start » program » presentations

Eye movement patterns as a reflection of expert behavior in viewing dynamic medical images

The present eye movement study investigated the role of professional experience in viewing dynamic scenes and the impact of presentation speed on diagnostic accuracy. The dynamic scenes were 9 videos of CT-scans from (the upper to the lower part of) the abdomen each consisting of approximately 700 multislice stack-view images. Three of the videos included enlarged lymph glands, three other pathologies, and three did not contain any abnormalities. An expert radiologist and a relative novice subject group of radiograph nurses were instructed to detect enlarged lymph glands in the CT-scans while their eye movements were registered. The radiologists were presumed to have better refined visual skills in detecting enlarged lymph glands due to their professional experience. The CT-scans were presented with a frame rate of 7, 14 or 28 frames per second. All participants indicated 7 frames per second to be too slow and detection success was also slightly worse in this than in other conditions. The results also showed that the radiologists more often detected the enlarged lymph glands than the radiographers did. The difference in success rate was associated with shorter average fixation durations for experts. This finding is in line with the theory of long-term working memory of Ericsson and Kintsch (1995), which proclaims that experts encode and retrieve information more rapidly than novices. The holistic model of image perception (Kundel et al., 2007) holds that experts have a more extensive perceptual span than novices allowing them to extract more information during a single fixation. This would predict them to make longer saccades than novices, but instead the average saccade length was slightly longer for novices than for experts. This suggests that radiologists perform the task actually more carefully than radiographers, a view supported by generally wider pupil dilation for the former group.