Sleep staging method and device based on graph structure, electronic equipment and storage medium

Through the sleep staging method based on graph structure, the morphological feature extractor and dual Graphormer network are used, combined with the sleep classifier, and the global information of polysomnography monitoring physiological signals is extracted, solving the problem of low accuracy of sleep staging results in the existing technology. , achieving higher sleep staging accuracy.

CN118986286APending Publication Date: 2024-11-22SHENZHEN TECH UNIV
View PDF 0 Cites 3 Cited by

Patent Information

Application Number
CN202411212777.4
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-30
Publication Date
2024-11-22

AI Technical Summary

Technical Problem

In the existing sleep staging methods, the detection process of polysomnography is time-consuming and subjective. The deep learning-based graph neural network leads to the accuracy of sleep staging results due to the limited number of training layers and the inadequate utilization of hidden information between nodes. Not high.

Method used

A sleep staging method based on graph structure is proposed. By obtaining the physiological signals of polysomnography monitoring, the fast Fourier transform is performed, combined with the morphological feature extractor, dual Graphormer network and sleep classifier, the graph structure model is used to extract the physiological methods of polysomnography monitoring. The global information of the signal, including spatial and temporal characteristics, is performed in sleep staging.

Benefits of technology

It effectively improves the accuracy of sleep staging results, and improves the accuracy of staging prediction of sleep stages by fully utilizing the characteristics and attention mechanism of the graph structure model.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118986286A_ABST
    Figure CN118986286A_ABST
Patent Text Reader

Abstract

The invention discloses a sleep staging method and device based on a graph structure, electronic equipment and a storage medium, and can be widely applied to the technical field of sleep analysis. According to the method, the sleep staging network model based on the graph structure is arranged, and the morphological feature extractor, the double-Grapher network and the sleep classifier are arranged in the model, so that after the polysleep monitoring physiological signal of the target object is obtained, the polysleep monitoring physiological signal is subjected to fast Fourier transform to obtain a frequency domain signal; then the polysleep monitoring physiological signals and the frequency domain signals are input into the sleep staging network model based on the graph structure together, and the sleep condition of the target object is staging according to the characteristics of the sleep staging network model based on the graph structure, so that the accuracy of the sleep staging result can be effectively improved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present application relates to the technical field of sleep analysis, and in particular to a sleep staging method and device based on a graph structure, an electronic device, and a storage medium. Background Art

[0002] In the related technology, the main sleep staging classification method used clinically is the "gold standard" polysomnography (PSG), which can monitor a variety of physiological signals. However, the detection process of this staging method is time-consuming and requires doctors to manually interpret the data, resulting in a certain degree of subjectivity in the staging results. The method of sleep staging classification based on deep learning graph neural networks is limited in the number of training layers, resulting in a slow training speed, and the graph convolutional network only performs an additional filtering on the fully connected network of each layer according to the convolution kernel with vertex relationship information, failing to fully utilize the effective hidden information between nodes, resulting in low accuracy of the sleep staging results.

[0003] In summary, the technical problems existing in the relevant technologies need to be improved. Summary of the invention

[0004] The main purpose of the embodiments of the present application is to propose a sleep staging method and device based on a graph structure, an electronic device and a storage medium, which can effectively improve the accuracy of sleep staging results.

[0005] To achieve the above objective, an embodiment of the present application provides a sleep staging method based on a graph structure, the method comprising the following steps:

[0006] Acquire polysomnography physiological signals of the target subject;

[0007] Performing fast Fourier transform on the polysomnography physiological signal to obtain a frequency domain signal;

[0008] Inputting the polysomnography physiological signal and the frequency domain signal into a sleep staging network model based on a graph structure to obtain a sleep staging result of the target object;

[0009] The sleep staging network model based on graph structure includes a morphological feature extractor, a dual Graphormer network and a sleep classifier;

[0010] The step of inputting the polysomnography physiological signal into a sleep staging network model based on a graph structure to obtain a sleep staging result of the target object includes:

[0011] Extracting a first morphological feature of the polysomnography physiological signal by the morphological feature extractor, and extracting a second morphological feature of the frequency domain signal by the morphological feature extractor;

[0012] Performing weighted addition on the first morphological feature and the second morphological feature to obtain a third morphological feature;

[0013] Inputting the third morphological feature into the dual Graphormer network to obtain global information of the polysomnography physiological signal, wherein the global information includes spatial feature information and temporal feature information;

[0014] The global information is input into the sleep classifier to obtain the sleep staging result of the target object.

