CN-116584891-B - Sleep apnea syndrome detection method based on multi-level feature fusion

CN116584891BCN 116584891 BCN116584891 BCN 116584891BCN-116584891-B

Abstract

The invention discloses a sleep apnea syndrome detection method based on multi-level feature fusion. The method comprises the following steps of obtaining different respiratory signal data of a human body to be detected and preprocessing the respiratory signal data, inputting the preprocessed data into a TF-Res2Net module, obtaining spatial characteristics of different scales of a time-frequency domain of respiratory signals, inputting the spatial characteristics into a mixed attention module, connecting the correlation characteristics inside the obtained different respiratory signals and the cross characteristics among the different respiratory signals, inputting the correlation characteristics into a self-adaptive fusion module, and outputting a detection result of sleep apnea syndrome through a fully-connected layer with an activation function of Softmax. The invention can automatically integrate the time-frequency domain characteristics of different scales of different respiratory signals of a human body, the correlation characteristics inside a single respiratory signal and the cross characteristics among different respiratory signals, filter the redundant characteristics among different signals, strengthen the generalization capability of a model and improve the efficiency of sleep apnea syndrome detection.

Inventors

HE KEJING
Shu Shiwen
ZHUO WEILUN

Assignees

华南理工大学
广州奥滴尔科技有限公司

Dates

Publication Date: 20260512
Application Date: 20230404

Claims (4)

