A depression disorder detection method and detection device based on sleep monitoring

Abstract

The application provides a sleep monitoring-based depression disorder detection method and device, and relates to the technical field of intelligent depression detection, and comprises the following steps: acquiring a polysomnogram of a to-be-detected object, the polysomnogram comprising a plurality of physiological signals; inputting the plurality of physiological signals into a first depression disorder detection model to obtain a first intermediate layer signal, wherein the first depression disorder detection model comprises an attention module, a CNN sequence, a first channel, and a data splicing module, the attention module is used to acquire an attention index of each physiological signal, calculate an attention weight of each physiological signal, and integrate all the attention weights into a second intermediate layer signal; inputting the first intermediate layer signal into a second depression disorder detection model to obtain a third intermediate layer signal; and inputting the third intermediate layer signal into a third depression disorder detection model to obtain a depression disorder detection result of the to-be-detected object. Through the application, the accuracy of the depression disorder detection result is improved.

Description

Technical Field

[0001] This application relates to the field of intelligent detection technology for depression, and more specifically, to a method and device for detecting depressive disorders based on sleep monitoring. Background Technology

[0002] Currently, sleep problems plague many people and have a close, two-way relationship with depressive disorders. Sleep problems are not only a common symptom of depression, but can also exacerbate or trigger depression if left unresolved for a long time. Therefore, accurately diagnosing whether a patient has depression is crucial for subsequent treatment during the diagnosis and treatment of sleep problems.

[0003] Clinical diagnostic criteria for depression are crucial for the clinical treatment of depressive disorders. Currently, diagnostic methods for depression mainly include clinical observation and questionnaires. However, these methods often rely on patient self-reports and clinicians' subjective judgment, which have certain limitations, such as strong subjectivity and limited accuracy in diagnosis.

[0004] Current methods for identifying physiological signals of depressive disorders suffer from problems such as complex design, the need to extract signal features in advance, and a lack of attention to physiological signals in the context of sleep. These issues make existing algorithms unsuitable for diagnosing depressive disorders based on sleep physiological signals. Summary of the Invention

[0005] In view of this, the purpose of this application is to provide a method and device for detecting depressive disorders based on sleep monitoring, so as to overcome at least one of the above-mentioned defects.

[0006] In a first aspect, embodiments of this application provide a method for detecting depressive disorders based on sleep monitoring. The method includes: acquiring a polysomnography of a subject within a target time period, the polysomnography including multiple physiological signals; inputting the multiple physiological signals into a first depressive disorder detection model to obtain a first intermediate layer signal, wherein the first depressive disorder detection model includes an attention module, a CNN sequence, a first channel, and a data stitching module, the attention module being used to acquire an attention index for each physiological signal, calculate an attention weight for each physiological signal, and integrate all attention weights into a second intermediate layer signal, the input of the CNN sequence being the second intermediate layer signal, the output of the CNN sequence being a time-domain signal, the input of the first channel being the second intermediate layer signal, the output of the first channel being a frequency-domain signal, the input of the data stitching module being the time-domain signal and the frequency-domain signal, and the output of the data stitching module being the first intermediate layer signal; inputting the first intermediate layer signal into a second depressive disorder detection model to obtain a third intermediate layer signal; and inputting the third intermediate signal into a third depressive disorder detection model to obtain a depressive disorder detection result for the subject.

[0007] In one optional embodiment of this application, both the CNN sequence and the first channel are 1D CNN sequences. The input of the plurality of physiological signals into the first depression disorder detection model to obtain a first intermediate layer signal includes: inputting the plurality of physiological signals into the attention module and outputting a second intermediate layer signal; inputting the second intermediate layer signal into the CNN sequence and the first channel respectively, and outputting the time-domain signal and the frequency-domain signal; and inputting the time-domain signal and the frequency-domain signal into the data splicing module to obtain the first intermediate layer signal.

[0008] In one optional embodiment of this application, the second depressive disorder detection model includes a memory module, wherein the first input of the memory module is the first intermediate layer signal, the first output of the memory module is the first feature, the second input of the memory module is the first feature, the second input of the memory module is the second feature, and the third intermediate layer signal is used to characterize the total feature obtained by splicing the first feature and the second feature.

[0009] In one optional embodiment of this application, the third depressive disorder detection model includes a first fully connected linear layer, a second fully connected linear layer, and a fully connected classification layer. The input of the first fully connected linear layer is the signal of the third intermediate layer, the output of the first fully connected linear layer is the input of the second fully connected linear layer, the output of the second fully connected linear layer is the input of the fully connected classification layer, and the output of the fully connected classification layer is the depressive disorder detection result.

[0010] In one optional embodiment of this application, the step of inputting the third intermediate layer signal into the third depressive disorder detection model to obtain the depressive disorder detection result of the test subject includes: inputting the third intermediate layer signal into the first fully connected linear layer and outputting the fourth intermediate layer signal; inputting the fourth intermediate layer signal into the second fully connected linear layer and outputting the fifth intermediate layer signal; and inputting the fifth intermediate layer signal into the fully connected classification layer to obtain the depressive disorder detection result of the test subject.