[0015] In some embodiments, the morphological feature extractor includes 4 residual modules and 1 pooling layer. The 4 residual modules are connected in sequence and then connected to the pooling layer. Each of the residual modules includes a positive convolution layer and a projection shortcut connection layer. The convolution filters in the 4 residual modules are doubled in sequence.

[0016] In some embodiments, the dual graphormer network includes three deep neural network modules, which are connected in series in sequence; the first deep neural network module is connected to an adaptive functional connection graph, and the third deep neural network module is connected to a distance-based connection graph; the third morphological feature is input into the first deep neural network module; and the global information is output through the third deep neural network module.

[0017] In some embodiments, each of the deep neural network modules includes a spatial attention mechanism module and a temporal channel attention mechanism module, the adaptive functional connection map is connected to the spatial attention mechanism module in the first deep neural network module, the distance-based connection map is connected to the spatial attention mechanism module in the third deep neural network module, and the spatial attention mechanism module is connected to the temporal channel attention mechanism module.

[0018] In some embodiments, the spatial attention mechanism module includes centrality encoding, spatial encoding, and edge encoding;

[0019] The centrality encoding is used to capture the proficiency of semantic relevance and importance of physiological signals;

[0020] The spatial encoding is used to guide and increase the attention of each node and adjacent nodes in the physiological signal;

[0021] The edge coding is used to enhance the integration of edge features.

[0022] In some embodiments, the time channel attention mechanism module is used to perform an aggregation operation on the spatial information feature map generated by the spatial attention mechanism module to obtain a first time feature, which includes a second time feature and a third time feature, and sends the first time feature to the sleep classifier.

[0023] In some embodiments, inputting the global information into the sleep classifier to obtain the sleep staging result of the target object includes:

[0024] The global information is input into the sleep classifier, and the predicted probability of the target object falling into each sleep stage is predicted as the sleep staging result.

[0025] To achieve the above-mentioned purpose, another aspect of the embodiment of the present application provides a sleep staging device based on a graph structure, the device comprising:

[0026] The first module is used to obtain polysomnography physiological signals of the target object;

[0027] The second module is used to perform fast Fourier transform on the polysomnography physiological signal to obtain a frequency domain signal;

[0028] The third module is used to input the polysomnography physiological signal and the frequency domain signal into a sleep staging network model based on a graph structure to obtain a sleep staging result of the target object;

[0029] The sleep staging network model based on graph structure includes a morphological feature extractor, a dual Graphormer network and a sleep classifier;

[0030] The step of inputting the polysomnography physiological signal into a sleep staging network model based on a graph structure to obtain a sleep staging result of the target object includes:

[0031] Extracting a first morphological feature of the polysomnography physiological signal by the morphological feature extractor, and extracting a second morphological feature of the frequency domain signal by the morphological feature extractor;

[0032] Performing weighted addition on the first morphological feature and the second morphological feature to obtain a third morphological feature;

[0033] Inputting the third morphological feature into the dual Graphormer network to obtain global information of the polysomnography physiological signal, wherein the global information includes spatial feature information and temporal feature information;

[0034] The global information is input into the sleep classifier to obtain the sleep staging result of the target object.

[0035] To achieve the above-mentioned purpose, another aspect of the embodiment of the present application provides an electronic device / computer device, comprising:

[0036] at least one processor;

[0037] at least one memory for storing at least one program;

[0038] When the at least one program is executed by the at least one processor, the at least one processor implements the above method.

[0039] To achieve the above objective, another aspect of an embodiment of the present application provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and the computer program implements the above method when executed by a processor.

[0040] The embodiments of the present application include at least the following beneficial effects: The present application provides a sleep staging method and device based on a graph structure, an electronic device and a storage medium. The scheme sets up a sleep staging network model based on a graph structure, and sets a morphological feature extractor, a dual graphormer network and a sleep classifier in the model. After obtaining the polysomnographic physiological signal of the target object, the polysomnographic physiological signal is subjected to a fast Fourier transform to obtain a frequency domain signal, and then the polysomnographic physiological signal and the frequency domain signal are input together into the sleep staging network model based on the graph structure, so as to utilize the characteristics of the sleep staging network model based on the graph structure to stage the sleep condition of the target object, thereby effectively improving the accuracy of the sleep staging results. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 is a flowchart of a sleep staging method based on a graph structure provided in an embodiment of the present application;

