Real-time use of a computer-aided system for polyp detection during colonoscopy, an ambispective study.

Abstract

This study aimed to evaluate ambispectively the effectiveness of a real-time computer-aided detection (CADe) system on the number of polyp (PPC) or adenoma per colonoscopy (APC), and polyp (PDR) or adenoma detection rate (ADR). Eight-five videos marked using the CADe system, together with the unmarked videos, were reviewed by two senior endoscopists. Polyps detected in the marked and unmarked videos were recounted in parallel. Additionally, 128 consecutive patients were enrolled for a prospective evaluation using a standard colonoscopy or the CADe monitor alternately every 2 weeks. The PC, APC, PDR and ADR were compared between the two groups. The total number of polyps reported in the unmarked and marked videos were 73 and 88, respectively (mean PPC 0.86 vs 1.04, P=0.001). The proportion of polyps detected per colonoscopy increased by 20.5%. Of the 128 prospectively enrolled patients, 186 polyps were detected. The mean PPC was higher in the CADe colonoscopy than in the standard colonoscopy (1.66 vs 1.13, P=0.039). The PDR using the CADe colonoscopy was significantly higher than that of the standard colonoscopy (78.1% vs 56.3%, P=0.008). Real-time CADe system significantly increases the PDR and PPC under the situation of a high rate of polyp detection.