O49 Gaze tracking analysis identifies mechanism for improved adenoma detection with linked colour imaging during colonoscopy

Abstract

<h3>Introduction</h3> An improvement in adenoma detection rate (ADR) is linked to a reduced risk of interval colorectal cancer and mortality. Linked colour imaging (LCI, Fujifilm) enhances red-white contrast and improved ADR in clinical trials. Identifying the underlying mechanism will help develop future interventions to improve ADR. We used eye tracking technology to establish gaze patterns on white-light (WLE) and LCI and the impact on polyp detection stratified by endoscopist skillset. <h3>Methods</h3> This prospective study enrolled participants blinded to study hypotheses, stratified by polyp detection rate (PDR): PDR = 0 (non-endoscopist), PDR 1 – 30 and PDR &gt; 30. Participants watched randomised colonoscopy recordings obtained from 2 patients of withdrawal from the caecum to hepatic flexure with WLE, LCI and dye chromoendoscopy whilst eye tracking was performed. We analysed the location and duration of gaze fixations (milliseconds) and time to first fixation for a single polyp present on all image modes (interval when polyp is first visible on a frame to first fixation). We compared gaze metrics between imaging modes stratified by PDR groupings. <h3>Results</h3> We recruited 33 participants (PRD 0 = 14, PDR 1–30 = 13 and PDR &gt; 30 = 6). There was no significant difference in lifetime procedure count (778.2±720.0 vs 1841.7±1450.0, p = 0.10) and years of practice (7.2±4.7 vs 10.1±6.4, p = 0.35) between PDR 1–30 and PDR&gt;30 groups. All participants had a bias to focus on the vertical midline of the monitor on WLE than the right and left columns based on total duration of fixations (7.22±1.75ms vs 3.47±0.79ms, p&lt;0.001).There was a significant difference between PDR groups for the total duration of fixations on the centre of the lumen (Kruskal-Wallis, p&lt;0.01), with a trend for lower duration of fixations associated with higher PDR; on post-hoc analysis this was only significant for PDR 0 vs PDR 1–30 (p&lt;0.01) and PDR 0 vs PDR &gt; 30 (p&lt;0.001). For all participants polyp detection was faster for LCI (0.53ms vs 0.99ms, p&lt;0.01) and dye chromoendoscopy (0.58ms vs 0.99ms p&lt;0.01) than WLE, with no significant difference between LCI and dye chromoendoscopy. Sub-group analysis by PDR revealed this was only significant in the PDR 1–30 group for LCI vs WLE (0.48ms vs 0.91ms, p&lt;0.01). <h3>Conclusions</h3> LCI is an effective tool to improve ADR and is associated with faster polyp detection. There may be a learning curve associated with this technology with the greatest benefit for low to average polyp detectors. Eye tracking technology is a powerful tool to evaluate novel image enhancement technologies.