[0042] Figure 2 It is an architecture diagram of a sleep staging network model based on a graph structure provided in an embodiment of the present application;

[0043] Figure 3 It is an architecture diagram of the spatial attention mechanism module provided in an embodiment of the present application;

[0044] Figure 4 is an architecture diagram of a time channel attention module provided in an embodiment of the present application;

[0045] Figure 5 is a structural schematic diagram of a sleep staging device based on a graph structure provided in an embodiment of the present application;

[0046] Figure 6 It is a schematic diagram of the hardware structure of the electronic device provided in the embodiment of the present application. DETAILED DESCRIPTION

[0047] In order to make the purpose, technical solutions and advantages of the present application clearer, the present application is further described in detail below in conjunction with the accompanying drawings and examples. It should be understood that the specific embodiments described herein are only used to explain the present application and are not intended to limit the present application. When the following description refers to the drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the embodiments of the present application, they are only examples of devices and methods consistent with some aspects of the embodiments of the present application.

[0048] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art to which this application belongs. The terms used herein are only for the purpose of describing the embodiments of this application and are not intended to limit this application.

[0049] In the embodiments of the present application, a sleep staging method and device based on a graph structure, an electronic device and a storage medium are provided. In the embodiments, a sleep staging network model based on a graph structure is set, and a morphological feature extractor, a dual graphormer network and a sleep classifier are set in the model. After the polysomnographic physiological signal of the target object is acquired, the polysomnographic physiological signal is fast Fourier transformed to obtain a frequency domain signal, and then the polysomnographic physiological signal and the frequency domain signal are input into the sleep staging network model based on the graph structure together, so as to utilize the characteristics of the sleep staging network model based on the graph structure to stage the sleep condition of the target object, thereby effectively improving the accuracy of the sleep staging results.

[0050] The sleep staging method based on graph structure provided in the embodiment of the present application relates to the field of sleep analysis technology. The sleep staging method based on graph structure provided in the embodiment of the present application can be applied to a terminal, can also be applied to a server, and can also be software running in a terminal or a server. In some embodiments, the terminal can be a smart phone, a tablet computer, a laptop computer, a desktop computer, a smart speaker, a smart watch, and a car terminal, etc., but is not limited to this; the server side can be configured as an independent physical server, or it can be configured as a server cluster or distributed system composed of multiple physical servers, and can also be configured as a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, CDN, and big data and artificial intelligence platforms. The server can also be a node server in a blockchain network; the software can be an application that implements a sleep staging method based on a graph structure, etc., but is not limited to the above forms.

[0051] The present application can be used in many general or special computer system environments or configurations. For example: personal computers, server computers, handheld or portable devices, tablet devices, multiprocessor systems, microprocessor-based systems, set-top boxes, programmable consumer electronics, network PCs, minicomputers, mainframe computers, distributed computing environments including any of the above systems or devices, etc. The present application can be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform specific tasks or implement specific abstract data types. The present application can also be practiced in distributed computing environments, in which tasks are performed by remote processing devices connected through a communication network. In a distributed computing environment, program modules can be located in local and remote computer storage media including storage devices.

[0052] Figure 1 is an optional flowchart of the sleep staging method based on graph structure provided in an embodiment of the present application. Figure 1 The method may include but is not limited to steps S110 to S130:

[0053] Step S110, obtaining polysomnography physiological signals of the target object;

[0054] Step S120, performing fast Fourier transform on the polysomnography physiological signal to obtain a frequency domain signal;

[0055] Step S130: inputting the polysomnography physiological signal and the frequency domain signal into the sleep staging network model based on the graph structure to obtain the sleep staging result of the target object.

[0056] In the embodiment of the present application, polysomnography physiological signals may include but are not limited to electroencephalogram (EEG), electromyography (EMG), electrocardiogram (ECG) and electrooculogram (EOG). Figure 2As shown, the sleep staging network model based on graph structure (DGraphomer-SleepNet network) includes a morphological feature extractor (MSFE), a dual Graphormer network and a sleep classifier (FC). After obtaining the polysomnographic physiological signal, the polysomnographic physiological signal is divided into two paths and input into the morphological feature extractor. One path directly extracts the time domain information from the original polysomnographic physiological signal as the first morphological feature through the morphological feature extractor, and the other path uses the fast Fourier transform to convert the polysomnographic physiological signal into a frequency domain signal, and then extracts the second morphological feature from the frequency domain signal through the morphological feature extractor. The first morphological feature and the second morphological feature are weightedly added to obtain the third morphological feature, and then the third morphological feature is input into the dual Graphormer network to obtain global information including spatial feature information and temporal feature information. The global information is then sent to the sleep classifier and output to obtain the sleep staging result of the target object.

