Local Sequence Alignment for Scan Path Similarity Assessment

Abstract

Abstract-It has long been shown that there is a close relationship between eye movement, human cognition and brain activity. The present work seeks to explore this relationship by investigating the students' saccadic eye movement sequences in a problem solving task. We aim to assess students' reasoning process in a clinical problem solving task using students' visual trajectories. We use students' scan path, followed while resolving medical cases, and a local sequence alignment algorithm, to evaluate their analytical reasoning during medical case resolution. An experimental protocol was conducted with 15 participants. Eye movements were recorded while they were interacting with our learning environment. The proposed approach, based on gaze data, can be reliably applied to eye movement sequence comparison. Our findings have implications for improving novice clinicians' reasoning abilities in particular and ultimately enhancing learning outcomes.Index Terms-Cognitive tasks, eye movements, local sequence alignment, medical reasoning, scan path similarity.

Deep semantic gaze embedding and scanpath comparison for expertise classification during OPT viewing

“…Rather than deal with the entirety of two sequences, local alignment determines the most optimal aligned subsequence between the two. Local alignment compensates to a greater degree for sequences of differing lengths and is not as strongly influenced by differences in the beginning or end of the sequences [Khedher et al 2018]. For example, [Khedher et al 2018] used the Smith-Waterman algorithm [Smith et al 1981] for local alignment of medical undergrads' scanpaths during a clinical reasoning task.…”

“…Local alignment compensates to a greater degree for sequences of differing lengths and is not as strongly influenced by differences in the beginning or end of the sequences [Khedher et al 2018]. For example, [Khedher et al 2018] used the Smith-Waterman algorithm [Smith et al 1981] for local alignment of medical undergrads' scanpaths during a clinical reasoning task. They found similarly well performing students had highly correlative scores.…”

“…For scanpath comparison, we perform local alignment using a variant of the Smith-Waterman Algorithm. We preferred local alignment scoring over global alignment due to its ability to find similar subsequences, even if the scanpaths may otherwise be highly varying [Khedher et al 2018]. Moreover, we did not want to enforce strict global alignment due to different viewing times required by students and experts.…”

Understanding Clinical Reasoning through Visual Scanpath and Brain Activity Analysis

“…The higher the aligned similarity score, the more the two sequences are similar, which means the closer the learner's visual sequence is to the optimal reference sequence. We refer to [83] for more details about this procedure.…”