Abstract
Purpose :
Improving Optometric training requires a greater understanding of what drives students’ attention when examining and interpreting clinical images. For example, despite the wealth of data indicating that examining the optic cup provides only limited information about the presence or absence of glaucoma, students often describe the optic cup in detail both during routine examination and glaucoma assessment. Our aim is to understand the pattern of attention. Eye movements provide a reliable measure of visual attention and allow insight as to what students are attending when examining the ocular fundus for the presence of glaucoma.
Methods :
First year Optometry students (n=17) were recruited prior to receiving any anatomy or pathophysiology training. Participants were provided a written definition of glaucoma then shown 12 fundus images, 8 with glaucoma and 4 without and asked to identify which images depicted glaucoma. Eye movements were recorded using a Tobii 1750 (60Hz) tracker. Participants then received a short, narrated presentation on anatomy and pathophysiology of glaucoma before repeating the task. Eye movements pre and post were compared for each participant and novice student responses compared with areas identified via a machine derived visual saliency toolbox (Saliency Toolbox v2.3). Participants undertook a short quiz before and after the presentation to ensure naivety and benchmark learning.
Results :
Training significantly increased dwell time on drance haemorrhages (p=0.012) and the temporal retina (p=0.004) and produced a trend towards a reduction in dwell time to the optic cup (p=0.108). Despite this reduction, the cup remained the most examined AOI. This is consistent with visual saliency maps identifying the cup as the most salient part of the image. Individual fixation durations overall show a significant interaction between areas of interest pre and post (p=0.034) with a significant decrease to the cup and significant increase towards drance haemorrhages.
Conclusions :
Initial examination of optic nerve head images during a glaucoma detection task appears largely driven by saliency. Education regarding anatomy and pathophysiology provides a top-down cognitive overlay, which alters the pattern of attention without specifically indicating where to look in a “recipe” manner. A lot of time is still spent examining highly salient, but uninformative for the task, parts of the image indicating further opportunities for training exist.
This abstract was presented at the 2019 ARVO Annual Meeting, held in Vancouver, Canada, April 28 - May 2, 2019.