[0057] In the embodiment of the present application, Figure 2 Taking the DGraphomer-SleepNet network with the structure shown in the figure as an example, the polysomnographic physiological signals (PSG) of patients who have stayed overnight are marked as S = {S1, S2, ···, S M}, where M is the number of individuals participating in the sleep monitoring registration. Si is the PSG signal for monitoring the individual's nighttime sleep. Each person's nighttime PSG signal Si consists of multi-channel physiological signals such as EEG, EOG, EMG, and ECG. Given the sleep stage classification task, each nighttime PSG signal Si is segmented into {Z (1) , Z (2) ,···,Z (N)} segment, with a length of 30 seconds, where N is the number of PSG segments at night. Here each PSG segment can be defined as Z(i) = {Z (i) 1 ,Z (i) 2 ,···,Z (i) L}, where x j (i) represents the PSG signal of the jth channel of the i-th PSG segment, and L is the number of physiological signal channels. Each PSG segment X (i)Assign a sleep stage label y(i), where y(i)∈{1,2,3,4,5}, corresponding to the five sleep stages of W, N1, N2, N3, and REM respectively. Therefore, the training data set of Graphomer-SleepNet can be expressed as D={(Z(i), y(i))|i=1,2,···,N}. Then, using the adjacent segment information, the PSG segment sequence {Z(iT),···,Z(i),··,Z(i+T)} is used as the input of the DGraphomer-SleepNet network, and the sleep label y(i) of the PSG segment Z(i) is used as the label of the corresponding sequential PSG segment to train the network model. Where T is the length of the PSG segment, and in this embodiment, T is set to 2.

[0058] In an embodiment of the present application, the morphological feature extractor is a shared feature extractor, which uses a one-dimensional residual network as a backbone. The extractor has a residual block composition of a projection shortcut, which is mainly used to learn information in the time domain and frequency domain. Specifically, the residual neural network (ResNet) with a projection shortcut can achieve better performance than the same shortcut connection. The two input streams of the morphological feature extractor (MSFE) based on the one-dimensional ResNet are shared, one of the two inputs is the original ECG signal, and the other is the frequency domain signal processed by the fast Fourier transform (FFT) algorithm. It can be understood that the morphological feature extractor of this embodiment includes 4 residual modules and 1 common pooling layer. The 4 residual modules are connected in sequence and then connected to the common pooling layer. Each residual module includes two positive convolution layers with a kernel size of 25 and a projection shortcut connection layer. The projection shortcut connection layer can also be a convolution layer with a kernel size of 1. The convolution filters in the four residual modules are doubled successively. For example, when the number of convolution filters in the first residual module is set to 16, the number of convolution filters in the next three residual modules is 32, 64 and 128 respectively. In this embodiment, the morphological features of each channel are extracted from a segmented psg-epoch sequence, which can be mathematically defined as shown in Formula 1:

[0059] F i =ωF MSFE (Z i ′)+(1-ω)F MSFE (F FFT (Z i ′)) Formula 1;

[0060] In the formula, ω is a weight learning parameter. MSFE (·) is a mapping function of MSFE, F FFT (·) is the mapping function of the FFT algorithm.

[0061] In the embodiments of the present application, Figure 2 As shown, the dual Graphormer network is based on Graphormer. In order to better extract the spatial information of multimodal physiological signals, this embodiment reconstructs the graph-based spatial attention mechanism (GSAttn). In order to extract temporal features, this embodiment introduces temporal channel attention (TCAttn) after GSAtn. Specifically, the topological map provides extensive spatial information for physiological signals and produces promising results. Therefore, in this embodiment, two topological brain maps, an adaptive functional connectivity map (AG) and a distance-based connectivity map (DG), are constructed. The use of AG and DG in this embodiment can help the dual Graphormer network to fully capture the spatial relationships of the brain. Specifically, AG is generated using a data-driven method that utilizes the feature correlations between different channels of physiological signals. DG is constructed based on the distance between brain regions, so as the distance between brain regions decreases, the impact will intensify. It can be understood that the dual Graphormer network of this embodiment is as follows Figure 2 As shown, it includes three deep neural network modules (Transformer modules), and the three deep neural network modules are connected in series. In order to better concentrate partial information, this embodiment embeds the adaptive functional connection map (AG) into the first deep neural network module, and embeds the distance-based connection map (DG) into the third deep neural network module; at the same time, the second deep neural network module is also merged to maintain the original structure and aggregate global information.

