(1381) Comparison Between Expert and Novice Visual Scanpaths in Lung Ultrasound Interpretation

Introduction: The cognitive process of medical image interpretation is incompletely understood. Previous studies have shown differences between experts and novices in the speed of finding pathology, but this information does not help to inform how best to teach interpretation. A greater understanding of how experts work through images would likely help learners develop heuristics and accelerate learning. We sought to determine whether eye tracking, which uses a device to track a participant's gaze location, would show differences in expert and novice search behavior when interpreting lung ultrasound images.

Methods: We conducted a prospective, observational, multi-institutional pilot study in which we compared expert pediatric critical care and emergency medicine ultrasound faculty with advanced training in ultrasound against novice resident physicians with no advanced training. Participants were asked to interpret a set of 10 lung ultrasound clips while eye tracking data was collected. From this eye tracking data, we used Multimatch, a previously validated algorithm, to compare the sequence and position of eye gaze fixations and saccades between each pair of participants. Following this, we ran a nonparametric permutation test to assess the difference between within-group (experts compared with experts and novices compared with novices) and between-group MultiMatch scores.

Results: 29 participants completed the study, including 14 experts and 15 novices. Experts had a significantly higher rate of correct interpretation when compared to novices (p < 0.001). Expert scanpaths were also significantly more similar to participants of their own expertise level than those of different expertise levels (mean Multimatch position score, within-group 0.9311, between group 0.9282, p=0.004).

Conclusions: This study shows that experts have a high similarity in the ordinal sequence of fixation points they use to look through lung ultrasound images. This suggests that despite not being taught a way to search through images, there is significant overlap in the way that experts do so. Future studies will need to assess whether this is true of different imaging modalities as well as to determine whether scan paths could be taught to learners to improve teaching of ultrasound interpretation.