1. A sleep apnea syndrome detection method based on multi-level feature fusion is characterized by comprising the following steps: s1, acquiring different respiratory signal data of a human body to be detected and preprocessing the respiratory signal data to obtain preprocessed data, wherein the data preprocessing step comprises the following steps: Performing z-score normalization on different respiratory signals of a human body; Uniformly dividing different respiratory signals of the standardized human body into non-overlapping fragments with the length of more than or equal to 10 seconds; Marking the sleep apnea category of a respiratory signal segment, if the segment has a hypopnea event, an obstructive sleep apnea event or a central sleep apnea event which exceeds 10 seconds, marking the segment as hypopnea, obstructive sleep apnea or central sleep apnea respectively, otherwise marking the segment as normal; S2, inputting the preprocessed data into a TF-Res2Net module to obtain spatial characteristics of different scales of a time domain and a frequency domain of a respiratory signal, wherein the TF-Res2Net module comprises a time domain branch and a frequency domain branch, the frequency domain branch is used as a residual error of the time domain branch, and the preprocessed first respiratory signal fragment is The following steps are: Wherein, the Is the point in time in the segment of the respiratory signal The corresponding value is used to determine, Is the number of time points in the respiratory signal segment; in the time domain branch, the first respiratory signal is segmented After inputting the one-dimensional convolution layer, the average division is divided into Parts, recorded as , Representing the segmented first respiratory signal segment, When each feature subset After additional convolution operations, it is used as residual to enter the next feature subset Is a convolution operation of (1): Wherein, the For feature subsets The output after the additional convolution operation described above, The output of all feature subsets is connected and input into one-dimensional convolution layer as output of time domain branch : Wherein, the For join (connection) operations; Branching in the frequency domain, dividing the first respiratory signal into segments Performing a discrete Fourier transform (Discrete Fourier Transform, DFT), preserving the real part, noted as Repeating the operations of dividing, convolution, coupling and convolution in the time domain branch to obtain the output of the frequency domain branch : Wherein, the Representing a frequency domain segment of the segmented first respiratory signal, Representing the first respiration signal after segmentation in the frequency domain segment The subset of the features is selected to be, Feature subset for frequency domain Output after the additional convolution operation, output of time domain branch And output of frequency domain branches Adding, inputting to the maximum pooling layer (MaxPool), and obtaining the output of the TF-Res2Net module : ; S3, inputting the time-frequency domain characteristics of the respiratory signals into a Mixed attention module (Mixed-Attn-Block) to acquire the correlation characteristics inside different respiratory signals and the cross characteristics among different respiratory signals, wherein the Mixed attention module Mixed-Attn-Block comprises a single-signal self-attention module and a multi-head cross attention module among the signals, and the single-signal self-attention module is realized by a graph attention network: a1, assuming that the output of the first respiratory signal segment through the TF-Res2Net module is As input to a single signal self-attention module, where Is that The number of time points in the time frame, Is that Each time point of (3) Vector representations of (a); a2, calculating Time point of middle For the point in time Attention weight of (a) : Wherein the method comprises the steps of Is that The weight matrix to be shared inside is a weight matrix, As a function of the non-linear activation, For join (connection) operations; a3, inputting the single signal self-attention module Weighting to obtain enhancement matrix representation of internal association features of fused single signals : Wherein, the For a matrix representation of the first respiration signal, For the point in time Is used to determine the neighbor node of a node (a), In order to activate the function, Is the fusion time point With its neighbors Time point after correlation feature between Vector representations of (a); a4, assuming that the output of the second respiratory signal and the third respiratory signal through the TF-Res2Net module are respectively And Repeating the steps A1-A3 to obtain a matrix representation of the second respiration signal And a matrix representation of a third respiration signal The multi-head cross attention module between signals comprises the following operations: b1, calculating a matrix representation of the first respiratory signal according to a multi-headed cross-attention mechanism Matrix representation with a second respiration signal Is of the correlation matrix of (a) : Wherein, the For a matrix representation of the query signal, For a relational matrix representation between the queried signal and the query signal, For a matrix representation of the signal being queried, Is a relation matrix Is a dimension of (2); respectively different random initialization parameter matrices, Is the first Single head cross attention representation; Is the number of heads in the multi-head cross attention mechanism; b2 representing the matrix of the first respiration signals Association matrix Adding, performing layer normalization, inputting into a full-connection layer, and performing layer normalization to obtain a cross signal matrix representation : B3 exchanging the query signal and the queried signal, calculating a matrix representation of the second respiration signal according to a multi-headed cross-attention mechanism Matrix representation with first respiration signal Is of the correlation matrix of (a) : B4 representing the matrix of the second respiration signals Association matrix Adding, performing layer normalization, inputting into a fully-connected layer, and performing layer normalization to obtain a cross signal matrix representation of the second respiratory signal and the first respiratory signal : B5, repeating the steps B1-B4 for a plurality of times to respectively obtain cross signal matrix representations of the second respiration signals and the third respiration signals Cross signal matrix representation of a third respiration signal and a second respiration signal Cross signal matrix representation of first and third respiration signals Cross signal matrix representation of third respiration signal and first respiration signal ; S4, connecting the correlation characteristics in different respiratory signals with the cross characteristics among different respiratory signals, inputting the connection characteristics into the self-adaptive fusion module, and outputting a detection result of sleep apnea syndrome through a full-connection layer with an activation function of Softmax.
2. The method for detecting sleep apnea syndrome based on multi-level feature fusion according to claim 1, wherein in step S1, the different respiratory signals of the human body include a thoracic activity signal, i.e. a first respiratory signal, an abdominal activity signal, i.e. a second respiratory signal, and an oronasal respiratory airflow, i.e. a third respiratory signal.
3. The sleep apnea syndrome detection method based on multi-level feature fusion according to claim 1, wherein in step S4, the adaptive fusion module obtains refined features F for subsequent classification by balancing information complementation and information redundancy between features of different levels: Wherein, the As a result of the parameters that can be trained, , Is element dot product.
4. The method for detecting sleep apnea syndrome based on multi-level feature fusion according to claim 1, wherein in step S4, the sleep apnea syndrome detection result of the respiratory signal segment includes normal, hypopnea, obstructive sleep apnea and central sleep apnea, and the loss function adopted in training is cross entropy.