[0062] In the embodiments of the present application, Figure 2 As shown, each deep neural network module includes a spatial attention mechanism module (Spatial Temporal Attention, GSAtn) and a temporal channel attention mechanism module (ChannelAttention, TCAtn), the adaptive functional connection map is connected to the spatial attention mechanism module in the first deep neural network module, the distance-based connection map is connected to the spatial attention mechanism module in the third deep neural network module, and the spatial attention mechanism module is connected to the temporal channel attention mechanism module.

[0063] In the embodiment of the present application, in view of the fact that the deep neural network module is good at processing sequential data, the present embodiment integrates graph structure information into the deep neural network module to guide the classification of sleep stages. Specifically, the present embodiment introduces centrality coding, spatial coding, and edge coding in the spatial attention mechanism module. It can be understood that the attention distribution is determined by the semantic association between nodes, which is crucial to understanding the graph. Traditional attention calculations often ignore the measurement of node importance. To solve this problem, the present embodiment introduces centrality coding, expressed as CAG and CDG, such as Figure 3As shown in Figure 2, centrality encoding assigns two real-valued embedding vectors to each node based on its in-degree and out-degree. This encoding is seamlessly added to the input node features, so that the proficiency of semantic relevance and physiological signal importance can be captured by centrality encoding.

[0064] It can be understood that spatial encoding is used to guide and increase the attention of each node and its adjacent nodes in the physiological signal. Specifically, for any graph G, this embodiment considers a function V×V→R represents v in graph g i ,v j In this embodiment, if two nodes are connected, select As v i ,v j If not, the output is set to a specified value, i.e. -1, based on experience. This embodiment assigns a learnable scalar to each feasible output, which will be used as a bias term in the attention module. By using Each node can adaptively pay attention to all other nodes based on the structural information of the graph. is learned as a relative i ,v j ), so that each node can pay more attention to the nodes near it and less attention to the nodes far away from it.

[0065] It can be understood that edge coding is used to enhance the integration of edge features. Specifically, for each ordered node pair (v i ,v j ), in order to enhance the integration of edge features in the attention layer, this embodiment finds i to v j Shortest Path SP ij =(e1,e2,...,e N ) (one of them), the average of the dot products of edge features and the learnable embedding along the path are calculated as edge encoding. This edge encoding is added to the attention module as a bias term that includes edge features.

[0066] In the embodiments of the present application, Figure 3 As shown in Figure 1, the spatial attention mechanism module is used to explore the independence of the temporal dependency and channel dependency of the signal. According to the attention calculation paradigm, the query vector and the key vector have the same dimension as the input tensor. In order to simplify the complexity of the attention calculation and capture the spatiotemporal correlation of the attention tensor, this embodiment uses a linear transformation method to reduce F i The characteristic dimension of , specifically using two adaptive learning vectors Uq, Uk. This embodiment uses three learning parameter tensors Wq, Wk and Wv to calculate the GSAtn tensor MGSAttn Therefore, the morphological characteristics of multimodal PSG signals are obtained The spatial-temporal relationship information is shown in formula (2):

[0067]

[0068] In the formula, W o represents a learning parameter vector, S represents spatial encoding, E represents edge encoding, ⊙ represents matrix multiplication, and LN(·) represents normalization.

[0069] In the embodiment of the present application, in order to utilize the correlation between channels in the time dimension of the feature, the present embodiment uses a Figure 4 The temporal channel attention module (TCAttn) shown in FIG. Since using average pooling and maximum pooling operations to capture channel information can obtain more competitive results than using a single operation, this embodiment uses average pooling and maximum pooling operations to extract the feature map of spatial information from GSAtn. Aggregate to generate the second time feature and third time characteristics The first time feature of the sleep classifier is used as the input of the shared sleep classifier (FC) network. The size of FC is set to Where r represents the reduction ratio. After applying FC, two feature vectors are generated and And merge them together. The merged result is combined with the original feature map Multiply to get channel attention in, The calculation can be expressed as shown in formula (3):

[0070]

[0071] In the formula, σ represents the sigmoid function and ⊙ represents matrix multiplication.

