit the number of bytes N to the server through the TCP\IP protocol. After receiving the number of bytes N, the server starts to monitor and receive the next N bytes of data. At this time, the client will start to send N bytes of EEG signal data. After receiving, the server will use the data online preprocessing module and the online depression diagnosis module based on convolutional neural network for analysis, and finally get a predicted probability value (the probability value ranges from 0 to 1, the higher the value, the greater the possibility of depression) and send it to the client, and display it on the client to give the final diagnosis result. The data online preprocessing module is used to filter the target EEG signal data using a preset filter to obtain a data set of N windows after filtering; the N data sets are respectively input into a preset prediction model to obtain N corresponding prediction labels, wherein the prediction model is a model obtained after training with a target training set, and the target training set is a collection of a first training set and a second training set, the first training set includes initial EEG signal data of multiple subjects, and the second training set includes denoising generated data output by a conditional diffusion model based on the initial EEG signal data of multiple subjects, first information and second information, the first information is the subject's number information, and the second information is label information indicating whether the subject suffers from depression; the number of target labels in the prediction label is obtained, and the proportion of the number of target labels in the N prediction labels is determined as the probability that the patient to be tested suffers from depression; wherein the target label is a label representing depression.

[0120] Figure 6FIG. 1 shows an internal structure diagram of a computer device in an embodiment. The computer device may be a terminal or a server. Figure 6 As shown, the computer device includes a processor, a memory and a network interface connected via a system bus. The memory includes a non-volatile storage medium and an internal memory. The non-volatile storage medium of the computer device stores an operating system and may also store a computer program. When the computer program is executed by the processor, the processor can implement all the steps of the above method. The internal memory may also store a computer program. When the computer program is executed by the processor, the processor can implement all the steps of the above method. Those skilled in the art will understand that Figure 6 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0121] In one embodiment, a computer device is provided, including a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the processor executes each step of the aforementioned method.

[0122] In one embodiment, a computer-readable storage medium is provided, which stores a computer program. When the computer program is executed by a processor, the processor executes each step of the aforementioned method.

[0123] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the program can be stored in a non-volatile computer-readable storage medium. When the program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

[0124] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0125] The above-mentioned embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the present application. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the attached claims.

Claims

1. A method for determining the probability of depression, characterized in that: The method comprises: Acquire the EEG signal data of the patient to be tested to obtain target EEG signal data, and use a preset filter to filter the target EEG signal data to obtain a data set of N windows after filtering; Inputting N data sets into a preset prediction model respectively, and obtaining N corresponding prediction labels, wherein the prediction model is a model obtained after training with a target training set, and the target training set is a collection of a first training set and a second training set, wherein the first training set includes initial EEG signal data of multiple subjects, and the second training set includes denoised generated data outputted by a conditional diffusion model based on the initial EEG signal data of multiple subjects, first information, and second information, wherein the first information is the subject's number information, and the second information is label information indicating whether the subject suffers from depression; The number of target labels in the predicted labels is obtained, and the proportion of the number of target labels in the N predicted labels is determined as the probability that the patient to be tested suffers from depression; wherein the target label is a label that indicates depression.

2. The method according to claim 1, characterized in that De-noising and generating data based on initial EEG signal data of multiple subjects, first information, and second information output by a conditional diffusion model, including: Acquire initial EEG signal data of multiple subjects, perform global average reference on the initial EEG signal data to obtain reference EEG signal data, and use a preset filter to filter the reference EEG signal data to obtain a filtered initial training set; Inputting the initial training set into the conditional diffusion model, performing forward noise addition on the initial training set to obtain forward noise added data, and inputting the first information and the second information into the conditional diffusion model to train the conditional diffusion model to predict the noise of the forward noise added data, thereby obtaining a trained conditional diffusion model; The random noise, the first information and the second information of the subject are input into the trained conditional diffusion model, and the random noise is reversely denoised to obtain denoised generated data.

3. The method according to claim 2, characterized in that The conditional diffusion model includes a first convolutional layer, a time step embedding layer and a feature embedding layer. The first information and the second information are input into the conditional diffusion model to train the conditional diffusion model to predict the noise of the forward noisy data, and obtain the trained conditional diffusion model, including: Inputting the forward noisy data into the first convolutional layer to obtain first noisy data, and processing the first noisy data using a SiLU activation function to obtain second noisy data; Inputting the first information and the second information into the feature embedding layer to obtain corresponding first features and second features, and inputting the time step into the time step embedding layer to obtain corresponding third features; A conditional diffusion model is trained based on the second noisy data, the first feature, the second feature, and the third feature to predict the noise of the forward noisy data, thereby obtaining a trained conditional diffusion model.