[0011] In an optional embodiment of this application, the method further includes: dividing the physiological signals under each channel of the polysomnography into multiple sleep segments according to a preset time, each sleep segment including EEG signals, EEG signals, ECG signals, and EMG signals under multiple channels within the preset time; sequentially inputting the EEG signals, EEG signals, ECG signals, and EMG signals corresponding to each sleep segment into a depression disorder detection model to obtain the depression disorder detection result of the subject, wherein the depression disorder detection model includes a first depression disorder detection model, a second depression disorder detection model, and a third depression disorder detection model.

[0012] In one optional embodiment of this application, the depression disorder detection result includes a first result and a second result, wherein the depression disorder detection result of the test subject is determined by the following method: each sleep segment is sequentially input into the depression disorder detection model to determine the result corresponding to the test subject in that sleep segment.

[0013] Secondly, this application also provides a sleep monitoring-based depressive disorder detection device, the device comprising: an acquisition module for acquiring a polysomnography of a subject within a target time period, the polysomnography including multiple physiological signals; and a first intermediate layer signal acquisition module for inputting the multiple physiological signals into a first depressive disorder detection model to obtain a first intermediate layer signal, wherein the first depressive disorder detection model includes an attention module, a CNN sequence, a first channel, and a data stitching module, the attention module being used to acquire an attention index for each physiological signal, calculate the attention weight for each physiological signal, and integrate all attention weights into a second intermediate layer signal. The intermediate layer signal is input to the CNN sequence as the second intermediate layer signal, and the output of the CNN sequence is a time-domain signal. The input of the first channel is the second intermediate layer signal, and the output of the first channel is a frequency-domain signal. The input of the data splicing module is the time-domain signal and the frequency-domain signal, and the output of the data splicing module is the first intermediate layer signal. The third intermediate layer signal acquisition module is used to input the first intermediate layer signal into the second depressive disorder detection model to obtain the third intermediate layer signal. The depressive disorder detection result module is used to input the third intermediate layer signal into the third depressive disorder detection model to obtain the depressive disorder detection result of the test subject.

[0014] Thirdly, embodiments of this application also provide an electronic device, including: a processor, a memory, and a bus, wherein the memory stores machine-readable instructions executable by the processor, and when the electronic device is running, the processor communicates with the memory via the bus, and when the machine-readable instructions are executed by the processor, the steps of the method described above are performed.

[0015] Fourthly, embodiments of this application also provide a computer-readable storage medium storing a computer program that, when executed by a processor, performs the steps of the method described above.

[0016] The present application provides a method and apparatus for detecting depressive disorders based on sleep monitoring, comprising: acquiring a polysomnography of a subject within a target time period, the polysomnography including multiple physiological signals; inputting the multiple physiological signals into a first depressive disorder detection model to obtain a first intermediate layer signal, wherein the first depressive disorder detection model includes an attention module, a CNN sequence, a first channel, and a data stitching module, the attention module being used to acquire an attention index for each physiological signal, calculate the attention weight for each physiological signal, and integrate all attention weights into a second intermediate layer signal; inputting the first intermediate layer signal into a second depressive disorder detection model to obtain a third intermediate layer signal; and inputting the third intermediate signal into a third depressive disorder detection model to obtain the depressive disorder detection result of the subject. Through this application, the processing and analysis of multiple physiological signals in the polysomnography improves the accuracy of the depressive disorder detection result.

[0017] To make the above-mentioned objectives, features and advantages of this application more apparent and understandable, preferred embodiments are described below in detail with reference to the accompanying drawings. Attached Figure Description

[0018] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of this application and should not be regarded as a limitation of the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 A flowchart illustrating the sleep monitoring-based method for detecting depressive disorders provided in an embodiment of this application;

[0020] Figure 2 This is a schematic diagram of the depression disorder detection model provided in the embodiments of this application;

[0021] Figure 3 A schematic diagram of the first depressive disorder detection model provided in the embodiments of this application;

[0022] Figure 4 This is a flowchart illustrating the process of obtaining the first intermediate layer signal, as provided in an embodiment of this application.

[0023] Figure 5 This is a schematic diagram of the second depressive disorder detection model provided in the embodiments of this application;

[0024] Figure 6 A schematic diagram of the third depressive disorder detection model provided in the embodiments of this application;

[0025] Figure 7This is a flowchart illustrating the process of obtaining the test results for a subject's depressive disorder, provided in an embodiment of this application.

[0026] Figure 8 This is a schematic diagram of the sleep monitoring-based depressive disorder detection device provided in an embodiment of this application.

[0027] Figure 9 This is a schematic diagram of the structure of the electronic device provided in the embodiments of this application. Detailed Implementation

[0028] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. The components of the embodiments of this application described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of this application provided in the accompanying drawings is not intended to limit the scope of the claimed application, but merely represents selected embodiments of this application. Based on the embodiments of this application, every other embodiment obtained by those skilled in the art without inventive effort falls within the scope of protection of this application.

[0029] First, the applicable application scenarios of this application will be introduced. This application can be applied to the field of intelligent detection technology for depression.

