A Comparison of Signal Combinations for Deep Learning-Based Simultaneous Sleep Staging and Respiratory Event Detection

Riku Huttunen; Timo Leppanen; Brett Duce; Erna S. Arnardottir; Sami Nikkonen; Sami Myllymaa; Juha Toyras; Henri Korkalainen

doi:10.1109/TBME.2022.3225268

A Comparison of Signal Combinations for Deep Learning-Based Simultaneous Sleep Staging and Respiratory Event Detection

Riku Huttunen, Timo Leppanen, Brett Duce, Erna S. Arnardottir, Sami Nikkonen, Sami Myllymaa, Juha Toyras, Henri Korkalainen

Reykjavik University

Research output: Contribution to journal › Article › peer-review

Abstract

Objective: Obstructive sleep apnea (OSA) is diagnosed using the apnea-hypopnea index (AHI), which is the average number of respiratory events per hour of sleep. Recently, machine learning algorithms for automatic AHI assessment have been developed, but many of them do not consider the individual sleep stages or events. In this study, we aimed to develop a deep learning model to simultaneously score both sleep stages and respiratory events. The hypothesis was that the scoring and subsequent AHI calculation could be performed utilizing pulse oximetry data only. Methods: Polysomnography recordings of 877 individuals with suspected OSA were used to train the deep learning models. The same architecture was trained with three different input signal combinations (model 1: photoplethysmogram (PPG) and oxygen saturation (SpO<inline-formula><tex-math notation="LaTeX">$_{2}$</tex-math></inline-formula>); model 2: PPG, SpO<inline-formula><tex-math notation="LaTeX">$_{2}$</tex-math></inline-formula>, and nasal pressure; model 3: SpO<inline-formula><tex-math notation="LaTeX">$_{2}$</tex-math></inline-formula>, nasal pressure, electroencephalogram (EEG), oronasal thermocouple, and respiratory belts). Results: Model 1 reached comparative performance with models 2 and 3 for estimating the AHI (model 1 intraclass correlation coefficient (ICC)=0.946; model 2 ICC=0.931; model 3 ICC=0.945), and REM-AHI (model 1 ICC=0.912; model 2 ICC=0.921; model 3 ICC=0.883). The automatic sleep staging accuracies (wake/N1/N2/N3/REM) were 69%, 70%, and 79% with models 1, 2, and 3, respectively. Conclusion: AHI can be estimated using pulse oximetry-based automatic scoring. Explicit scoring of sleep stages and respiratory events allows visual validation of the automatic analysis, and provides information on OSA phenotypes. Significance: Automatic scoring of sleep stages and respiratory events with a simple pulse oximetry setup could allow cost-effective, large-scale screening of OSA.

Original language	English
Pages (from-to)	1-12
Number of pages	12
Journal	IEEE Transactions on Biomedical Engineering
DOIs	https://doi.org/10.1109/TBME.2022.3225268
Publication status	Published - 1 Jan 2022

Bibliographical note

Publisher Copyright: Author

Other keywords

Brain modeling
Deep learning
Hospitals
Indexes
Rapid eye movement sleep
Recording
Sleep
Sleep apnea
obstructive sleep apnea
pulse oximetry
sleep staging

Access to Document

10.1109/TBME.2022.3225268

Cite this

@article{80343a6e470f451e8420df758ccf9522,

title = "A Comparison of Signal Combinations for Deep Learning-Based Simultaneous Sleep Staging and Respiratory Event Detection",

abstract = "Objective: Obstructive sleep apnea (OSA) is diagnosed using the apnea-hypopnea index (AHI), which is the average number of respiratory events per hour of sleep. Recently, machine learning algorithms for automatic AHI assessment have been developed, but many of them do not consider the individual sleep stages or events. In this study, we aimed to develop a deep learning model to simultaneously score both sleep stages and respiratory events. The hypothesis was that the scoring and subsequent AHI calculation could be performed utilizing pulse oximetry data only. Methods: Polysomnography recordings of 877 individuals with suspected OSA were used to train the deep learning models. The same architecture was trained with three different input signal combinations (model 1: photoplethysmogram (PPG) and oxygen saturation (SpO\$\_\{2\}\$); model 2: PPG, SpO\$\_\{2\}\$, and nasal pressure; model 3: SpO\$\_\{2\}\$, nasal pressure, electroencephalogram (EEG), oronasal thermocouple, and respiratory belts). Results: Model 1 reached comparative performance with models 2 and 3 for estimating the AHI (model 1 intraclass correlation coefficient (ICC)=0.946; model 2 ICC=0.931; model 3 ICC=0.945), and REM-AHI (model 1 ICC=0.912; model 2 ICC=0.921; model 3 ICC=0.883). The automatic sleep staging accuracies (wake/N1/N2/N3/REM) were 69\%, 70\%, and 79\% with models 1, 2, and 3, respectively. Conclusion: AHI can be estimated using pulse oximetry-based automatic scoring. Explicit scoring of sleep stages and respiratory events allows visual validation of the automatic analysis, and provides information on OSA phenotypes. Significance: Automatic scoring of sleep stages and respiratory events with a simple pulse oximetry setup could allow cost-effective, large-scale screening of OSA.",

keywords = "Brain modeling, Deep learning, Hospitals, Indexes, Rapid eye movement sleep, Recording, Sleep, Sleep apnea, obstructive sleep apnea, pulse oximetry, sleep staging",

author = "Riku Huttunen and Timo Leppanen and Brett Duce and Arnardottir, \{Erna S.\} and Sami Nikkonen and Sami Myllymaa and Juha Toyras and Henri Korkalainen",

note = "Publisher Copyright: Author",

year = "2022",

month = jan,

day = "1",

doi = "10.1109/TBME.2022.3225268",

language = "English",

pages = "1--12",

journal = "IEEE Transactions on Biomedical Engineering",

issn = "0018-9294",

publisher = "IEEE Computer Society",

}

TY - JOUR

T1 - A Comparison of Signal Combinations for Deep Learning-Based Simultaneous Sleep Staging and Respiratory Event Detection