[0072] In the embodiments of the present application, Figure 2 As shown, the sleep classifier of the embodiment of the present application includes a fully connected layer (FC) of 128 hidden units, through which the probability of the target object falling into each sleep stage is predicted. Specifically, the predicted probability distribution of the sleep stage is defined as shown in formula (4):

[0073]

[0074] Formula (4) uses the cross entropy loss function shown in formula (5) as the objective function:

[0075]

[0076] In the formula, N represents the total number of samples, C represents the total number of classes, and I(·) is the indicator function.

[0077] In some embodiments, this embodiment uses ten-fold cross validation on two sleep datasets, ISRUC-S1 and ISRUC-S3, to evaluate the performance of the sleep stage prediction of the graph-based sleep staging network model (DGraphomer-SleepNet) provided in the embodiment of the present application. A large number of experiments conducted on the sleep datasets of ISRUC-S1 and ISRUC-S3 show that DGraphomer-SleepNet can achieve competitive intra-domain and cross-domain results. Compared with other invention experimental data, DGraphomer-SleepNet is superior to most of the most advanced similar inventions. On ISRUC-S1, cVAN achieved an accuracy of 81.4% and an F1 of 78.8%, and on ISRUC-S3, it achieved an accuracy of 85.4% and an F1 of 84.5%. At the same time, it is noted that DGraphomer-SleepNet performs better in sleep staging on the ISRUC-S3 dataset than on the ISRUC-S1 dataset. This is because the sleep dataset on ISRUC-S3 records PSG signals of healthy subjects, while ISRUC-S1 records PSG signals of patients with sleep disorders. Compared with the PSG signals of healthy subjects, sleep disorders suppress or change some morphological patterns of the signals. The experimental results of cVAN on the ISRUC-S1 and ISRUC-S3 datasets are shown in Table 1:

[0078] Table 1

[0079]

[0080] In summary, the embodiment of the present application sets a shared morphological spectral feature extractor (MSFE) in the graph-based sleep staging network model to fully extract the time domain information and frequency domain information of multimodal physiological signals, and sets a dual Graphomer network with two adaptive attentions, GSAtn and TCAtn, to better extract the spatiotemporal characteristics of multimodal physiological signals, thereby effectively improving the accuracy of sleep staging results.

[0081] Reference Figure 5 , an embodiment of the present application provides a sleep staging device based on a graph structure, the device comprising:

[0082] The first module 510 is used to obtain polysomnography physiological signals of the target object;

[0083] The second module 520 is used to perform fast Fourier transform on the polysomnography physiological signal to obtain a frequency domain signal;

[0084] The third module 530 is used to input the polysomnography physiological signal and the frequency domain signal into the sleep staging network model based on the graph structure to obtain the sleep staging result of the target object;

[0085] Among them, the sleep staging network model based on graph structure includes morphological feature extractor, dual graphormer network and sleep classifier;

[0086] Input polysomnography physiological signals into the sleep staging network model based on graph structure to obtain the sleep staging results of the target object, including:

[0087] extracting a first morphological feature of a polysomnography physiological signal by a morphological feature extractor, and extracting a second morphological feature of a frequency domain signal by a morphological feature extractor;

[0088] Performing weighted addition on the first morphological feature and the second morphological feature to obtain a third morphological feature;

[0089] The third morphological feature is input into the dual Graphormer network to obtain the global information of the polysomnography physiological signal, which includes spatial feature information and temporal feature information;

[0090] The global information is input into the sleep classifier to obtain the sleep stage results of the target object.

[0091] It can be understood that the contents of the above method embodiments are all applicable to the present device embodiments, the functions specifically implemented by the present device embodiments are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above method embodiments.

[0092] The embodiment of the present application also provides an electronic device, the electronic device includes a memory and a processor, the memory stores a computer program, and the processor implements the above-mentioned sleep staging method based on graph structure when executing the computer program. The electronic device can be any intelligent terminal including a tablet computer, a car computer, etc.

[0093] It can be understood that the contents of the above method embodiments are all applicable to the present device embodiments, the functions specifically implemented by the present device embodiments are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above method embodiments.

