Automatic Sleep Stage Classification with Optimized Selection of EEG Channels DOI Creative Commons

Håkon Stenwig,

Andrés Soler,

Junya Furuki

et al.

bioRxiv (Cold Spring Harbor Laboratory), Journal Year: 2022, Volume and Issue: unknown

Published: June 17, 2022

Abstract Visual inspection of Polysomnography (PSG) recordings by sleep experts based on established guidelines has been the gold standard in stage classification. This approach is expensive, time consuming and mostly limited to experimental research clinical cases major disorders. Various automatic approaches scoring have emerging past years are opening way a quick computational assessment architecture that may find its clinics. With hope make fully automated process clinics, we report here an ensemble algorithm aims at not only predicting stages but doing so with optimized minimal number EEG channels. For that, combine genetic optimization classification framework minimizes channels used machine learning quantify stages. resulted F1 score 0.793 for model 0.806 trained 10 percent unseen subject, both 3 The combination extremely randomized trees MiniRocket classifiers. was trained, validated tested night PSG data collected from 7 subjects. novelty our lies use minimum information needed scoring, systematic search concurrently selects optimal-minimum best performing features classifier. presented this work enable new designs devices suited studies comfort homes, easily inexpensively facilitate large populations.

Language: Английский

Automatic Sleep Stage Classification with Optimized Selection of EEG Channels DOI Creative Commons

Håkon Stenwig,

Andrés Soler,

Junya Furuki

et al.

bioRxiv (Cold Spring Harbor Laboratory), Journal Year: 2022, Volume and Issue: unknown

Published: June 17, 2022

Abstract Visual inspection of Polysomnography (PSG) recordings by sleep experts based on established guidelines has been the gold standard in stage classification. This approach is expensive, time consuming and mostly limited to experimental research clinical cases major disorders. Various automatic approaches scoring have emerging past years are opening way a quick computational assessment architecture that may find its clinics. With hope make fully automated process clinics, we report here an ensemble algorithm aims at not only predicting stages but doing so with optimized minimal number EEG channels. For that, combine genetic optimization classification framework minimizes channels used machine learning quantify stages. resulted F1 score 0.793 for model 0.806 trained 10 percent unseen subject, both 3 The combination extremely randomized trees MiniRocket classifiers. was trained, validated tested night PSG data collected from 7 subjects. novelty our lies use minimum information needed scoring, systematic search concurrently selects optimal-minimum best performing features classifier. presented this work enable new designs devices suited studies comfort homes, easily inexpensively facilitate large populations.

Language: Английский

Citations

0