4. The method according to claim 3, characterized in that The conditional diffusion model further includes a linear layer and a dilated convolution layer. The conditional diffusion model is trained based on the second noisy data, the first feature, the second feature, and the third feature to predict the noise of the forward noisy data to obtain the trained conditional diffusion model, including: Performing a matrix element-by-element addition operation on the first feature, the second feature, and the third feature to obtain a first matrix, inputting the first matrix into a linear layer for processing to obtain a corresponding second matrix, performing a matrix element-by-element addition operation on the second matrix and the second noisy data to obtain a third matrix; inputting the third matrix into a dilated convolution layer to obtain a fourth matrix; The fourth matrix is ​​convolved to obtain a feature matrix and third noisy data; the third noisy data is used as the second noisy data, and the step of performing a matrix element-by-element addition operation on the second matrix and the second noisy data to obtain a third matrix is ​​returned to the step, until the cycle is repeated M times; a conditional diffusion model is trained based on the M feature matrices to predict the noise of the forward noisy data to obtain a trained conditional diffusion model.

5. The method according to claim 4, characterized in that The conditional diffusion model further includes a second convolution layer and a third convolution layer, and the convolution processing is performed on the fourth matrix to obtain a feature matrix and third noise-added data, including: After performing nonlinear transformation on the fourth matrix by using a hyperbolic tangent function and a sigmoid activation function respectively, a fifth matrix and a sixth matrix are obtained correspondingly; Perform a matrix element-by-element multiplication operation on the fifth matrix and the sixth matrix to obtain a seventh matrix, input the seventh matrix into the second convolution layer to obtain a feature matrix; input the seventh matrix into the third convolution layer to obtain third noisy data.

6. The method according to claim 5, characterized in that The conditional diffusion model further includes a fourth convolutional layer and a fifth convolutional layer. The conditional diffusion model is trained based on the M feature matrices to predict the noise of the forward noisy data to obtain the trained conditional diffusion model, including: Performing a matrix element-by-element addition operation on the obtained M feature matrices and a preset initial feature matrix to obtain a target feature matrix, and inputting the target feature matrix into a fourth convolutional layer to obtain fourth noisy data; The fourth noisy data is calculated and processed by using a SiLU activation function to obtain fifth noisy data, the fifth noisy data is input into a fifth convolutional layer to obtain predicted noise data, and a conditional diffusion model is trained based on the predicted noise data until convergence to obtain a trained conditional diffusion model.

7. A depression probability determination system, characterized in that: The system includes an EEG signal acquisition device, a communication module, a host computer and a server; The host computer is connected to the EEG signal acquisition device and is used to send a control signal representing the EEG signal of the patient to be tested to the EEG signal acquisition device; The EEG signal acquisition device is used to receive the control signal and acquire EEG signal data of the patient to be tested; The host computer and the EEG signal acquisition device are respectively connected to a communication module, and the EEG signal acquisition device is also used to send the acquired target EEG signal data to the host computer through the communication module; The host computer is also used to receive target EEG signal data and send the target EEG signal data to the server; The server is used to receive target EEG signal data, and filter the target EEG signal data with a preset filter to obtain a data set of N windows after filtering; the N data sets are respectively input into a preset prediction model to obtain N corresponding prediction labels, wherein the prediction model is a model obtained after training with a target training set, and the target training set is a collection of a first training set and a second training set, the first training set includes initial EEG signal data of multiple subjects, and the second training set includes denoising generated data output by a conditional diffusion model based on the initial EEG signal data of multiple subjects, first information and second information, the first information is the subject's number information, and the second information is label information indicating whether the subject suffers from depression; the number of target labels in the prediction label is obtained, and the proportion of the number of target labels in the N prediction labels is determined as the probability that the patient to be tested suffers from depression; wherein the target label is a label representing depression.

8. A device for determining the probability of depression, characterized in that: The device comprises: Acquisition module: used to acquire the EEG signal data of the patient to be tested, obtain target EEG signal data, and filter the target EEG signal data using a preset filter to obtain a data set of N windows after filtering; Label prediction module: used to input N data sets into preset prediction models respectively, and obtain N corresponding prediction labels, wherein the prediction model is a model obtained after training with a target training set, and the target training set is a collection of a first training set and a second training set, wherein the first training set includes initial EEG signal data of multiple subjects, and the second training set includes denoised generated data output by a conditional diffusion model based on the initial EEG signal data of multiple subjects, first information, and second information, wherein the first information is the subject's number information, and the second information is label information indicating whether the subject suffers from depression; Probability prediction module: used to obtain the number of target labels in the predicted labels, and determine the proportion of the number of target labels in N predicted labels as the probability that the patient to be tested suffers from depression; wherein the target label is a label that represents depression.

9. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the processor is caused to perform the steps of the method according to any one of claims 1 to 6.

10. A computer device comprising a memory and a processor, characterized in that: The memory stores a computer program, and when the computer program is executed by the processor, the processor executes the steps of the method according to any one of claims 1 to 6.

Images