[0030] Research has found that current diagnostic methods for depression mainly include clinical observation and questionnaires. However, these methods often rely on patient self-reports and clinicians' subjective judgments, which have certain limitations, such as strong subjectivity and limited accuracy. Current methods for identifying physiological signals of depressive disorders are complex in design, require prior extraction of signal features, and pay little attention to physiological signals in the context of sleep. These problems make existing algorithms unsuitable for diagnosing depressive disorders based on sleep physiological signals.

[0031] Based on this, embodiments of this application provide a method and device for detecting depressive disorders based on sleep monitoring. The method acquires a polysomnography (containing multiple physiological signals) of the subject within a target time period and processes and analyzes these signals using a multi-level depressive disorder detection model. These models include attention modules, CNN sequences, etc., which can accurately calculate the attention weight of each physiological signal and integrate the information to ultimately obtain the depressive disorder detection result for the subject. This method improves the accuracy of depressive disorder detection.

[0032] Please see Figure 1 , Figure 1This is a flowchart illustrating a sleep monitoring-based method for detecting depressive disorders provided in an embodiment of this application. Figure 1 As shown in the embodiments of this application, the method for detecting depressive disorders based on sleep monitoring includes:

[0033] S101. Obtain the polysomnography of the subject under test within the target time period.

[0034] Polysomnography (PSG) includes multiple physiological signals.

[0035] In addition to electroencephalography (EEG), current standard PSG monitoring should include physiological signals from channels such as electrocardiography (ECG), electromyography (EMG), oculomotor imaging (EOG), thoracic and abdominal respiratory tension mapping (TIP), nasal and oral ventilation, body position and movement, blood oxygen saturation, and corpus cavernosum volume.

[0036] PSG examinations are conducted in a specially designed sleep laboratory. Electrodes and sensors are placed on the subject's scalp, face, chest, abdomen, limbs, and other areas according to monitoring needs. The placement of electrodes and sensors should follow internationally accepted standards to ensure the accuracy and comparability of monitoring data. The electrodes and sensors are connected to the PSG monitoring equipment to ensure stable and reliable signal transmission.

[0037] According to the monitoring needs, the sampling rate, filter and other parameters of the equipment are adjusted to obtain the best monitoring effect. After the subject falls asleep, the PSG monitoring equipment will start to continuously and synchronously record physiological signals such as electroencephalogram (EEG), electromyography (EMG), electrooculogram (EOG), electrocardiogram (ECG), and respiratory recording device. If necessary, it can also monitor physiological indicators reflecting cardiovascular function such as blood pressure and pulse at the same time.

[0038] The PSG monitoring device automatically collects physiological signal data throughout the night's sleep, which is called a polysomnography, and stores it in the device's internal or external storage. The collected data is preprocessed, including filtering and signal calibration, to improve the accuracy and reliability of the data. Professional sleep analysis software is used to perform detailed analysis on the preprocessed data, including assessments of sleep structure, process, abnormal EEG, respiratory function, and cardiovascular function. Based on the analysis results and clinical information, the sleep disorders of the subjects are diagnosed, classified, and differentially diagnosed.

[0039] Specifically, polysomnography can be obtained in the following ways:

[0040] First, ensure that the equipment used to record polysomnography (such as an electroencephalogram, electrooculogram, electromyogram, etc.) is in normal working order and correctly connected to the subject. Second, set the parameters for recording polysomnography according to the research or diagnostic needs, such as sampling rate, recording time, number of channels, etc. Finally, start the equipment and record the polysomnography of the subject within the preset time period, ensuring that the equipment is stable and the data is complete during the recording process. Store the collected polysomnography data in a storage medium for subsequent analysis and processing.

[0041] Multiple physiological signals are identified using the following methods:

[0042] Determine the filtering range for each physiological signal;

[0043] Current standard PSG monitoring, in addition to electroencephalography (EEG), should include physiological signals from channels such as electrocardiography (ECG), electromyography (EMG), oculomotor imaging (EOG), thoracic and abdominal respiratory tension mapping (TPT), nasal and oral ventilation, body position and movement, blood oxygen saturation, and corpus cavernosum volume. During preprocessing, the desired channels need to be filtered to an appropriate range.

[0044] Preferably, the filtering range can be set as follows:

[0045] Electroencephalogram (EEG): The frequency range of recording is usually between 0.3 and 35 Hz. The low-frequency filter (LFF) range is generally set to 0.3 Hz, and the high-frequency filter (HFF) range is generally set to 30 Hz or 35 Hz.

[0046] Electrooculography (EOG): Typically records frequencies between 0.3 and 35 Hz, recording the movement of the corneal-retinal potential difference. The frequency range may vary depending on the sleep stage. For example, the frequency may be 813 Hz during wakefulness, while it may decrease to 47 Hz in late sleep. The filter range should be set according to specific monitoring needs, but it usually includes these frequency ranges.

[0047] Electromyography (EMG): The frequency range of recording is usually between 10 and 100 Hz. It focuses on high-frequency signals, so the high cutoff filter setting is usually high, such as 70 Hz, to filter out unnecessary low-frequency interference.

[0048] Electrocardiogram (ECG): The filtering range of ECG signals is usually set according to clinical needs. For example, the frequency range that can be recorded is between 0.3 and 70 Hz.

