부산대학교 도서관

The processing of eye-tracking signals could be relevant in multiple fields. The report presents a description of pilot software for eye-tracking analysis based on python libraries. The outlines the various applications of the software, including its processing of input data of eye-tracking studies, variety of visualization tools, as well as features evaluation for assessment of eye-tracking signals. Overview of the basic pipeline was presented, which includes data calibration, demonstration of key visualizations graphs as well as features extraction in table format. Furthermore, the software was used to process data of pilot study involving healthy subjects and patients after a hemispheric stroke. Research has shown that when comparing target hit statistics, as well comparing as extracted data from static and dynamic tests for a group of healthy subjects and patients, the parameters are significantly different for the two groups, in particular amount of fixation events and misses of target is lower for healthy subjects, while successful hit of target is higher. Based on the extracted data, it can be noted that for the observed group of data, simple statistics of hitting the target, using the angular velocity algorithm to divide into fixations and saccades may be enough to divide the subjects into healthy and patients. Presented results revealed multiple informative features to discriminate subjects from different groups. Future development of the presented software includes increase of the extracted features to make it possible not only in binary classification but also for a multi-class tasks to distinguish patients with different pathologies.