Citation

Skorucak J, Hertig-Godeschalk A, Achermann P, Mathis J, Schreier DR. Front. Neurosci. 2020; 14: e8.

Affiliation

Department of Neurology, Inselspital, Bern University Hospital, University of Bern, Bern, Switzerland.

Copyright

(Copyright © 2020, Frontiers Research Foundation)

DOI

10.3389/fnins.2020.00008

PMID

32038155

PMCID

PMC6990913

Abstract

Study Objectives: Microsleep episodes (MSEs) are short fragments of sleep (1-15 s) that can cause dangerous situations with potentially fatal outcomes. In the diagnostic sleep-wake and fitness-to-drive assessment, accurate and early identification of sleepiness is essential. However, in the absence of a standardised definition and a time-efficient scoring method of MSEs, these short fragments are not assessed in clinical routine. Based on data of moderately sleepy patients, we recently developed the Bern continuous and high-resolution wake-sleep (BERN) criteria for visual scoring of MSEs and corresponding machine learning algorithms for automatic MSE detection, both mainly based on the electroencephalogram (EEG). The present study aimed to investigate the relationship between automatically detected MSEs and driving performance in a driving simulator, recorded in parallel with EEG, and to assess algorithm performance for MSE detection in severely sleepy participants. Methods: Maintenance of wakefulness test (MWT) and driving simulator recordings of 18 healthy participants, before and after a full night of sleep deprivation, were retrospectively analysed. Performance of automatic detection was compared with visual MSE scoring, following the BERN criteria, in MWT recordings of 10 participants. Driving performance was measured by the standard deviation of lateral position and the occurrence of off-road events. Results: In comparison to visual scoring, automatic detection of MSEs in participants with severe sleepiness showed good performance (Cohen's kappa = 0.66). The MSE rate in the MWT correlated with the latency to the first MSE in the driving simulator (r
_s
= -0.54, p < 0.05) and with the cumulative MSE duration in the driving simulator (r
_s
= 0.62, p < 0.01). No correlations between MSE measures in the MWT and driving performance measures were found. In the driving simulator, multiple correlations between MSEs and driving performance variables were observed. Conclusion: Automatic MSE detection worked well, independent of the degree of sleepiness. The rate and the cumulative duration of MSEs could be promising sleepiness measures in both the MWT and the driving simulator. The correlations between MSEs in the driving simulator and driving performance might reflect a close and time-critical relationship between sleepiness and performance, potentially valuable for the fitness-to-drive assessment.

Copyright © 2020 Skorucak, Hertig-Godeschalk, Achermann, Mathis and Schreier.

Language: en

Keywords

driving simulator; electroencephalography; fitness to drive; machine learning; maintenance of wakefulness test; microsleep episodes; sleepiness; wake-sleep transition zone