Generalizable Sleep Staging via Multi-Level Domain Alignment
Jiquan Wang, Sha Zhao, Haiteng Jiang, Shijian Li, Tao Li, Gang Pan
TL;DR
The paper tackles the challenge of generalizing automatic sleep staging to unseen PSG datasets, where cross-dataset domain shifts impede transfer. It introduces SleepDG, a domain-generalization framework that learns domain-invariant representations via multi-level feature alignment, combining epoch-level (mean and covariance) and sequence-level (Pearson correlation) alignment with an AE-based encoder to capture local and sequential sleep features. Evaluated on five public datasets under DG settings, SleepDG achieves state-of-the-art performance, outperforming both non-DG and existing DG methods and showing robust improvements in unseen-domain tests. The results, complemented by visualization, demonstrate that jointly aligning epoch-level and sequence-level features yields more discriminative, domain-invariant representations and enhanced generalization for clinical sleep staging.
Abstract
Automatic sleep staging is essential for sleep assessment and disorder diagnosis. Most existing methods depend on one specific dataset and are limited to be generalized to other unseen datasets, for which the training data and testing data are from the same dataset. In this paper, we introduce domain generalization into automatic sleep staging and propose the task of generalizable sleep staging which aims to improve the model generalization ability to unseen datasets. Inspired by existing domain generalization methods, we adopt the feature alignment idea and propose a framework called SleepDG to solve it. Considering both of local salient features and sequential features are important for sleep staging, we propose a Multi-level Feature Alignment combining epoch-level and sequence-level feature alignment to learn domain-invariant feature representations. Specifically, we design an Epoch-level Feature Alignment to align the feature distribution of each single sleep epoch among different domains, and a Sequence-level Feature Alignment to minimize the discrepancy of sequential features among different domains. SleepDG is validated on five public datasets, achieving the state-of-the-art performance.