Sleep stage classification is a fundamental but cumbersome task in sleep analysis. To score the sleep stage automatically, this study presents a stage classification method based on a two-stage neural network. The feature learning stage as the first stage can fuse network trained features with traditional hand-crafted features. A recurrent neural network (RNN) in the second stage is fully utilized for learning temporal information between sleep epochs and obtaining classification results. To solve serious sample imbalance problem, a novel pre-training process combined with data augmentation was introduced. The proposed method was evaluated by two public databases, the Sleep-EDF and Sleep Apnea (SA). The proposed method can achieve the F1-score and Kappa coefficient of 0.806 and 0.80 for healthy subjects, respectively, and achieve 0.790 and 0.74 for the subjects with suspect sleep disorders, respectively. The results show that the method can achieve better performance compared to the state-of-the-art methods for the same databases. Model analysis displayed that the combination of the hand-crafted features and network trained features can improve the classification performance via the comparison experiments. In addition, the RNN is a good choice for learning temporal information in sleep epochs. Besides, the pre-training process with data augmentation is verified that can reduce the impact of sample imbalance. The proposed model has potential to exploit sleep information comprehensively. INDEX TERMS Sleep stage classification, feature learning, sequence learning, EEG signal. JIAHAO FAN received the B.S. degree in communication engineering in 2013, and the M.S. degree in electronic and communication engineering from Jilin University, in 2016. He is currently pursuing the Ph.D. degree with the

No abstract

Objective. Automatic sleep staging models suffer from an inherent class imbalance problem (CIP), which hinders the classifiers from achieving a better performance. To address this issue, we systematically studied sleep electroencephalogram data augmentation (DA) approaches. Furthermore, we modified and transferred novel DA approaches from related research fields, yielding new efficient ways to enhance sleep datasets. Approach. This study covers five DA methods, including repeating minority classes, morphological change, signal segmentation and recombination, dataset-to-dataset transfer, as well as generative adversarial network (GAN). We evaluated these mentioned DA methods by a sleep staging model on two datasets, the Montreal archive of sleep studies (MASS) and Sleep-EDF. We used a classification model with a typical convolutional neural network architecture to evaluate the effectiveness of the mentioned DA approaches. We also conducted a comprehensive analysis of these methods. Main results. The classification results showed that DA methods, especially DA by GAN, significantly improved the total classification performance in comparison with the baseline. The improvement of accuracy, F1 score and Cohen Kappa coefficient range from 0.90% to 3.79%, 0.73% to 3.48%, 2.61% to 5.43% on MASS and 1.36% to 4.79%, 1.47% to 4.23%, 2.22% to 4.04% on Sleep-EDF, respectively. DA methods improved the classification performance in most cases, whereas the performance of class N1 showed a subtle degradation in the F1 scores. Significance. Overall, our study proved that DA approaches are efficient in alleviating CIP lying in sleep staging tasks. Meanwhile, this study provided avenues for further improving the sleep staging accuracy using DA methods.

Objective. Currently, the automatic sleep staging methods mainly face two problems: the first problem is that although the algorithms which use electroencephalogram (EEG) signals perform well, acquiring EEG signals is complicated and uncomfortable; the second problem is that if the methods utilize physiological signals collected by user-friendly devices, such as cardiorespiratory signals, whose accuracies are hard to be accepted by clinicians, although the employed signals are easy and comfortable to acquire. Approach. To overcome the two issues, an automatic sleep staging method is proposed by developing a hierarchical sequential neural network to process only the electrooculogram (EOG) and R-R interval (RR) signals. The two signals are convenient and comfortable to acquire. The proposed network mainly contains two parts: comprehensive feature learning and sequence learning. The first part extracts hand-crafted features, and network trained features are simultaneously learned by a two-scale network. Then the two kinds of features are fused. The second part utilized a two-flow recurrent neural network (RNN) to learn temporal information between sleep epochs. Main results. The proposed method was evaluated on 86 subjects from two public databases, the Montreal archive of sleep studies (MASS) and sleep apnea (SA). The proposed method can discriminate five sleep stages with the F1-score of 0.781 and 0.740 for MASS and SA, respectively. And discriminate four stages with the F1-score of 0.858 and 0.802 for MASS and SA, respectively. Significance. The proposed method can achieve comparable performance as using EEG signals for sleep staging and have better performance compared to five related state-of-the-art methods. Model analysis displayed that the network can learn effective features and sequence information from EOG and RR signals. In summary, the proposed method is promising to enable new sleep monitoring in a more convenient way while having a good performance on sleep staging.

No abstract