Webcam-based eye-tracking for online assessment of smooth pursuit and saccade performance

Webcam-based eye-tracking is a scalable alternative to traditional eye-trackers, but its precision requires validation for clinical deployment. As methodological validation for ongoing studies involving Mild Cognitive Impairment (MCI) patients, we evaluated an online protocol to capture subtle gaze dynamics during static fixation, smooth pursuit, and saccadic paradigms online in a control sample . Participants were enrolled via LIAA’s DataPruebas (N=49). Excluding 8 subjects with poor calibration, the remaining sample (N=41, 34.8 ± 14.1 years) showed a mean Euclidean error of 0.20 ± 0.07 (normalized to screen height) during fixation. Smooth pursuit yielded comparable errors (euclidean: 0.22, horizontal: 0.12, vertical: 0.16). Despite higher error margins than state-of-the-art eye-trackers, this approach identified subtle variations in saccadic behavior. Prosaccade (PS) accuracy was significantly higher than that of antisaccade (AS) for strictly correct (69 vs. 60%) and correct or corrected trials (91 vs. 83%). Timing and kinematics presented significant differences: correct AS exhibited longer latencies than correct PS, whereas incorrect AS had shorter latencies than correct ones. The system also detected variations in amplitude and velocity, revealing corrected PS had higher amplitudes and faster gaze velocities than strictly correct trials. These findings support webcam-based eye-tracking as a sensitive tool for oculomotor assessment in scalable cognitive testing.