[0049] In addition, a notch filter is applied to the relevant 50Hz channel to remove power frequency interference. When setting the filter range, the frequency characteristics of the signal, noise interference, and the monitoring objective must be considered. For example, a notch filter is typically used to eliminate voltage interference of 50-60Hz generated by standard cables.

[0050] For each physiological signal, the corresponding physiological signal in the polysomnography is filtered according to the corresponding filtering range to obtain multiple physiological signals of that physiological signal within the corresponding filtering range.

[0051] Based on the determined filtering range, an appropriate filter is selected to filter the physiological signal. Common filter types include low-pass filters, high-pass filters, and band-pass filters.

[0052] A low-pass filter allows signals with frequencies below a certain threshold to pass through while blocking signals with frequencies above that threshold. A high-pass filter, conversely, allows signals with frequencies above a certain threshold to pass through. A band-pass filter allows signals within a specific frequency range to pass through.

[0053] Filtering:

[0054] The physiological signals recorded in the polysomnography are input into the filter and filtered according to the set filtering range. The filtered signal will only contain information within the specified frequency range, thereby improving the signal-to-noise ratio and clarity.

[0055] Analyzing filtered physiological signals can extract useful information for diagnosis and research. For example, by analyzing filtered EEG signals, characteristic waveforms of different sleep stages (such as spindle waves and K-complexes) can be identified. Filtering can also help eliminate noise interference and improve diagnostic accuracy. For instance, when monitoring respiratory airflow, filtering can eliminate artifacts and interference caused by respiratory movements.

[0056] Specifically, for each channel of the polysomnography, the physiological signals under that channel are divided according to a preset time to obtain multiple sleep segments.

[0057] Each sleep segment includes EEG signals, EOG signals, ECG signals, and EMG signals from multiple channels within a preset time period;

[0058] Here, data segmentation serves two purposes: to facilitate data processing within the deep learning model and to ensure the accuracy and reliability of the obtained experimental data. This lays a solid foundation and provides a strong guarantee for further experimental and analytical work. Baseline correction is also performed on the segmented data. In the AASM (American Academy of Sleep Medicine) standards, each sleep stage is defined as 30 seconds; therefore, the data is segmented every 30 seconds.

[0059] The electroencephalogram (EEG), electrooculogram (EOG), electrocardiogram (ECG), and electromyogram (EMG) signals corresponding to each sleep segment are sequentially input into the depression disorder detection model to obtain the depression disorder detection results of the subject.

[0060] The depressive disorder detection models include the first depressive disorder detection model, the second depressive disorder detection model, and the third depressive disorder detection model.

[0061] Please see Figure 2 , Figure 2 This is a schematic diagram of the depressive disorder detection model provided in an embodiment of this application. Figure 2 As shown in the embodiments of this application, the depression disorder detection model includes a first depression disorder detection model (feature extraction), a second depression disorder detection model (reuse LSTM), and a third depression disorder detection model (classification).

[0062] S102. Input multiple physiological signals into the first depressive disorder detection model to obtain the first intermediate layer signal.

[0063] Please see Figure 3 , Figure 3 This is a schematic diagram of the first depressive disorder detection model provided in the embodiments of this application.

[0064] like Figure 3 As shown in the figure, the first depressive disorder detection model provided in this application embodiment includes an attention module, a CNN sequence, a first channel, and a data stitching module. The attention module is used to obtain the attention index of each physiological signal, calculate the attention weight of each physiological signal, and integrate all attention weights into a second intermediate layer signal. The input of the CNN sequence is the second intermediate layer signal, and the output of the CNN sequence is a time domain signal. The input of the first channel is the second intermediate layer signal, and the output of the first channel is a frequency domain signal. The input of the data stitching module is the time domain signal and the frequency domain signal, and the output of the data stitching module is the first intermediate layer signal.

[0065] like Figure 3 As shown, each square represents a module within the first depressive disorder detection model, and its specific content is as follows:

[0066] Multi-head attention is a mechanism used to capture information from different subspaces of the input signal. It linearly projects the input signal into multiple subspaces, performs an attention mechanism calculation in each subspace, and then concatenates the results before passing them through a linear layer to obtain the final output.

[0067] Bi-LSTM: Bidirectional Long Short-Term Memory network, used to process sequence data. Bi-LSTM can read the input sequence from both the forward and backward directions, thus capturing the context information at each position in the sequence.

[0068] Conv1d: A one-dimensional convolutional layer used to process sequential data. By sliding the convolution kernel across the sequence and performing convolution operations, local features can be captured.

[0069] ELU: Exponential Linear Unit Activation Function. ELU is a non-linear activation function, with the formula f(x) = x if x > 0, otherwise f(x) = α(ex − 1). ELU enables the network output to have zero centrality and alleviates the gradient vanishing problem.

[0070] Concat: A concatenation operation used to concatenate multiple feature vectors or tensors along a certain dimension to obtain a wider feature representation.

[0071] It can be understood that the output of the attention layer is a signal multiplied by a weight. The first attention layer multiplies the signal by taking the weight of each channel. (C, T) is the input format of the signal, where C represents the channel and T represents the time. The two attention layers on the left multiply the signal by taking the weight of T.

