Cannabis Impairment Monitoring Using Objective Eye Tracking Analytics

TL;DR

This study develops an eye-tracking based method to objectively and rapidly assess cannabis impairment, demonstrating significant changes in eye movement metrics and achieving high classification accuracy between sober and impaired states.

Contribution

Introduces a novel eye-tracking analytics approach combined with machine learning to objectively detect cannabis impairment, improving speed and accuracy over traditional methods.

Findings

Significant post-cannabis changes in eye movement metrics

Supervised models achieved 89% accuracy in impairment classification

Eye-tracking proved effective for rapid impairment assessment

Abstract

The continuing growth in cannabis legalization necessitates the development of rapid, objective methods for assessing impairment to ensure public and occupational safety. Traditional measurement techniques are subjective, time-consuming, and do not directly measure physical impairment. This study introduces objective metrics derived from eye-tracking analytics to address these limitations. We employed a head-mounted display to present 20 subjects with smooth pursuit performance, horizontal saccade, and simple reaction time tasks. Individual and group performance was compared before and after cannabis use. Results demonstrated significant changes in oculomotor control post-cannabis consumption, with smooth pursuit performance showing the most substantial signal. The objective eye-tracking data was used to develop supervised learning models, achieving a classification accuracy of 89% for distinguishing between sober and impaired states when normalized against baseline measures. Eye-tracking is the optimal candidate for a portable, rapid, and objective tool for assessing cannabis impairment, offering significant improvements over current subjective and indirect methods.