Thesis Defense: Tablet-Based Eye Tracking for Scalable Neurological Assessment

Doctoral Thesis Title: Tablet-Based Eye Tracking for Scalable Neurological Assessment
Presenter: Jamie Koerner
Presenter’s Affiliation: RLE
Thesis Supervisors: Prof. Thomas Heldt and Prof. Vivienne Sze
Date: Thursday, August 20, 2026
Time: 12–1 p.m. ET
Location: E25-117

Abstract:
Eye movements provide quantitative measures of neurological function, but the infrared eye trackers used to measure them accurately are largely confined to specialized laboratories. Tablets could support scalable neurological assessment, but their lower frame rates and nonoptimized cameras may reduce measurement accuracy. Using synchronized recordings from a tablet and a clinical-grade EyeLink eye tracker across five eye-movement tasks, this thesis determines which oculomotor metrics are preserved and which require correction. Saccade latency, directional-error rates, and target-acquisition measures were generally well preserved, whereas saccade amplitude, peak velocity, and smooth-pursuit measures were less accurate.
To improve these measurements, this thesis evaluates two complementary correction approaches. Metric-specific constants correct final metric values, while an EyeLink-supervised temporal convolutional network corrects the tablet gaze trace. Both approaches reduce errors in saccade and smooth-pursuit metrics for held-out participants. EyeLink is required only during training; the resulting corrections operate from tablet recordings alone. Using the corrected measurements, the thesis evaluates 37 metrics and identifies which support within-participant, between-participant, and group-level analyses. Applications in pediatric concussion assessment and repeated at-home testing further demonstrate the potential of tablet-based eye tracking for scalable neurological assessment.

Details

• Date: Thursday, August 20
• Time: 12:00 pm - 1:00 pm
• Category: Thesis Defense
• Location: E25-117