[0072] For example, the PSG channel count is 6 EEG, 2 EOG, 1 ECG, and 2 EMG. This model is adaptable to data with various channel counts, and the specific channel count can be limited according to the actual situation.

[0073] Specifically, the PSG signal is passed through the first layer of Attention to obtain the attention index of each PSG channel. The processed data is then processed through a one-dimensional CNN sequence and an LSTM+Attention sequence to extract the frequency domain and time domain features of the PSG signal, respectively. The two features are then integrated and processed before being transmitted to the reuse LSTM.

[0074] The entire process can be described as follows:

[0075] The input is a PSG signal; it is first processed by a multi-head attention mechanism; then it is processed by a one-dimensional convolutional layer with an ELU activation function and a bidirectional LSTM, and then processed by a multi-head attention mechanism again on each of them; the outputs of the convolutional layer and the bidirectional LSTM are merged by a concatenation operation to form the final feature representation.

[0076] Please see Figure 4 , Figure 4 This is a flowchart illustrating the process of obtaining the first intermediate layer signal, as provided in an embodiment of this application. Figure 4 As shown in the embodiments of this application, the sleep staging detection method includes:

[0077] S201. Input multiple physiological signals into the attention module and output the second intermediate layer signal.

[0078] In this step, the system receives multiple physiological signals as input. These signals include electroencephalogram (EEG), electrooculogram (EOG), electrocardiogram (ECG), and electromyogram (EMG), among others. The input signals are then processed by the attention module. The attention module employs a multi-head attention mechanism, which captures information from the input signals in different subspaces, enhancing focus on important features.

[0079] After processing by the attention module, the second intermediate layer signal is output. This signal is a preliminary feature extraction and enhancement of the input signal, providing a foundation for subsequent processing.

[0080] S202. Input the second intermediate layer signal into the CNN sequence and the first channel respectively, and output the time domain signal and the frequency domain signal.

[0081] The signals in the second intermediate layer are processed separately.

[0082] In the first embodiment, the signal from the second intermediate layer is fed into a CNN sequence (specifically referring to a one-dimensional convolutional layer Conv1d) and undergoes a nonlinear transformation using the ELU activation function. The CNN sequence can capture local features of the signal and output a time-domain signal.

[0083] In the second embodiment, the second intermediate layer signal is sent to the first channel for processing. Here, the first channel may refer to a module containing other processing steps, such as a bidirectional LSTM (Bi-LSTM). Bidirectional LSTM can process sequential data, capture contextual information in the signal, and output a frequency domain signal.

[0084] Ultimately, time-domain and frequency-domain signals were output, providing input for subsequent splicing operations.

[0085] S203. Input the time-domain signal and frequency-domain signal into the data splicing module to obtain the first intermediate layer signal.

[0086] In this step, the time-domain signal and the frequency-domain signal are sent to the data splicing module for processing.

[0087] The data stitching module uses a concat operation to join two signals along a certain dimension, forming a wider feature representation. After processing by the concat operation, the first intermediate layer signal is output. This signal contains comprehensive features of time and frequency domain information, providing a comprehensive information foundation for subsequent processing or analysis.

[0088] S103. Input the first intermediate layer signal into the second depressive disorder detection model to obtain the third intermediate layer signal.

[0089] Please see Figure 5 , Figure 5This is a schematic diagram of the second depressive disorder detection model provided in the embodiments of this application.

[0090] like Figure 5 As shown in the figure, the second depressive disorder detection model provided in this application embodiment includes a memory module, wherein the first input of the memory module is a first intermediate layer signal, the first output of the memory module is a first feature, the second input of the memory module is a first feature, the second input of the memory module is a second feature, and the third intermediate layer signal is used to characterize the total feature obtained by splicing the first feature and the second feature.

[0091] like Figure 5 As shown, each square represents a module within the second depressive disorder detection model, and its specific content is as follows:

[0092] LSTM: This box represents a Long Short-Term Memory (LSTM) network layer. LSTM is a special type of recurrent neural network (RNN) used for processing and predicting time series data. LSTM addresses the long-term dependency problem present in standard RNNs through gating mechanisms (forget gate, input gate, and output gate), enabling it to effectively remember key information from long-term sequences.

[0093] 1: Represents the first element or first feature output from the LSTM layer. In an LSTM network, a feature vector is output at each time step.

[0094] 2: This indicates that the element or feature corresponding to the output of "1" is input into the same LSTM layer.

[0095] 3: This indicates that the second element or feature output from the LSTM layer is obtained after the element or feature corresponding to the output of "1" is input into the same LSTM layer. The output marked "3" is input together with the output of "1" into the adder below.

[0096] + (Adder): This circular symbol represents an adder, used to add two input values. In this diagram, the adder adds the two inputs "1" and "3" to produce a new output value.

[0097] Specifically, in the second depressive disorder detection model, the extracted first intermediate layer signal is passed through an LSTM layer to obtain a feature. This feature is then input into the same LSTM layer again (to achieve the effect of reuse) to obtain a second feature. The two features are added together to output the third intermediate layer signal.

[0098] S104. Input the third intermediate layer signal into the third depression disorder detection model to obtain the depression disorder detection results of the test subject.

[0099] Please see Figure 6 , Figure 6 This is a schematic diagram of the third depressive disorder detection model provided in the embodiments of this application.

