Abstract
Purpose:
Driving presents safety relevant stimuli at all eccentricities and contrasts. Divided attention, contrast sensitivity and visual processing speed are faculties negatively impacted by glaucoma. Functional degradation assessment is plagued by difficulties to separate detection of low contrast stimuli (vision based) from reaction to press a button (neuromuscular). We present a Monte Carlo approach to estimate distributions of key response-model coefficients per group/condition. The adopted response model is an evidence accumulation model for detection of low contrast peripheral stimuli followed by a lognormal reaction time model.
Methods:
109 patients with glaucoma and 69 age-matched controls performed a 3min car-following task in a driving simulator pressing a steering wheel button each time a peripheral stimulus at different Weber contrast levels (0.1,0.4,0.9) was detected. Monte Carlo simulations of responses (per group/contrast) where embedded in a search algorithm to find the parameters of the Gaussian population distribution of the response model’s coefficients: evidence-rate (rate at which evidence accumulates while stimulus is present) and mean reaction time (time to press button after detection). Noise in the evidence accumulation process and the detection-threshold of the detection model were fixed as was the variance in the reaction time model.
Results:
Focus was directed to the lowest contrast condition where modeling detection time separately was most meaningful. Glaucoma patients exhibited a significantly slower evidence-rate (P<0.001) as well as a longer reaction time (P<0.001) compared to age-matched controls. The population distributions clearly showed that the highest performing glaucoma patients performed as well as highest performing normal subjects (same high evidence rate and same low mean reaction time). The reaction time effect was only about 50-100ms for the worst performing glaucoma patients, while the effect of their low evidence-rate resulted in delays on the order of seconds and complete misses.
Conclusions:
The Monte Carlo simulation methodology to estimate population distributions of response-generation model-coefficients offers meaningful insights into the effect magnitudes that glaucoma has on the different processes involved in generating a response to low contrast peripheral stimuli during car following.
Keywords: 629 optic nerve •
496 detection •
758 visual fields