Dataset Information

A Convolutional Neural Network Architecture to Enhance Oximetry Ability to Diagnose Pediatric Obstructive Sleep Apnea.

ABSTRACT: This study aims at assessing the usefulness of deep learning to enhance the diagnostic ability of oximetry in the context of automated detection of pediatric obstructive sleep apnea (OSA). A total of 3196 blood oxygen saturation (SpO₂) signals from children were used for this purpose. A convolutional neural network (CNN) architecture was trained using 20-min SpO₂ segments from the training set (859 subjects) to estimate the number of apneic events. CNN hyperparameters were tuned using Bayesian optimization in the validation set (1402 subjects). This model was applied to three test sets composed of 312, 392, and 231 subjects from three independent databases, in which the apnea-hypopnea index (AHI) estimated for each subject (AHI_CNN) was obtained by aggregating the output of the CNN for each 20-min SpO₂ segment. AHI_CNN outperformed the 3% oxygen desaturation index (ODI3), a clinical approach, as well as the AHI estimated by a conventional feature-engineering approach based on multi-layer perceptron (AHI_MLP). Specifically, AHI_CNN reached higher four-class Cohen's kappa in the three test databases than ODI3 (0.515 vs 0.417, 0.422 vs 0.372, and 0.423 vs 0.369) and AHI_MLP (0.515 vs 0.377, 0.422 vs 0.381, and 0.423 vs 0.306). In addition, our proposal outperformed state-of-the-art studies, particularly for the AHI severity cutoffs of 5 e/h and 10 e/h. This suggests that the information automatically learned from the SpO₂ signal by deep-learning techniques helps to enhance the diagnostic ability of oximetry in the context of pediatric OSA.

SUBMITTER: Vaquerizo-Villar F

PROVIDER: S-EPMC8460136 | biostudies-literature | 2021 Aug

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

A Convolutional Neural Network Architecture to Enhance Oximetry Ability to Diagnose Pediatric Obstructive Sleep Apnea.

Vaquerizo-Villar Fernando F Alvarez Daniel D Kheirandish-Gozal Leila L Gutierrez-Tobal Gonzalo C GC Barroso-Garcia Veronica V Santamaria-Vazquez Eduardo E Campo Felix Del FD Gozal David D Hornero Roberto R

IEEE journal of biomedical and health informatics 20210805 8

This study aims at assessing the usefulness of deep learning to enhance the diagnostic ability of oximetry in the context of automated detection of pediatric obstructive sleep apnea (OSA). A total of 3196 blood oxygen saturation (SpO<sub>2</sub>) signals from children were used for this purpose. A convolutional neural network (CNN) architecture was trained using 20-min SpO<sub>2</sub> segments from the training set (859 subjects) to estimate the number of apneic events. CNN hyperparameters were ...[more]

PMID: 33406046

Similar Datasets

Project description:ImportanceGood sleep is essential for health, yet associations between sleep and dementia risk remain incompletely understood. The Sleep and Dementia Consortium was established to study associations between polysomnography (PSG)-derived sleep and the risk of dementia and related cognitive and brain magnetic resonance imaging endophenotypes.ObjectiveTo investigate association of sleep architecture and obstructive sleep apnea (OSA) with cognitive function in the Sleep and Dementia Consortium.Design, setting, and participantsThe Sleep and Dementia Consortium curated data from 5 population-based cohorts across the US with methodologically consistent, overnight, home-based type II PSG and neuropsychological assessments over 5 years of follow-up: the Atherosclerosis Risk in Communities study, Cardiovascular Health Study, Framingham Heart Study (FHS), Osteoporotic Fractures in Men Study, and Study of Osteoporotic Fractures. Sleep metrics were harmonized centrally and then distributed to participating cohorts for cohort-specific analysis using linear regression; study-level estimates were pooled in random effects meta-analyses. Results were adjusted for demographic variables, the time between PSG and neuropsychological assessment (0-5 years), body mass index, antidepressant use, and sedative use. There were 5946 participants included in the pooled analyses without stroke or dementia. Data were analyzed from March 2020 to June 2023.ExposuresMeasures of sleep architecture and OSA derived from in-home PSG.Main outcomes and measuresThe main outcomes were global cognitive composite z scores derived from principal component analysis, with cognitive domains investigated as secondary outcomes. Higher scores indicated better performance.ResultsAcross cohorts, 5946 adults (1875 females [31.5%]; mean age range, 58-89 years) were included. The median (IQR) wake after sleep onset time ranged from 44 (27-73) to 101 (66-147) minutes, and the prevalence of moderate to severe OSA ranged from 16.9% to 28.9%. Across cohorts, higher sleep maintenance efficiency (pooled β per 1% increase, 0.08; 95% CI, 0.03 to 0.14; P < .01) and lower wake after sleep onset (pooled β per 1-min increase, -0.07; 95% CI, -0.13 to -0.01 per 1-min increase; P = .02) were associated with better global cognition. Mild to severe OSA (apnea-hypopnea index [AHI] ≥5) was associated with poorer global cognition (pooled β, -0.06; 95% CI, -0.11 to -0.01; P = .01) vs AHI less than 5; comparable results were found for moderate to severe OSA (pooled β, -0.06; 95% CI, -0.11 to -0.01; P = .02) vs AHI less than 5. Differences in sleep stages were not associated with cognition.Conclusions and relevanceThis study found that better sleep consolidation and the absence of OSA were associated with better global cognition over 5 years of follow-up. These findings suggest that the role of interventions to improve sleep for maintaining cognitive function requires investigation.

Project description:Study objectivesThe efficacy of portable-monitor (PM) sleep testing in children is not well understood. While most studies have evaluated PM in a lab setting, the utility of PM in the home environment is relatively unknown. We sought to determine whether home PM accurately diagnoses obstructive sleep apnea in adolescents and to assess patient satisfaction with home PM sleep testing.MethodsWe evaluated adolescents (age 12-18 years) with suspected obstructive sleep apnea using a PM device. In addition to in-laboratory polysomnography (PSG), all participants had PM testing performed twice, once in their home and once concurrent to in-laboratory PSG. PM was compared to PSG using 2 primary outcomes: the apnea-hypopnea index and oxygen desaturation index. All participants were approached for interview to evaluate their experience with PM sleep testing.ResultsTwenty adolescents participated. Bland-Altman analysis comparing the apnea-hypopnea index and oxygen desaturation index determined by home or in-laboratory PM to in-laboratory PSG revealed mostly agreement; however, some deviations were observed when either parameter was markedly increased. While PM testing tended to underestimate the apnea-hypopnea index, the diagnostic agreement between home PM and PSG was 80% (by the White-Westbrook method). Most preferred PM to PSG and found PM easy to very easy to set up.ConclusionsIn a small cohort of adolescents, our study supports the application of home PM in the diagnosis of suspected obstructive sleep apnea. Until studies implementing PM using larger cohorts become readily available, the findings from this preliminary study could contribute to adolescents receiving sleep apnea therapy more promptly.Clinical trial registrationRegistry: ClinicalTrials.gov; Identifier: NCT03748771. At the time of issue publication, this registration is not publicly available because the trial includes a device that is not approved or cleared for use in pediatric populations. Once the device is FDA cleared, the registration will become public.

Dataset Information

A Convolutional Neural Network Architecture to Enhance Oximetry Ability to Diagnose Pediatric Obstructive Sleep Apnea.

Publications

A Convolutional Neural Network Architecture to Enhance Oximetry Ability to Diagnose Pediatric Obstructive Sleep Apnea.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets