Abstract
Purpose :
Sustained oculomotor control and visual attention are critical to safe motor vehicle operation. Reliable measurement of visual performance during driving can give insight into human factors contributing to driver performance and provide accessible biomarkers of fatigue. This study assessed the reliability of visual performance metrics collected within a driving simulation paradigm.
Methods :
Fourteen participants with a full New Zealand drivers licence took part in one familiarisation session and two identical experimental sessions held at the same time of day, six weeks apart. Experimental sessions comprised operating a driving simulator (M300WS, STISIM) in two different scenarios, each completed twice. The scenarios were representative of roads typically encountered in urban and rural New Zealand. The urban scenario consisted of obstacles and intersections with audio navigation. The rural scenario consisted of winding roads, minimal obstacles and a visual attention task requiring an overt shift of attention towards visual targets 10°, 25° and 35° left or right from the centre of the windshield. The targets were left or right arrows and participants identified the arrow direction by pressing steering wheel buttons. Driving performance and visual attention metrics were recorded via the simulator and eye movement kinematics were measured using infrared oculography.
Results :
No systematic bias was detected in peak saccade velocity between and/or within sessions (main & interaction effects, p > 0.1) and measures of absolute agreement indicated a good reliability within sessions (ICC = 0.75 & 0.74, r = 0.78 & 0.81 session 1 & 2, respectively) and between sessions (ICC = 0.77 & 0.70, r = 0.76 & 0.74 session 1 & 2, respectively). Response time in the attention task also showed good reliability across all sessions. However, for task response accuracy, a learning effect was present between, but not within sessions. Thus, poor reliability between sessions was observed in attention task accuracy.
Conclusions :
Our measurements of reliability suggest that peak saccade velocity, used as an outcome variable in a driving simulator paradigm, is repeatable with a high degree of absolute agreement across multiple time points. However, visual attention task accuracy is only reliable within sessions. Baseline measures of visual attention can improve reliability when interpreting visual attention task performance.
This abstract was presented at the 2019 ARVO Annual Meeting, held in Vancouver, Canada, April 28 - May 2, 2019.