5 March 2021 in IEEE Transactions on Instrumentation and Measurement
The detection of obstructive sleep apnea (OSA) based on single-lead electrocardiogram (ECG) is better suited to the noninvasive needs and hardware conditions of wearable mobile devices. From previous ECG-based OSA detection methods, we can find that deep learning methods have shown great potential and advantages. However, due to the nonstationarity of sympathetic nerve signals and the complex characteristics of heart rate variability (HRV), the neural network under a single scale cannot effectively capture the features of HRV. In this study, an OSA detection method based on a multiscale dilation attention 1-D convolutional neural network (MSDA-1DCNN) and a weighted-loss time-dependent (WLTD) classification model were proposed. The introduction of dilated convolution effectively balanced the relationship between model parameters and performance. Attention mechanism technology modified the multiscale features after fusion and improved the weight of features under important channels. In the final classification part of the network, the combination of weighted cross-entropy loss function and hidden Markov model effectively alleviated the problem of data imbalance and improved the classification accuracy of the classifier. In segment identification, the accuracy, sensitivity, and specificity of the proposed method are 89.4%, 89.8%, and 89.1%, respectively; as for individual identification, the accuracy of that achieved 100%. The results demonstrated that the method proposed in this study can identify sleep apnea accurately.
18 August 2020 in IEEE Transactions on Instrumentation and Measurement
Sleep apnea (SA) is a harmful respiratory disorder that has caused widespread concern around the world. Considering that electrocardiogram (ECG)-based SA diagnostic methods were effective and human-friendly, many machine learning or deep learning methods based on ECG have been proposed by prior works. However, these methods are based on feature engineering or supervised and semi-supervised learning techniques, and the feature sets are always incomplete, subjective and highly dependent on labeled data. In addition, some related studies ignored the data imbalance problem which leads to poor performance of classifier on minority classes. In this study, a SA detection model based on frequential stacked sparse auto-encoder (FSSAE) and time-dependent cost-sensitive (TDCS) classification model was proposed. The FSSAE extracts feature set automatically with unsupervised learning technique, and the TDCS classification model is proposed by combining the hidden Markov model (HMM) and the Metacost algorithm to improve the performance of the classifier by considering temporal dependence and the imbalance problem. In the test set, the result of per-segment classification achieved 85.1%, 86.2% and 84.4% for accuracy, sensitivity and specificity respectively, proving that our method is helpful for SA detection.