STDA-Net: Spectrogram-Based Domain Adaptation for cross-dataset Sleep Stage Classification

2026-05-07

AI Summary

This study addresses the challenge of model generalization in cross-dataset sleep staging, which is hindered by discrepancies in EEG channel configurations, sampling rates, recording environments, and subject populations. To overcome this, the authors propose STDA-Net, a novel framework that first transforms EEG signals into two-dimensional spectrograms, then leverages CNNs to extract spatial features and BiLSTMs to capture temporal dynamics. Crucially, an unsupervised domain-adversarial neural network (DANN) is integrated to align feature distributions between source and target domains without requiring labeled data from the target domain. Evaluated across six transfer settings involving the Sleep-EDF, SHHS-1, and SHHS-2 datasets, the method achieves an average accuracy of 89.03% and a macro F1-score of 87.64%, significantly outperforming existing one-dimensional approaches and demonstrating superior generalization and robustness.

Abstract

Accurate sleep stage classification across datasets remains challenging due to variability in EEG channel montages, sampling rates, recording environments, and subject populations. Although deep learning has shown considerable promise for automated sleep staging, most existing cross-dataset methods rely on one-dimensional EEG signal representations, whereas the use of two-dimensional spectrogram-based inputs within an unsupervised domain adaptation framework has remained largely unexplored. Here, we propose STDA-Net (Spectrogram-based Temporal Domain Adaptation Network), a framework that combines a convolutional neural network (CNN) for spectrogram-based feature extraction, a bidirectional long short-term memory (BiLSTM) module for temporal modeling of sleep dynamics, and a domain-adversarial neural network (DANN) for source-to-target feature alignment without requiring any labeled target-domain data during training. Experiments are conducted on three publicly available datasets Sleep-EDF, SHHS-1, and SHHS-2 under six cross-dataset transfer settings. Results show that the proposed framework achieves an average accuracy of 89.03% and an average macro F1-score of 87.64%, consistently outperforming existing 1D baseline methods in terms of balanced classification performance, with substantially lower variance across five independent runs, indicating improved stability and reproducibility. Overall, these findings demonstrate that 2D spectrogram-based representations, combined with temporal modeling and adversarial domain adaptation, provide a robust and competitive alternative to conventional 1D EEG inputs for cross-dataset sleep staging.

Problem

• sleep stage classification
• cross-dataset
• domain adaptation
• EEG variability
• unsupervised learning

Innovation

• spectrogram-based representation
• unsupervised domain adaptation
• cross-dataset sleep staging
• domain-adversarial neural network
• temporal modeling