AU - Huttunen, Riku

AU - Leppanen, Timo

AU - Duce, Brett

AU - Arnardottir, Erna S.

AU - Nikkonen, Sami

AU - Myllymaa, Sami

AU - Toyras, Juha

AU - Korkalainen, Henri

N1 - Publisher Copyright: Author

PY - 2022/1/1

Y1 - 2022/1/1

N2 - Objective: Obstructive sleep apnea (OSA) is diagnosed using the apnea-hypopnea index (AHI), which is the average number of respiratory events per hour of sleep. Recently, machine learning algorithms for automatic AHI assessment have been developed, but many of them do not consider the individual sleep stages or events. In this study, we aimed to develop a deep learning model to simultaneously score both sleep stages and respiratory events. The hypothesis was that the scoring and subsequent AHI calculation could be performed utilizing pulse oximetry data only. Methods: Polysomnography recordings of 877 individuals with suspected OSA were used to train the deep learning models. The same architecture was trained with three different input signal combinations (model 1: photoplethysmogram (PPG) and oxygen saturation (SpO$_{2}$); model 2: PPG, SpO$_{2}$, and nasal pressure; model 3: SpO$_{2}$, nasal pressure, electroencephalogram (EEG), oronasal thermocouple, and respiratory belts). Results: Model 1 reached comparative performance with models 2 and 3 for estimating the AHI (model 1 intraclass correlation coefficient (ICC)=0.946; model 2 ICC=0.931; model 3 ICC=0.945), and REM-AHI (model 1 ICC=0.912; model 2 ICC=0.921; model 3 ICC=0.883). The automatic sleep staging accuracies (wake/N1/N2/N3/REM) were 69%, 70%, and 79% with models 1, 2, and 3, respectively. Conclusion: AHI can be estimated using pulse oximetry-based automatic scoring. Explicit scoring of sleep stages and respiratory events allows visual validation of the automatic analysis, and provides information on OSA phenotypes. Significance: Automatic scoring of sleep stages and respiratory events with a simple pulse oximetry setup could allow cost-effective, large-scale screening of OSA.

AB - Objective: Obstructive sleep apnea (OSA) is diagnosed using the apnea-hypopnea index (AHI), which is the average number of respiratory events per hour of sleep. Recently, machine learning algorithms for automatic AHI assessment have been developed, but many of them do not consider the individual sleep stages or events. In this study, we aimed to develop a deep learning model to simultaneously score both sleep stages and respiratory events. The hypothesis was that the scoring and subsequent AHI calculation could be performed utilizing pulse oximetry data only. Methods: Polysomnography recordings of 877 individuals with suspected OSA were used to train the deep learning models. The same architecture was trained with three different input signal combinations (model 1: photoplethysmogram (PPG) and oxygen saturation (SpO$_{2}$); model 2: PPG, SpO$_{2}$, and nasal pressure; model 3: SpO$_{2}$, nasal pressure, electroencephalogram (EEG), oronasal thermocouple, and respiratory belts). Results: Model 1 reached comparative performance with models 2 and 3 for estimating the AHI (model 1 intraclass correlation coefficient (ICC)=0.946; model 2 ICC=0.931; model 3 ICC=0.945), and REM-AHI (model 1 ICC=0.912; model 2 ICC=0.921; model 3 ICC=0.883). The automatic sleep staging accuracies (wake/N1/N2/N3/REM) were 69%, 70%, and 79% with models 1, 2, and 3, respectively. Conclusion: AHI can be estimated using pulse oximetry-based automatic scoring. Explicit scoring of sleep stages and respiratory events allows visual validation of the automatic analysis, and provides information on OSA phenotypes. Significance: Automatic scoring of sleep stages and respiratory events with a simple pulse oximetry setup could allow cost-effective, large-scale screening of OSA.

KW - Brain modeling

KW - Deep learning

KW - Hospitals

KW - Indexes

KW - Rapid eye movement sleep

KW - Recording

KW - Sleep

KW - Sleep apnea

KW - obstructive sleep apnea

KW - pulse oximetry

KW - sleep staging

UR - https://www.scopus.com/pages/publications/85144024411

U2 - 10.1109/TBME.2022.3225268

DO - 10.1109/TBME.2022.3225268

M3 - Article

SN - 0018-9294

SP - 1

EP - 12

JO - IEEE Transactions on Biomedical Engineering

JF - IEEE Transactions on Biomedical Engineering

ER -

A Comparison of Signal Combinations for Deep Learning-Based Simultaneous Sleep Staging and Respiratory Event Detection

Abstract

Bibliographical note

Other keywords

Access to Document

Fingerprint

Cite this