[0094] See also Figure 6 , Figure 6 The hardware structure of an electronic device of another embodiment is illustrated, and the electronic device includes:

[0095] The processor 610 may be implemented by a general-purpose CPU (Central Processing Unit), a microprocessor, an application-specific integrated circuit (ASIC), or one or more integrated circuits, and is used to execute relevant programs to implement the technical solutions provided in the embodiments of the present application;

[0096] The memory 620 can be implemented in the form of a read-only memory (ROM), a static storage device, a dynamic storage device, or a random access memory (RAM). The memory 620 can store an operating system and other applications. When the technical solution provided in the embodiment of this specification is implemented by software or firmware, the relevant program code is stored in the memory 620, and the processor 610 calls and executes the sleep staging method based on the graph structure of the embodiment of this application;

[0097] Input / output interface 630, used to implement information input and output;

[0098] Communication interface 640, used to realize communication interaction between the device and other devices, which can be realized through wired mode (such as USB, network cable, etc.) or wireless mode (such as mobile network, WIFI, Bluetooth, etc.);

[0099] bus 650 , which transmits information between the various components of the device (e.g., processor 610 , memory 620 , input / output interface 630 , and communication interface 640 );

[0100] The processor 610 , the memory 620 , the input / output interface 630 , and the communication interface 640 are connected to each other in communication within the device via a bus 650 .

[0101] An embodiment of the present application further provides a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, the above-mentioned sleep staging method based on the graph structure is implemented.

[0102] It can be understood that the contents of the above method embodiments are all applicable to the present storage medium embodiments, the functions specifically implemented by the present storage medium embodiments are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above method embodiments.

[0103] The memory, as a non-transient computer-readable storage medium, can be used to store non-transient software programs and non-transient computer executable programs. In addition, the memory may include a high-speed random access memory, and may also include a non-transient memory, such as at least one disk storage device, a flash memory device, or other non-transient solid-state storage device. In some embodiments, the memory may optionally include a memory remotely disposed relative to the processor, and these remote memories may be connected to the processor via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0104] The embodiments described in the embodiments of the present application are intended to more clearly illustrate the technical solutions of the embodiments of the present application and do not constitute a limitation on the technical solutions provided in the embodiments of the present application. Those skilled in the art will appreciate that with the evolution of technology and the emergence of new application scenarios, the technical solutions provided in the embodiments of the present application are also applicable to similar technical problems.

[0105] Those skilled in the art will appreciate that the technical solutions shown in the figures do not constitute a limitation on the embodiments of the present application, and may include more or fewer steps than shown in the figures, or a combination of certain steps, or different steps.

[0106] The device embodiments described above are merely illustrative, and the units described as separate components may or may not be physically separated, that is, they may be located in one place or distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0107] Those skilled in the art will appreciate that all or some of the steps in the methods disclosed above, and the functional modules / units in the systems and devices may be implemented as software, firmware, hardware, or a suitable combination thereof.

[0108] The terms "first", "second", "third", "fourth", etc. (if any) in the specification of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0109] It should be understood that in the present application, "at least one (item)" means one or more, and "plurality" means two or more. "And / or" is used to describe the association relationship of associated objects, indicating that three relationships may exist. For example, "A and / or B" can mean: only A exists, only B exists, and A and B exist at the same time, where A and B can be singular or plural. The character " / " generally indicates that the objects associated before and after are in an "or" relationship. "At least one of the following" or similar expressions refers to any combination of these items, including any combination of single or plural items. For example, at least one of a, b or c can mean: a, b, c, "a and b", "a and c", "b and c", or "a and b and c", where a, b, c can be single or multiple.

[0110] In the several embodiments provided in the present application, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the above units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0111] The units described above as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0112] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0113] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including multiple instructions to enable a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of various embodiments of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (Read-Only Memory, referred to as ROM), random access memory (Random Access Memory, referred to as RAM), disk or optical disk and other media that can store programs.

[0114] The preferred embodiments of the present invention are described above with reference to the accompanying drawings, but the scope of the rights of the present invention is not limited thereto. Any modification, equivalent substitution and improvement made by a person skilled in the art without departing from the scope and essence of the present invention should be within the scope of the rights of the present invention.

