How to measure brain activity in drivers with sleep disorders: A methodological presentation

Author(s): Rizzo, Grad, Baltzan, Libman

Poster Presentation:

Poster link

Abstract:

Background:

Drivers with obstructive sleep apnea (OSA) - estimated to affect 5.4 million Canadian adults - are traditionally believed to be at higher risk for road accidents. To date, we have no accurate metric for establishing level of driver risk. Policies regarding driving with OSA are currently being decided all around the world and clear guidelines are limited.

Aims:

We plan to measure a critical cognitive aspect of driving performance: attention, simultaneously with driving performance, by tracking brain activity (microsleep events) directly through electroencephalography (EEG).

Methods:

This paper describes the 1) CARLA driving simulation program is an open-source urban driving simulator built under the Unreal Engine that provides a sandbox for autonomous driving algorithms. The simulation platform supports flexible specification of sensor suites, environmental conditions, full control of all static and dynamic actors, map generation, etc.; and 2) the ABM B-Alert X24 Electroencephalography (EEG) system (Advanced Brain Monitoring, Inc, Carlsbad, CA, USA) sampled at 256 Hz. The flexible electrode strip consists of a set of flat electrodes in standard 20 scalp locations and is affixed to an adjustable headband. A wireless transmitter attached to the headband sends EEG signals to a separate device for data capture. Datasets for CARLA and EEG will be synchronized to the 1-epoch second and exported into an SPSS file, ready for analyses and interpretation.

Results:

EEG data will be matched with the driving performance data (matched epochs), which will also have time-stamped events (e.g. deviation from lateral position, speed increase, sudden breaking).
Data will be analyzed using a Superposed epoch analysis: When comparing two time series, the occurrences of the microsleep events will be identified as key times; we will then extract subsets of data from the driving simulation dataset within a fixed range near each key time and superpose all extracted subsets by compositing them and comparing the means of those categories using mixed-design ANOVAs.

Discussion:

This methodological presentation describes how we plan to analyze and interpret the link between microsleeps and driving performance. In this methodology, driver cognitive activity and driving performance, in individuals with and without OSA, are evaluated in a single experimental design.

Conclusions:

Developing an accurate metric for establishing level of driver risk is an important issue in drivers diagnosed with a sleep disorder: such measurable events underlie the subjective experience of sleepiness and/or fatigue and are the precursors of performance deficits and compromised driving skill.