Using simultaneous scanpath visualization to investigate the influence of visual behaviour on medical image interpretation

Abstract

In this paper, we explore how a number of novel methods for visualizing and analyzing differences in eye-tracking data, including scanpath length, Levenshtein distance, and visual transition frequency, can help to elucidate the methods clinicians use for interpreting 12-lead electrocardiograms (ECGs). Visualizing the differences between multiple participants’ scanpaths simultaneously allowed us to answer questions including: do clinicians fixate randomly on the ECG, or do they apply a systematic approach?; is there a relationship between interpretation accuracy and visual behavior? Results indicate that practitioners have very different visual search strategies. Clinicians who incorrectly interpret the image have greater scanpath variability than those who correctly interpret it, indicating that differences between practitioners in terms of accuracy are reflected in different eye-movement behaviors.  The variation across practitioners is likely to be the result of differential training, clinical role and expertise.