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

• Published in: bioRxiv
• Main Authors: Coe, Brian C., Huang, Jeff, Brien, Donald C., White, Brian J., Yep, Rachel, Munoz, Douglas P.
• Format: Paper
• Language: English
• Published: Cold Spring Harbor Laboratory 25.02.2022
• Edition: 1.1
• Subjects: Neuroscience; video-based eye tracking; blincades; automated saccade classification; post-saccadic slosh; automated blink classification; Saccade; pupil area; boomerang saccades; blink; automated trial marking; post saccadic oscillations; pupil
• ISSN: 2692-8205
• DOI: 10.1101/2022.02.22.481518

The tremendous increase of video-based eye tracking has made it possible to collect data from thousands of participants. Traditional manual detection and classification procedures for saccades and trial categorization is not viable for the large data sets being collected. Additionally, high-speed video-based eye trackers now allow for the novel analysis of pupil responses and blink behavior. Here we present a detailed description of our pipeline for collecting data, storing data, organizing participant codes, and cleaning data, which are fairly lab-specific but nonetheless important precursory steps in establishing standardized pipelines. More importantly, we also include descriptions of the automated detection and classification of saccades, blinks, ‘blincades’ (blinks occurring during saccades), and boomerang saccades (two saccades in opposite directions that occur almost simultaneously so that speed-based algorithms fail to split them), which are almost entirely task-agnostic and can be used on a wide variety of data. We additionally describe novel findings about post-saccadic oscillations, and provide a method to get more accurate estimates for end-points of saccades. Lastly, we describe the automated behavior classification for the Interleaved Pro- and Anti-Saccade Task (IPAST), a well-known task that probes voluntary and inhibitory control. This pipeline was evaluated using data collected from 592 human participants between 5 and 93 years of age, making it robust enough to handle large clinical patient data sets as well. In sum, this pipeline has been optimized to consistently handle large data sets obtained from diverse study cohorts (i.e., development, aging, clinical), and collected across multiple laboratory sites.