查看文章

In this prospective multicentre trial, a new computer-aided detection (CADe) software for colorectal polyp detection, incorporating a temporal algorithm, was validated and compared against a composite reference standard represented by both experienced endoscopists and CADe detection, by performing real-time unblinding of CADe findings by a second observer. We found an accuracy of 98.2% and 96.5% for detection by the endoscopist and by CADe, respectively. CADe was non-inferior to expert endoscopists (margin 5%; p< 0.001; 95% CI− 0.0378 to 0.0047), with a miss rate of 3.5% and 1.8%, respectively. This methodology allows to assess the additional value of CADe over human detection.