SacLab: An eye-tracking analysis toolbox to increase usability of eye-tracking systems in clinical and research applications

A short period of eye-tracking can produce large amounts of data, making analysis challenging, particularly for clinical applications. Automated methods for extracting basic oculomotor features of saccades, fixation, or blinking are available. For more in-depth oculomotor analysis, manual cleaning and processing is often required. This is time-consuming and variable across individuals and over time, which limits scalability and clinical utility, even if identified oculomotor markers have high potential clinical significance. Automation provides an opportunity to reduce manual labour and improve replicability and precision through algorithmic approaches. Here we outline the design and validation of our SacLab Toolbox for processing data from well-established oculomotor paradigms commonly used in clinical research such as prosaccade, antisaccade, and smooth pursuit tasks. For prosaccade and antisaccade tasks, results showed good agreement between algorithmic and manual cleaning regarding which trials were to be included (86% for prosaccade and 91% for antisaccade with F1-scores of 0.89 and 0.79, respectively). The toolbox was also tested on data for both tasks from an independent laboratory with a different manual cleaning protocol, and the level of agreement remained high (87% for prosaccade and 93% for antisaccade with F1-scores of 0.93 and 0.94, respectively). The extent of concordance for the metrics extracted in both methods is discussed. The predominant reason for discrepancy between the manual process and our toolbox related to human error in the manual process. Overall, the results provide support for the benefits of this toolbox for these eye-tracking tasks in clinical and research applications.