Eye-Tracking Analytics

Mining where groups of people look: scanpath pattern mining, webcam eye tracking, and visual-salience studies with Geisinger’s Autism & Developmental Medicine Institute.

From 2015 to 2023 I worked with Dr. Vanessa Troiani (Geisinger Autism & Developmental Medicine Institute) and Dr. Evan Peck on new ways to analyze eye-tracking data, especially to understand visual attention in autism.

Group-level scanpath mining. Most eye-tracking analysis summarizes one viewer at a time. We developed methods that cluster fixations across many viewers and then mine the sequences of regions people look at, keeping only patterns that are statistically significant. That work was presented at the Vision Sciences Society 2019 meeting (Journal of Vision abstract). Later, Kartikeya Sharma ’22 extended it into a gaze-trajectory toolkit (Kalman Research Symposium, 2022).

Webcam eye tracking. Can an ordinary webcam stand in for a lab eye tracker? Students built and evaluated browser-based eye tracking, including gamified calibration, as part of a Bucknell–Geisinger Healthcare Research Initiative project. Demo site →

Visual salience and autism. Student projects used deep learning and image features to study which images draw high interest from children with autism. Another ported a vision-training game to VR to explore shifting visual attention bias.

Students: Tongyu Yang ’18, Morgan Eckenroth ’17, Khai Nguyen, Brooke Bullek ’18, Zilin Ma, Yili Wang ’21, Kartikeya Sharma ’22. See Research Students.