Adenoma detection rates augmented by AI in colonoscopy

Compared to conventional colonoscopy, the addition of artificial intelligence entailed significantly higher detection and lower rates of missed adenomas.

APC and ADR rates were superior with computer aided detection

An international, multicentre, randomised-tandem trial evaluated the potential enhancement of polyp detection by using AI in colonoscopy1. In 2 different groups, 916 patients with a mean age of about 60 years, underwent a colonoscopy with or without use of AI by the MAGENTIQ-COLO™ study device (

“A subset of patients was further randomised to undergo a tandem colonoscopy, so either a conventional colonoscopy followed by AI colonoscopy, or vice versa,” Dr Michiel Maas (Raboud UMC, the Netherlands) informed. The primary outcome measure was adenoma per colonoscopy (APC). Secondary objectives were adenoma detection rate (ADR) and adenoma miss rate. Reasons for the colonoscopy in the conventional group and computer aided detection group were surveillance (44.2% and 43.1%) and non-immunological faecal occult blood test (55.8% and 56.9%). 

Results showed a statistically significant APC rate for the computer aided detection compared with the conventional group (0.70 vs 0.51; P=0.014). The ADR was also superior with computer aided detection, not only in the entire cohort (37% vs 30%; P=0.014), but also for both indications: surveillance (P=0.001) and non-immunological faecal occult blood test (P=0.014).

Adenoma miss rate was almost halved with computer aid

Dr Maas conveyed that the adenoma miss rate was almost halved with computer aided detection (19%) in comparison with conventional colonoscopy (36%), while withdrawal times without intervention were equal in both groups.

In terms of adenoma characteristics that led to better detection with computer aided detection, size mattered. “We found an increased detection of diminutive adenomas sized 5 mm or less, but we also found an increase in small adenomas sized 6–9 mm in the computer aided detection group compared with the conventional group; there were no differences in the detection of advanced adenomas or sessile serrated lesions,” Dr Maas stated. Furthermore, computer aided detection found more adenomas in the proximal colon (P=0.006).

“This study further emphasises the beneficial role of AI or computer aided detection in improving our detection rates in a regular screening and surveillance population,” Dr Maas concluded.  

  1. Maas MHJ, et al. A novel computer-aided polyp detection system in daily clinical care: an international multicentre, randomised, tandem trial. LB06, UEG Week 2022, Vienna, Austria, 8–11 October.