Automatic sleep staging method based on CNN-BiGRU and multi-head self-attention mechanism
10.3969/j.issn.1005-202X.2025.04.011
- VernacularTitle:基于CNN-BiGRU和多头自注意力机制的自动睡眠分期方法
- Author:
Xiaoli ZHANG
1
;
Xizhen ZHANG
;
Dongmei LIN
;
Fuming CHEN
Author Information
1. 甘肃中医药大学医学信息工程学院,甘肃 兰州 730000;中国人民解放军联勤保障部队第九四〇医院医疗保障中心,甘肃 兰州 730050
- Publication Type:Journal Article
- Keywords:
sleep stage;
class balance;
residual network;
bidirectional gated recurrent network
- From:
Chinese Journal of Medical Physics
2025;42(4):496-504
- CountryChina
- Language:Chinese
-
Abstract:
The study aims to address the issues of class imbalance in sleep EEG data and gradient vanishing or explosion phenomena that may occur when deep networks extract more features.An improved adaptive synthetic sampling technique is firstly employed to perform data augmentation on the minority classes of sleep EEG data.Subsequently,convolutional neural networks and residual networks are utilized to learn data features,while a 3-layer bidirectional gated recurrent network is applied to explore deep temporal information and establish correlations between different sleep stages,enabling automatic feature learning and sleep cycle extraction.Finally,a multi-head self-attention mechanism is adopted to enhance the model's focus on critical parts of the sequence,thereby completing the classification of various sleep stages.Experimental results show that according to the AASM sleep staging criteria,the automatic sleep staging model integrating CNN-BiGRU and multi-head self attention achieves an overall accuracy of 90.77%and a Kappa coefficient of 0.88 on the Sleep-EDF-20 dataset after data class balancing,with the precision of N1 stage reaching 87.1%.On the Sleep-EDFx dataset,the model attains an MF1 score of 0.84 while maintaining a precision of 77.2%for N1 stage classification.These metrics demonstrate significant improvements in performance as compared with CNN-BiGRU model tested on the original dataset.When benchmarked against other related studies,the proposed architecture exhibits superior sleep stage classification accuracy.These findings collectively validate the effectiveness and generalization capability of the proposed method.