[0100] like Figure 6 As shown, the third depressive disorder detection model includes a first fully connected linear layer, a second fully connected linear layer, and a fully connected classification layer. The input of the first fully connected linear layer is the signal of the third intermediate layer, the output of the first fully connected linear layer is the input of the second fully connected linear layer, the output of the second fully connected linear layer is the input of the fully connected classification layer, and the output of the fully connected classification layer is the depressive disorder detection result.

[0101] In one optional embodiment, the first fully connected linear layer includes a first linear layer and a first exponential linear unit, the second fully connected linear layer includes a second linear layer and a second exponential linear unit, and the fully connected classification layer includes a third linear layer and a classification function.

[0102] The third depressive disorder detection model is a 1-dimensional CNN sequence.

[0103] like Figure 6 As shown, each square represents a module within the third depressive disorder detection model, and its specific content is as follows:

[0104] FC: Fully Connected Layer, also known as a linear layer, is a fundamental structure in neural networks. Each neuron in a layer is connected to all neurons in the previous layer, and through weights and biases, a linear transformation is performed to integrate features and map them to the output space. It is commonly found at the end of classification or regression tasks, such as in convolutional neural networks (CNNs), where after convolutional layers extract spatial features, FC layers convert those features into class probabilities.

[0105] ELU: Exponential Linear Unit, is an activation function. The ELU function enables the network to capture negative values ​​while avoiding the "dead neuron" problem, thereby improving the training speed and final performance of the model.

[0106] SoftMax: The SoftMax function is a function that transforms a vector of real numbers into a probability distribution, typically used in the final layer of multi-class classification problems. It outputs a probability distribution representing the probability of each class being predicted.

[0107] The results of reusing the LSTM are input into the last three linear layers, and the output is the final diagnostic result.

[0108] Please see Figure 7 , Figure 7 This is a flowchart illustrating the process of obtaining the depressive disorder detection results of a test subject, as provided in an embodiment of this application. Figure 7As shown in the embodiments of this application, the sleep staging detection method includes:

[0109] S701: Input the third intermediate layer signal into the first fully connected linear layer and output the fourth intermediate layer signal.

[0110] Third intermediate layer signal: This is the output of the previous layer (which may be a convolutional layer, pooling layer, or other fully connected layer) and serves as the input for the current step.

[0111] First fully connected linear layer: This is a fully connected layer (also known as a dense layer or linear layer) where each input node is fully connected to the output node. Additionally, this layer typically includes a non-linear activation function (such as ReLU, Sigmoid, or Tanh) to increase the model's non-linearity.

[0112] Fourth intermediate layer signal: This is the output of the first fully connected linear layer, which will be used as the input for the next step.

[0113] S702: Input the fourth intermediate layer signal into the second fully connected linear layer and output the fifth intermediate layer signal.

[0114] The second fully connected linear layer: Similar to the first fully connected linear layer, this is also a fully connected layer containing a non-linear activation function.

[0115] Fifth intermediate layer signal: This is the output of the second fully connected linear layer, which will serve as the input for the next step.

[0116] S703. Input the signal from the fifth intermediate layer into the fully connected classification layer to obtain the depressive disorder detection results of the test subject.

[0117] Fully connected classification layer: This is the last layer of the model, typically a fully connected layer with the same number of output nodes as the number of classes. This layer usually uses the Softmax activation function to transform the output into a probability distribution representing the predicted probability for each class.

[0118] Depression disorder detection result: This is the output of the fully connected classification layer, representing the probability that the test subject belongs to each depression disorder category (e.g., normal, mild depressive disorder, moderate depressive disorder, severe depressive disorder, etc.). Ultimately, the model selects the category with the highest probability as the depression disorder detection result.

[0119] In this application, the results of the depressive disorder test are divided into two categories: depressive disorder and health. The result categories can be divided according to the actual application, and no specific limitation is made here.

[0120] This model, based on physiological signals during sleep, can adaptively process physiological signals from multiple channels, reducing the complexity of data processing required by existing algorithms. The dataset used in this study contains 60 cases of overnight PSG data from 30 patients with depressive disorders and 30 healthy participants, totaling over 48,000 samples after data segmentation. The model was optimized using the Adam optimizer with a learning rate of 1e-6. The final diagnostic results for depressive disorders are shown in Table 1 below.

[0121] Table 1:

[0122]

[0123] Here, it's equivalent to splitting the original 60 cases into 48,000 samples, each of which is an independent diagnosis. The final accuracy is also calculated based on the segmented samples. It is a common practice in related fields to segment the data of the same case and input it into the deep learning model to calculate the results.

