PaperarXiv:2411.04644
wav2sleep: A Unified Multi-Modal Approach to Sleep Stage Classification from Physiological Signals
arXiv preprint (cs.LG, cs.AI); accepted ML4H 2024
Open codeOpen weights
Key results4 SOTA
All results0.742
Cohen's κ ↑
Sleep staging · MESA
wav2sleep (trained on all signals)
PPG only at test time · 4-class (Wake, Light, Deep, REM) · end-to-end training
Best comparison: SleepPPG-Net (ECG→PPG transfer) 0.724
0.78
Cohen's κ ↑
Sleep staging · SHHS
wav2sleep
ECG + thoracic respiratory effort · 4-class (Wake, Light, Deep, REM) · end-to-end training
Best comparison: Carter et al. (2024) 0.75
0.74
Cohen's κ ↑
Sleep staging · SHHS
wav2sleep
ECG only · 4-class (Wake, Light, Deep, REM) · end-to-end training
Best comparison: Sridhar et al. (2020) 0.66
0.78
Cohen's κ ↑
Sleep staging · MESA
wav2sleep
PPG + thoracic respiratory effort · 4-class (Wake, Light, Deep, REM) · end-to-end training
Best comparison: Bakker et al. (2021) 0.68
Results as reported by the authors; not independently reproduced. A SOTA star marks the best value among results reported under the same evaluation protocol; ties go to the earlier publication.
Embed this badge (4 badges for READMEs and project pages)
Markdown[](https://chatppg.com/research/papers/wav2sleep-multimodal-variable-sensor-sleep-2024#key-results)HTML<a href="https://chatppg.com/research/papers/wav2sleep-multimodal-variable-sensor-sleep-2024#key-results"><img src="https://chatppg.com/research/badges/wav2sleep--sleep-staging-mesa-cohen-kappa-ppg-only-at-test-time.svg" alt="SOTA on ChatPPG: Sleep staging · MESA" height="20"></a>
Markdown[](https://chatppg.com/research/papers/wav2sleep-multimodal-variable-sensor-sleep-2024#key-results)HTML<a href="https://chatppg.com/research/papers/wav2sleep-multimodal-variable-sensor-sleep-2024#key-results"><img src="https://chatppg.com/research/badges/wav2sleep--sleep-staging-shhs-cohen-kappa-ecg-thoracic-respiratory-effort.svg" alt="SOTA on ChatPPG: Sleep staging · SHHS" height="20"></a>
Markdown[](https://chatppg.com/research/papers/wav2sleep-multimodal-variable-sensor-sleep-2024#key-results)HTML<a href="https://chatppg.com/research/papers/wav2sleep-multimodal-variable-sensor-sleep-2024#key-results"><img src="https://chatppg.com/research/badges/wav2sleep--sleep-staging-shhs-cohen-kappa-ecg-only.svg" alt="SOTA on ChatPPG: Sleep staging · SHHS" height="20"></a>
Markdown[](https://chatppg.com/research/papers/wav2sleep-multimodal-variable-sensor-sleep-2024#key-results)HTML<a href="https://chatppg.com/research/papers/wav2sleep-multimodal-variable-sensor-sleep-2024#key-results"><img src="https://chatppg.com/research/badges/wav2sleep--sleep-staging-mesa-cohen-kappa-ppg-thoracic-respiratory-effort.svg" alt="SOTA on ChatPPG: Sleep staging · MESA" height="20"></a>
TL;DR
Unified model that accepts variable sensor sets at train and test time (addresses PPG scarcity vs ECG). Jointly trained on >10,000 overnight recordings from six public PSG datasets including SHHS and MESA; outperforms prior staging models across test-time combinations including ECG, PPG, and respiratory signals without modality-fixed architectures.
- Form factor: other
Abstract
Accurate classification of sleep stages from less obtrusive sensor measurements such as the electrocardiogram (ECG) or photoplethysmogram (PPG) could enable important applications in sleep medicine. Existing approaches to this problem have typically used deep learning models designed and trained to operate on one or more specific input signals. However, the datasets used to develop these models often do not contain the same sets of input signals. Some signals, particularly PPG, are much less prevalent than others, and this has previously been addressed with techniques such as transfer learning. Additionally, only training on one or more fixed modalities precludes cross-modal information transfer from other sources, which has proved valuable in other problem domains. To address this, we introduce wav2sleep, a unified model designed to operate on variable sets of input signals during training and inference. After jointly training on over 10,000 overnight recordings from six publicly available polysomnography datasets, including SHHS and MESA, wav2sleep outperforms existing sleep stage classification models across test-time input combinations including ECG, PPG, and respiratory signals.
Verbatim from arXiv (opens in a new tab) · arXiv metadata (abstracts are released under CC0)
Results on ChatPPG leaderboards
All leaderboardsNo results from this paper on ChatPPG leaderboards yet. Browse the leaderboards.
Tasks & signals
Tasks
Signals
Keywords
How it works
Cite this paper
APA
Carter, J. F., & Tarassenko, L. (2024). wav2sleep: A Unified Multi-Modal Approach to Sleep Stage Classification from Physiological Signals. arXiv preprint. https://doi.org/10.48550/arXiv.2411.04644
BibTeX
@misc{carter2024wav2sleep,
title = {{wav2sleep: A Unified Multi-Modal Approach to Sleep Stage Classification from Physiological Signals}},
author = {Carter, Jonathan F. and Tarassenko, Lionel},
year = {2024},
howpublished = {arXiv preprint},
doi = {10.48550/arXiv.2411.04644},
eprint = {2411.04644},
archivePrefix = {arXiv}
}Built from the hub record. Check it against the publisher before submitting.
Record updated
Added to the ChatPPG research hub
