Effect of a novel spatial-temporal computer-aided detection system on adenoma detection during colonoscopy: A multicenter, randomized controlled trial

Abstract

Background Most computer-aided detection (CADe) systems depend on the spatial information from static images rendering them unreliable for real-time diagnosis. This research aims to assess the performance of a novel spatial-temporal CADe system. Methods This randomized study recruited patients 18 years or older scheduled for colonoscopy at four endoscopy centers from June 2023 to September 2023. Participants were randomly assigned to receive CADe colonoscopy or conventional colonoscopy. The primary outcome was ADR. Furthermore, the correlation between endoscopists’ acceptance rate of the CADe system and ADR was observed. Results Among 3317 patients, the ADR was 32.1% in the CADe group and 24.7% in the control group (P < 0.001). The CADe group detected significantly more adenomas < 10 mm and flat-type adenomas [(23.3% vs. 18.4%, P = 0.001) and (15.5% vs. 9.9%, P = 0.001), respectively]. In addition, a significant correlation (r = 0.916, P < 0.001) was observed between the grit score and ADR with the CADe system. Conclusion The spatial-temporal CADe system significantly improved overall polyp and adenoma detection, especially for diminutive and flat-type lesions. Moreover, endoscopists with a greater propensity to embrace the CADe system tend to detect a higher proportion of adenomas during colonoscopy.