[0124] The present application provides a method and apparatus for detecting depressive disorders based on sleep monitoring. The method acquires a polysomnography (PSG) of the subject within a target time period. The PSG includes multiple physiological signals. These physiological signals are input into a first depressive disorder detection model to obtain a first intermediate layer signal. The first depressive disorder detection model includes an attention module, a CNN sequence, a first channel, and a data stitching module. The attention module acquires the attention index of each physiological signal, calculates the attention weight of each physiological signal, and integrates all attention weights into a second intermediate layer signal. The first intermediate layer signal is input into a second depressive disorder detection model to obtain a third intermediate layer signal. The third intermediate layer signal is then input into a third depressive disorder detection model to obtain the depressive disorder detection result of the subject. This application, through the introduction of an attention module and a multi-layer model structure, enables a more accurate extraction of features related to depressive disorders from polysomnography, thereby improving the accuracy of depressive disorder detection. The use of the attention module allows the model to focus on physiological signals that are more critical to the detection of depressive disorders, which helps to enhance the interpretability of the model and make the detection results more convincing. The sleep monitoring-based method for detecting depressive disorders can detect depressive symptoms at an early stage, providing doctors with timely intervention and treatment suggestions, which helps to improve the prognosis of patients. Compared with traditional methods for detecting depression, this method is based on sleep monitoring, does not require invasive examinations or treatments, and is more patient-friendly and comfortable.

[0125] Based on the same inventive concept, this application also provides a sleep monitoring-based depression detection device corresponding to the sleep monitoring-based depression detection method. Since the principle of the device in this application is similar to the sleep monitoring-based depression detection method described above, the implementation of the device can refer to the implementation of the method, and the repeated parts will not be described again.

[0126] Please see Figure 8 , Figure 8 This is a schematic diagram of the structure of the sleep monitoring-based depressive disorder detection device provided in an embodiment of this application. Figure 8 As shown, the sleep monitoring-based depressive disorder detection device 800 includes:

[0127] The acquisition module 801 is used to acquire a polysomnogram of the subject under test within a target time period, wherein the polysomnogram includes multiple physiological signals;

[0128] The first intermediate layer signal acquisition module 802 is used to input the multiple physiological signals into the first depression disorder detection model to obtain the first intermediate layer signal. The first depression disorder detection model includes an attention module, a CNN sequence, a first channel, and a data splicing module. The attention module is used to acquire the attention index of each physiological signal, calculate the attention weight of each physiological signal, and integrate all attention weights into a second intermediate layer signal. The input of the CNN sequence is the second intermediate layer signal, and the output of the CNN sequence is a time-domain signal. The input of the first channel is the second intermediate layer signal, and the output of the first channel is a frequency-domain signal. The input of the data splicing module is the time-domain signal and the frequency-domain signal, and the output of the data splicing module is the first intermediate layer signal.

[0129] The third intermediate layer signal acquisition module 803 is used to input the first intermediate layer signal into the second depression disorder detection model to obtain the third intermediate layer signal.

[0130] The depression disorder detection result module 804 is used to input the third intermediate layer signal into the third depression disorder detection model to obtain the depression disorder detection result of the test subject.

[0131] Please see Figure 9 , Figure 9 This is a schematic diagram of the structure of the electronic device provided in an embodiment of this application. Figure 9 As shown, the electronic device 300 includes a processor 310, a memory 320, and a bus 330.

[0132] The memory 320 stores machine-readable instructions executable by the processor 310. When the electronic device 300 is running, the processor 310 and the memory 320 communicate via the bus 330. When the machine-readable instructions are executed by the processor 310, they can perform the operations described above. Figure 1 The steps of the sleep monitoring-based depressive disorder detection method in the illustrated method embodiment can be found in the method embodiment for specific implementation, and will not be repeated here.

[0133] This application also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, can perform the above-described actions. Figure 1 The steps of the sleep monitoring-based depressive disorder detection method in the illustrated method embodiment can be found in the method embodiment for specific implementation, and will not be repeated here.

[0134] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0135] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. The apparatus embodiments described above are merely illustrative. For example, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. Furthermore, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Additionally, the shown or discussed mutual couplings, direct couplings, or communication connections may be through some communication interfaces; indirect couplings or communication connections between devices or units may be electrical, mechanical, or other forms.

[0136] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0137] In addition, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0138] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a processor-executable, non-volatile, computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0139] Finally, it should be noted that the above-described embodiments are merely specific implementations of this application, used to illustrate the technical solutions of this application, and not to limit them. The scope of protection of this application is not limited thereto. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that any person skilled in the art can still modify or easily conceive of changes to the technical solutions described in the foregoing embodiments, or make equivalent substitutions for some of the technical features, within the scope of the technology disclosed in this application. Such modifications, changes, or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application, and should all be covered within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A sleep monitoring-based system for detecting depressive disorders, characterized in that, include: Acquire polysomnography of the subject within a target time period. The polysomnography includes multiple physiological signals, including electroencephalogram (EEG), electrooculogram (EOG), electrocardiogram (ECG), and electromyogram (EMG). The multiple physiological signals are input into a first depressive disorder detection model to obtain a first intermediate layer signal. The first depressive disorder detection model includes an attention module, a CNN sequence, a first channel, and a data stitching module. The attention module is used to obtain the attention index of each physiological signal, calculate the attention weight of each physiological signal, and integrate all attention weights into a second intermediate layer signal. The input of the CNN sequence is the second intermediate layer signal, and the output of the CNN sequence is a time-domain signal obtained by extracting the time-domain features of the multiple physiological signals. The input of the first channel is the second intermediate layer signal, and the output of the first channel is a frequency-domain signal obtained by extracting the frequency-domain features of the multiple physiological signals. The input of the data stitching module is the time-domain signal and the frequency-domain signal, and the output of the data stitching module is the first intermediate layer signal. The first intermediate layer signal is input into the second depressive disorder detection model to obtain a third intermediate layer signal. The second depressive disorder detection model includes a memory module. The first input of the memory module is the first intermediate layer signal, the first output of the memory module is a first feature, the second input of the memory module is the first feature, the second input of the memory module is a second feature, and the third intermediate layer signal is used to characterize the total feature obtained by splicing the first feature and the second feature. The third intermediate layer signal is input into the third depressive disorder detection model to obtain the depressive disorder detection result of the test subject, and the depressive disorder detection result is either depressive disorder or healthy.

