Addressing Wearable Sleep Tracking Inequity: A New Dataset and Novel Methods for a Population with Sleep Disorders

Will Ke Wang, Jiamu Yang, Leeor Hershkovich, Hayoung Jeong, Bill Chen, Karnika Singh, Ali R Roghanizad, Md Mobashir Hasan Shandhi, Andrew R Spector, Jessilyn Dunn
Proceedings of the fifth Conference on Health, Inference, and Learning, PMLR 248:380-396, 2024.

Abstract

Sleep is crucial for health, and recent advances in wearable technology and machine learning offer promising methods for monitoring sleep outside the clinical setting. However, sleep tracking using wearables is challenging, particularly for those with irregular sleep patterns or sleep disorders. In this study, we introduce a dataset collected from 100 patients from the Duke Sleep Disorders Center who wore an Empatica E4 smartwatch during an overnight sleep study with concurrent clinical-grade polysomnography (PSG) recording. This dataset encompasses diverse demographics and medical conditions. We further introduce a new methodology that addresses the limitations of existing modeling methods when applied on patients with sleep disorders. Namely, we address the inability of existing models to account for 1) temporal relationships while leveraging relatively small data, by introducing a LSTM post-processing method, and 2) group-wise characteristics that impact classification task performance (i.e., random effects), by ensembling mixed-effects boosted tree models. This approach was highly successful for sleep onset and wakefulness detection in this sleep disordered population, achieving an F1 score of 0.823 ± 0.019, an AUROC of 0.926 ± 0.016, and a 0.695 ± 0.025 Cohen’s Kappa. Overall, we demonstrate the utility of both the data that we collected, as well as our unique approach to address the existing gap in wearable-based sleep tracking in sleep disordered populations.

Cite this Paper

BibTeX
@InProceedings{pmlr-v248-wang24a, title = {Addressing Wearable Sleep Tracking Inequity: A New Dataset and Novel Methods for a Population with Sleep Disorders}, author = {Wang, Will Ke and Yang, Jiamu and Hershkovich, Leeor and Jeong, Hayoung and Chen, Bill and Singh, Karnika and Roghanizad, Ali R and Shandhi, Md Mobashir Hasan and Spector, Andrew R and Dunn, Jessilyn}, booktitle = {Proceedings of the fifth Conference on Health, Inference, and Learning}, pages = {380--396}, year = {2024}, editor = {Pollard, Tom and Choi, Edward and Singhal, Pankhuri and Hughes, Michael and Sizikova, Elena and Mortazavi, Bobak and Chen, Irene and Wang, Fei and Sarker, Tasmie and McDermott, Matthew and Ghassemi, Marzyeh}, volume = {248}, series = {Proceedings of Machine Learning Research}, month = {27--28 Jun}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v248/main/assets/wang24a/wang24a.pdf}, url = {https://proceedings.mlr.press/v248/wang24a.html}, abstract = {Sleep is crucial for health, and recent advances in wearable technology and machine learning offer promising methods for monitoring sleep outside the clinical setting. However, sleep tracking using wearables is challenging, particularly for those with irregular sleep patterns or sleep disorders. In this study, we introduce a dataset collected from 100 patients from the Duke Sleep Disorders Center who wore an Empatica E4 smartwatch during an overnight sleep study with concurrent clinical-grade polysomnography (PSG) recording. This dataset encompasses diverse demographics and medical conditions. We further introduce a new methodology that addresses the limitations of existing modeling methods when applied on patients with sleep disorders. Namely, we address the inability of existing models to account for 1) temporal relationships while leveraging relatively small data, by introducing a LSTM post-processing method, and 2) group-wise characteristics that impact classification task performance (i.e., random effects), by ensembling mixed-effects boosted tree models. This approach was highly successful for sleep onset and wakefulness detection in this sleep disordered population, achieving an F1 score of 0.823 ± 0.019, an AUROC of 0.926 ± 0.016, and a 0.695 ± 0.025 Cohen’s Kappa. Overall, we demonstrate the utility of both the data that we collected, as well as our unique approach to address the existing gap in wearable-based sleep tracking in sleep disordered populations.} }
Endnote
%0 Conference Paper %T Addressing Wearable Sleep Tracking Inequity: A New Dataset and Novel Methods for a Population with Sleep Disorders %A Will Ke Wang %A Jiamu Yang %A Leeor Hershkovich %A Hayoung Jeong %A Bill Chen %A Karnika Singh %A Ali R Roghanizad %A Md Mobashir Hasan Shandhi %A Andrew R Spector %A Jessilyn Dunn %B Proceedings of the fifth Conference on Health, Inference, and Learning %C Proceedings of Machine Learning Research %D 2024 %E Tom Pollard %E Edward Choi %E Pankhuri Singhal %E Michael Hughes %E Elena Sizikova %E Bobak Mortazavi %E Irene Chen %E Fei Wang %E Tasmie Sarker %E Matthew McDermott %E Marzyeh Ghassemi %F pmlr-v248-wang24a %I PMLR %P 380--396 %U https://proceedings.mlr.press/v248/wang24a.html %V 248 %X Sleep is crucial for health, and recent advances in wearable technology and machine learning offer promising methods for monitoring sleep outside the clinical setting. However, sleep tracking using wearables is challenging, particularly for those with irregular sleep patterns or sleep disorders. In this study, we introduce a dataset collected from 100 patients from the Duke Sleep Disorders Center who wore an Empatica E4 smartwatch during an overnight sleep study with concurrent clinical-grade polysomnography (PSG) recording. This dataset encompasses diverse demographics and medical conditions. We further introduce a new methodology that addresses the limitations of existing modeling methods when applied on patients with sleep disorders. Namely, we address the inability of existing models to account for 1) temporal relationships while leveraging relatively small data, by introducing a LSTM post-processing method, and 2) group-wise characteristics that impact classification task performance (i.e., random effects), by ensembling mixed-effects boosted tree models. This approach was highly successful for sleep onset and wakefulness detection in this sleep disordered population, achieving an F1 score of 0.823 ± 0.019, an AUROC of 0.926 ± 0.016, and a 0.695 ± 0.025 Cohen’s Kappa. Overall, we demonstrate the utility of both the data that we collected, as well as our unique approach to address the existing gap in wearable-based sleep tracking in sleep disordered populations.
APA
Wang, W.K., Yang, J., Hershkovich, L., Jeong, H., Chen, B., Singh, K., Roghanizad, A.R., Shandhi, M.M.H., Spector, A.R. & Dunn, J.. (2024). Addressing Wearable Sleep Tracking Inequity: A New Dataset and Novel Methods for a Population with Sleep Disorders. Proceedings of the fifth Conference on Health, Inference, and Learning, in Proceedings of Machine Learning Research 248:380-396 Available from https://proceedings.mlr.press/v248/wang24a.html.

Related Material