Claims

1. A sleep staging method based on graph structure, It is characterized in that The method comprises the following steps: Acquire polysomnography physiological signals of the target subject; Performing fast Fourier transform on the polysomnography physiological signal to obtain a frequency domain signal; Inputting the polysomnography physiological signal and the frequency domain signal into a sleep staging network model based on a graph structure to obtain a sleep staging result of the target object; The sleep staging network model based on graph structure includes a morphological feature extractor, a dual Graphormer network and a sleep classifier; The step of inputting the polysomnography physiological signal into a sleep staging network model based on a graph structure to obtain a sleep staging result of the target object includes: Extracting a first morphological feature of the polysomnography physiological signal by the morphological feature extractor, and extracting a second morphological feature of the frequency domain signal by the morphological feature extractor; Performing weighted addition on the first morphological feature and the second morphological feature to obtain a third morphological feature; Inputting the third morphological feature into the dual Graphormer network to obtain global information of the polysomnography physiological signal, wherein the global information includes spatial feature information and temporal feature information; The global information is input into the sleep classifier to obtain the sleep staging result of the target object.

2. The method according to claim 1, It is characterized in that The morphological feature extractor includes 4 residual modules and 1 common pooling layer. The 4 residual modules are connected in sequence and then connected to the common pooling layer. Each of the residual modules includes a positive convolution layer and a projection shortcut connection layer. The convolution filters in the 4 residual modules are doubled in sequence.

3. The method according to claim 1, It is characterized in that The dual Graphormer network includes three deep neural network modules, which are connected in series in sequence; the first deep neural network module is connected to an adaptive functional connection graph, and the third deep neural network module is connected to a distance-based connection graph; the third morphological feature is input into the first deep neural network module; and the global information is output through the third deep neural network module.

4. The method according to claim 3, It is characterized in that Each of the deep neural network modules includes a spatial attention mechanism module and a temporal channel attention mechanism module. The adaptive functional connection map is connected to the spatial attention mechanism module in the first deep neural network module, the distance-based connection map is connected to the spatial attention mechanism module in the third deep neural network module, and the spatial attention mechanism module is connected to the temporal channel attention mechanism module.

5. The method according to claim 4, It is characterized in that The spatial attention mechanism module includes centrality coding, spatial coding and edge coding; The centrality encoding is used to capture the proficiency of semantic relevance and importance of physiological signals; The spatial encoding is used to guide and increase the attention of each node and adjacent nodes in the physiological signal; The edge coding is used to enhance the integration of edge features.

6. The method according to claim 4, It is characterized in that The time channel attention mechanism module is used to perform an aggregation operation on the spatial information feature map generated by the spatial attention mechanism module to obtain a first time feature, where the first time feature includes a second time feature and a third time feature, and the first time feature is sent to the sleep classifier.

7. The method according to claim 1, It is characterized in that The step of inputting the global information into the sleep classifier to obtain the sleep staging result of the target object includes: The global information is input into the sleep classifier, and the predicted probability of the target object falling into each sleep stage is predicted as the sleep staging result.

8. A sleep staging device based on graph structure, It is characterized in that The device comprises: The first module is used to obtain polysomnography physiological signals of the target object; The second module is used to perform fast Fourier transform on the polysomnography physiological signal to obtain a frequency domain signal; The third module is used to input the polysomnography physiological signal and the frequency domain signal into a sleep staging network model based on a graph structure to obtain a sleep staging result of the target object; The sleep staging network model based on graph structure includes a morphological feature extractor, a dual Graphormer network and a sleep classifier; The step of inputting the polysomnography physiological signal into a sleep staging network model based on a graph structure to obtain a sleep staging result of the target object includes: Extracting a first morphological feature of the polysomnography physiological signal by the morphological feature extractor, and extracting a second morphological feature of the frequency domain signal by the morphological feature extractor; Performing weighted addition on the first morphological feature and the second morphological feature to obtain a third morphological feature; Inputting the third morphological feature into the dual Graphormer network to obtain global information of the polysomnography physiological signal, wherein the global information includes spatial feature information and temporal feature information; The global information is input into the sleep classifier to obtain the sleep staging result of the target object.

9. An electronic device / computer device, It is characterized in that include: at least one processor; at least one memory for storing at least one program; When the at least one program is executed by the at least one processor, the at least one processor implements the method according to any one of claims 1 to 7.

10. A computer-readable storage medium storing a computer program. It is characterized in that When the computer program is executed by a processor, the method according to any one of claims 1 to 7 is implemented.

Citation Information

Cited By

  • Sleep staging detection method and sleep staging detection device

    CN119719921A

  • Sleep staging method based on spatial-temporal feature coding and multi-source fusion

    CN120217200A

  • A sleep staging method based on spatiotemporal feature coding and multi-source fusion

    CN120217200B