HCP Live
Contagion LiveCGT LiveNeurology LiveHCP LiveOncology LiveContemporary PediatricsContemporary OBGYNEndocrinology NetworkPractical CardiologyRheumatology Netowrk

Photoplethysmography Differentiates Obstructive Sleep Apnea Severity Categories

The differences between the obstructive sleep apnea categories became more apparent when a shorter epoch-to-epoch interval was used to assess this deep learning-based signal.

An automated deep learning-based solution developed using a photoplethysmography (PPG)-signal was found to be effective in differentiating between obstructive sleep apnea (OSA) severity categories based on sleep continuity.

Senior author Henri Korkalainen, MSc, PhD, postdoctoral researcher, University of Eastern Finland, and colleagues created a PPG-based automatic sleep staging model using a combination of convolutional and recurrent neural networks. Sleep was tested separately on 3-class (wake/non-rapid eye movement [NREM]/REM), 4-class (wake/narcolepsy type 1 [N1]+N2/N3/REM) and 5-class (wake/N1/N2/N3/REM) classifications. Using 2 clinical datasets from Israel (n = 2149) and Australia (n = 877), investigators assessed the relationship between OSA severity categories and sleep fragmentation using survival analysis of mean continuous sleep.

In the 3-stage classification, the overall accuracy of the classifier was 83.3% on the test set. Precision values for each class on the test set were 0.85 (recall, 0.75) for wake, 0.85 (recall, 0.89) for NREM, and 0.78 (recall, 0.86) for REM sleep. The automatic PPG-based sleep staging showed accuracy of 74.1% on 4-stage classification, with specific class-wise test set precision values of 0.84 (recall, 0.77) for wake, 0.72 (recall, 0.79) for light sleep (N1+N2), 0.71 (recall, 0.57) for N3, and 0.76 (recall, 0.83) for REM sleep.

The overall accuracy of the classifier was 68.7% in the 5-stage classification. Precision values for each stage in the test set were 0.81 (recall, 0.82) for wake, 0.44 (recall, 0.14) for N1, 0.62 (recall, 0.75) for N2, 0.71 (recall, 0.58) for N3, and 0.77 (recall, 0.87) for REM.

READ MORE: JZP-258 Granted Orphan Designation in Narcolepsy

Korkalainen and colleagues applied overlapping PPG epochs to artificially obtain denser hypnograms for better identification of fragmented sleep. With both manually and automatically scored hypnograms, the hazard ratios (HRs) for decreased mean continuous sleep compared to the non-OSA group were larger when the OSA severity increased. “This supports the first hypothesis that the automated PPG-based sleep staging models can be used to differentiate between the OSA severity categories in terms of sleep continuity. Thus, it can be reasoned that the PPG-signal captures the sleep fragmentation induced by OSA-related breathing obstructions,” the study authors wrote.

When the epoch-to-epoch intervals were decreased, the differences between the HRs of different OSA severity groups increased. The HRs for PSG-based manually scored hypnograms were 1.18, 1.78, and 2.90 for mild, moderate, and severe OSA, respectively. With the PPG-based automatic scoring with 5-second epoch-to-epoch interval, the corresponding HRs were 1.70, 3.30, and 8.11, respectively. Kaplan-Meier plots also showed that the mean continuous sleep estimated by the deep learning models decreased drastically when the epoch-to-epoch interval was decreased.

Investigators noted that these results were in line with their second hypothesis which stated that a denser temporal resolution of the sleep staging would highlight the differences between the OSA severity categories with respect to sleep fragmentation.

REFERENCE
Huttenen R, Leppanen T, Duce B, Oksenberg A, Myllymaa S, Toyros J, Korkalainen H. Assessment of obstructive sleep apnea­–related sleep fragmentation utilizing deep learning-based sleep staging from photplethysmography. SLEEP. Published online June 5, 2021. doi: 10.1093/sleep/zsab142.