SleepPairNet improves pediatric sleep signal analysis using cross-attention
SleepPairNet: Learning Multimodal Pediatric Sleep Representations with Random-Pair Cross-Attention
Machine Learning
Summary
Sleep quality and sleep disorders are often studied by looking at different body signals during sleep, like breathing and brain waves. The authors created SleepPairNet, a system that learns how different signals relate to each other by looking at random pairs during training. This method helps the system understand sleep better and perform well even when tested on new data. The approach works particularly well for tasks involving breathing signals and is competitive for brain wave analysis.
What this means in practice
- •For medical device developers: Develop pediatric sleep monitoring devices that better interpret combinations of physiological signals for improved respiratory disorder detection.$Commercial implications: Allows creation of enhanced sleep diagnostic tools for children, offering more accurate and flexible analysis of multiple sleep signals.
- •For clinical data analysts: Improve analysis workflows for pediatric sleep studies by integrating learned representations from multiple physiological signal pairs.
Authors
Saurav Raj Pandey, Harlin Lee
Abstract
Analyzing physiological sleep signals is important for assessing sleep quality and detecting sleep-related disorders. Learning inter-modal relationships across physiological signals can provide richer representations for downstream sleep analysis. We introduce \textbf{SleepPairNet}, a multimodal self-supervised learning (SSL) framework that learns pairwise interactions between physiological signals through cross-attention. We propose a random modality-pair selection strategy that exposes the model to diverse signal pairs during pretraining, enabling flexibility across modality combinations at test time. Pretrained on the NCH dataset, SleepPairNet outperforms the evaluated SSL baselines on all respiratory tasks while remaining competitive on EEG-related tasks, and shows even stronger results when transferred to the external CHAT dataset.