Impact of Computer-Aided Detection on Endoscopist's Gaze-Shift Distance During Colonoscopy: a Randomized Controlled...

Computer-aided detection (CADe) facilitates colorectal lesion detection, but it remains unclear whether CADe affects endoscopist's eye strain by altering gaze patterns. This study aimed to determine whether CADe-assisted colonoscopy is superior to conventional colonoscopy in reducing endoscopists' gaze-shift distance.

This single-center randomized controlled trial enrolled participants and randomly assigned them to two groups: first observation with CADe, then second observation without CADe (CADe-first group) in the ascending colon, or observations in the reverse order (conventional-first group). The endoscopist's gaze positions were recorded via a dedicated eye-tracking system. The primary endpoint was the endoscopist's gaze-shift distance within the first allocated group. The secondary endpoints were the gaze-shift distance based on CADe experience and eye strain, ascertained using a visual analog scale (VAS) score.

Among 59 patients that were assigned to two groups (CADe-first, n = 30; conventional-first, n = 29), there was no intergroup difference in the gaze-shift distance (30.7 ± 10.7 pixels/30 ms vs. 32.7 ± 8.7 pixels/30 ms, p = 0.450). Among CADe-experienced endoscopists, the distance was significantly shorter in the CADe-first than in the conventional-first group (25.9 ± 3.4 pixels/30 ms vs. 29.4 ± 3.6 pixels/30 ms, p < 0.01), whereas no difference was observed among CADe-inexperienced endoscopists (40.3 ± 13.8 pixels/30 ms vs. 45.4 ± 11.1 pixels/30 ms, p = 0.463). The VAS score was significantly lower in the CADe-first than in the conventional-first group (3.2 ± 1.2 vs. 4.6 ± 1.5, p < 0.05).

CADe assistance did not significantly change the gaze-shift distance overall, but it reduced gaze-shift distance among CADe-experienced endoscopists. Thus, CADe may be associated with reduced eye strain, depending on the endoscopist's CADe experience.