Automated Analysis Pipeline for Extracting Saccade, Pupil, and Blink Parameters Using Video-Based Eye Tracking

Area of Science:

• Cognitive Neuroscience
• Ophthalmology
• Human-Computer Interaction

Background:

• Large-scale video-based eye-tracking generates massive datasets.
• Manual analysis of saccades and blinks is unfeasible for big data.
• Automated methods are crucial for efficient eye-tracking data processing.

Purpose of the Study:

• To present a standardized, automated pipeline for processing large-scale eye-tracking data.
• To enable accurate detection and classification of saccades, blinks, and complex eye movements.
• To facilitate robust data analysis across diverse participant groups and research sites.

Main Methods:

• Developed a comprehensive pipeline for data collection, storage, cleaning, and participant coding.
• Implemented automated algorithms for detecting saccades, blinks, blincades, and boomerang saccades.
• Included novel methods for analyzing post-saccadic oscillations and improving saccade endpoint estimation.
• Automated behavior classification for the interleaved pro/anti-saccade task (IPAST).

Main Results:

• The pipeline successfully processed data from 592 participants aged 5-93.
• Demonstrated robustness in handling diverse datasets, including developmental, aging, and clinical cohorts.
• Achieved accurate detection and classification of various eye movement types, including novel phenomena.

Conclusions:

• The automated pipeline is optimized for large, diverse eye-tracking datasets.
• This approach enhances the efficiency and standardization of eye-tracking data analysis in multi-site and clinical studies.
• The system supports a wide range of eye-tracking research applications.