2. The system according to claim 1, characterized in that, Both the CNN sequence and the first channel are 1-dimensional CNN sequences. The input of the plurality of physiological signals into the first depressive disorder detection model to obtain the first intermediate layer signal includes: The multiple physiological signals are input into the attention module, and the second intermediate layer signal is output. The second intermediate layer signal is input into the CNN sequence and the first channel respectively, and the time domain signal and the frequency domain signal are output. The time-domain signal and the frequency-domain signal are input into the data splicing module to obtain the first intermediate layer signal.

3. The system according to claim 1, characterized in that, The third depressive disorder detection model includes a first fully connected linear layer, a second fully connected linear layer, and a fully connected classification layer. The input of the first fully connected linear layer is the signal of the third intermediate layer, the output of the first fully connected linear layer is the input of the second fully connected linear layer, the output of the second fully connected linear layer is the input of the fully connected classification layer, and the output of the fully connected classification layer is the depressive disorder detection result.

4. The system according to claim 3, characterized in that, The step of inputting the third intermediate layer signal into the third depressive disorder detection model to obtain the depressive disorder detection result of the test subject includes: The third intermediate layer signal is input into the first fully connected linear layer, and the fourth intermediate layer signal is output. The fourth intermediate layer signal is input into the second fully connected linear layer, and the fifth intermediate layer signal is output. The signal from the fifth intermediate layer is input into the fully connected classification layer to obtain the depressive disorder detection result of the subject.

5. The system according to claim 1, characterized in that, The system also includes: For the physiological signals under each channel of the polysomnography, the multiple physiological signals under that channel are divided according to a preset time to obtain multiple sleep segments. Each sleep segment includes EEG signals, EEG signals, ECG signals and EMG signals under multiple channels within the preset time. The electroencephalogram (EEG), electrooculogram (EOG), electrocardiogram (ECG), and electromyogram (EMG) signals corresponding to each sleep segment are sequentially input into the depression disorder detection model to obtain the depression disorder detection results of the subject. The depression disorder detection model includes the first depression disorder detection model, the second depression disorder detection model, and the third depression disorder detection model.

6. The system according to claim 4, characterized in that, The results of the depressive disorder test include a first result and a second result. The results of the depressive disorder test for the subject were determined in the following manner: Each sleep segment is sequentially input into the depression disorder detection model to determine the result corresponding to the test subject in that sleep segment.

7. A sleep monitoring-based device for detecting depressive disorders, characterized in that, include: The acquisition module is used to acquire the polysomnography of the subject within a target time period. The polysomnography includes multiple physiological signals, including electroencephalogram (EEG), electrooculogram (EOG), electrocardiogram (ECG), and electromyogram (EMG). A first intermediate layer signal acquisition module is used to input the multiple physiological signals into a first depressive disorder detection model to obtain a first intermediate layer signal. The first depressive disorder detection model includes an attention module, a CNN sequence, a first channel, and a data splicing module. The attention module is used to acquire the attention index of each physiological signal, calculate the attention weight of each physiological signal, and integrate all attention weights into a second intermediate layer signal. The input of the CNN sequence is the second intermediate layer signal, and the output of the CNN sequence is a time-domain signal obtained by extracting the time-domain features of the multiple physiological signals. The input of the first channel is the second intermediate layer signal, and the output of the first channel is a frequency-domain signal obtained by extracting the frequency-domain features of the multiple physiological signals. The input of the data splicing module is the time-domain signal and the frequency-domain signal, and the output of the data splicing module is the first intermediate layer signal. The third intermediate layer signal acquisition module is used to input the first intermediate layer signal into the second depression disorder detection model to obtain a third intermediate layer signal. The second depression disorder detection model includes a memory module, wherein the first input of the memory module is the first intermediate layer signal, the first output of the memory module is a first feature, the second input of the memory module is the first feature, the second input of the memory module is a second feature, and the third intermediate layer signal is used to characterize the total feature obtained by splicing the first feature and the second feature. The depression disorder detection result module is used to input the third intermediate layer signal into the third depression disorder detection model to obtain the depression disorder detection result of the test subject, wherein the depression disorder detection result is either depression disorder or health.

8. An electronic device, characterized in that, include: The system includes a processor, a memory, and a bus, wherein the memory stores machine-readable instructions executable by the processor, and when the electronic device is in operation, the processor communicates with the memory via the bus, and the processor executes the machine-readable instructions to perform the steps of the system as described in any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, performs the steps of the system as described in any one of claims 1 to 6.

Images