Description

Sleep apnea syndrome detection method based on multi-level feature fusion Technical Field The invention relates to the field of sleep monitoring, in particular to a sleep apnea syndrome detection method based on multi-level feature fusion. Background Sleep apnea Syndrome (SLEEP APNEA Syndrome, SAS) is one of the most common sleep disordered breathing diseases, embodied by periodic reduction (hypopnea) or cessation (apnea) of airflow during sleep. Currently, about 10% of middle-aged people worldwide are diagnosed with sleep apnea syndrome, and the incidence rate rises year by year. Polysomnography (Polysomnography, PSG) is the golden standard for diagnosing SAS, however, due to the high price, long-term monitoring of patients, tedious data recording and difficult interpretation, a large number of potential patients do not seek professional treatment, and timely diagnosis is obtained, so that the potential threat of complications such as daytime sleepiness, cardiovascular diseases, cognitive dysfunction and the like is caused. Therefore, the detection of the SAS has important significance for guaranteeing the health of a human body and preventing related complications. To improve patient convenience and reduce costs, various physiological signals have been studied to diagnose SAS instead of PSG, wherein respiratory signals are the signals most directly related to respiratory dynamics. The existing SAS detection method based on the breathing signals is divided into a traditional machine learning analysis method and a deep neural network model analysis method. The traditional machine learning analysis method relies on a large number of manually set feature engineering, the deep neural network model can automatically extract the features of signals, but the existing deep neural network model is mainly based on a convolutional neural network and a cyclic neural network, the spatial features of a time domain are extracted through the convolutional neural network, the time features of the time domain are extracted through the cyclic neural network, the frequency domain features of respiratory signals cannot be effectively utilized, the frequency domain features of the respiratory signals and the dependency relationship among different respiratory signals are difficult to capture, and the detection effect of sleep apnea syndrome is limited to a certain extent, for example, the method comprises the following steps: after simply connecting different respiratory signals, inputting a model comprising six convolution layers, three maximum pooling layers and one full connection layer for sleep apnea syndrome detection (R.Haidar,S.McCloskey,I.Koprinska and B.Jeffries,"Convolutional Neural Networks on Multiple Respiratory Channels to Detect Hypopnea and Obstructive Apnea Events,"2018International Joint Conference on Neural Networks(IJCNN),Rio de Janeiro,Brazil,2018,pp.1-7,doi:10.1109/IJCNN.2018.8489248.). has the defects that only the time domain characteristics of the respiratory signals are extracted, and the connection between the different respiratory signals is not considered, so that the method has great limitation. Disclosure of Invention The invention aims to provide a sleep apnea syndrome detection method based on multi-level feature fusion, which mainly solves the problems of insufficient feature extraction capability and limited detection effect of the existing method. The method can effectively extract the time-frequency domain characteristics of different respiratory signals and the potential dependency relationship among different respiratory signals, is helpful for detecting sleep apnea syndrome, and improves the detection effect of the sleep apnea syndrome. The invention is realized at least by one of the following technical schemes. A sleep apnea syndrome detection method based on multi-level feature fusion comprises the following steps: S1, acquiring different respiratory signal data of a human body to be detected, and preprocessing to obtain preprocessed data; s2, inputting the preprocessed data into a TF-Res2Net module to acquire spatial features of different scales of a time domain and a frequency domain of the respiratory signal; S3, inputting the time-frequency domain characteristics of the respiratory signals into a Mixed attention module (Mixed-Attn-Block) so as to obtain the correlation characteristics inside different respiratory signals and the cross characteristics among different respiratory signals; s4, connecting the correlation characteristics in different respiratory signals with the cross characteristics among different respiratory signals, inputting the connection characteristics into the self-adaptive fusion module, and outputting a detection result of sleep apnea syndrome through a full-connection layer with an activation function of Softmax. Further, in step S1, the different respiratory signals of the human body include a thoracic activity signal, i.e. a first respiratory signal, an a