B1692
Title: Video segmentation and functional data pipeline for assessing pupil changes due to cannabis consumption
Authors: Julia Wrobel - Colorado School of Public Health (United States) [presenting]
Abstract: Research findings on the impact of acute cannabis use on pupillary size have been inconsistent. We developed a video processing and analysis pipeline that extracts pupil sizes from videos obtained during a light stimulus test administered with goggles utilizing infrared videography. The light stimulus test was administered for those with occasional ($N=36$), daily ($N=33$), and no cannabis use ($N=32$), and before and after acute cannabis smoking or a waiting period for the no-use group. Pupils were segmented using a combination of image pre-processing techniques and segmentation algorithms and linear least-squares ellipse fitting estimated pupil size. The segmentation pipeline achieved 99 percent precision on a validating set of images ($N=517$). Pupil size trajectories in response to the light stimulus were then analyzed across marijuana use groups using techniques from functional data analysis (FDA). Our results show that FDA-based regression models are more sensitive to differences across marijuana use groups than using scalar features extracted from the pupil trajectories. In addition, we find that acute cannabis use from both occasional and daily use groups results in less pupil constriction and slower rebound dilation in the light stimulus test.
