Incremental Yield of Artificial Intelligence in Follow-Up Screening Colonoscopies – an Interim Analysis

Aims Although colonoscopy is the gold standard for detection of precancerous colonics lesions, they are still missed and this is directly related with an increased risk of interval colorectal cancer. To improve the diagnostic performances of colonoscopy, novel technologies have been recently developed, in particular Artificial Intelligence (AI). The objective of this study was to compare the diagnostic yield obtained by using the GI Genius (Medtronic, Minneapolis, USA) AI software during colonoscopy to the yield obtained by the standard colonoscopy (SC).

Methods This is a single-center RCT evaluating consecutive patients undergoing follow-up screening colonoscopy at Fondazione Poliambulanza Istituto Ospedaliero, Brescia, Italy. Patients were randomly assigned to SC or AI. Subjects with ≤ 1 in any segment according to Boston Bowel Preparation Scale were excluded from analysis. Polyp Detection Rate (PDR), Adenoma Detection Rate (ADR), Serrated Detection Rate (SDR), patients with advanced adenomas (i.e. villous histology, high-grade dysplasia or low-grade dysplasia > 1cm) and patients with ≥ 3 adenomas were compared between the group using χ2-test. P < 0.05 were considered statistically significant.

Results Data about 275 patients (M:F = 157:118) were collected. 126 patients were assigned to AI, while 149 to SC arm. 15 patients (6 AI and 9 SC group) were excluded due to inadequate cleansing. Statistically significant improvements were shown in terms of PDR (83.3 % [100/120] vs 73.5 % [103/140]; p = 0.029), ADR (71.6 % [86/120] vs 58.5 % [82/140]; p = 0.014) and patients with ≥ 3 adenomas (28.3 % [34/120] vs 17.1 % [24/140]; p = 0.016). Although an increase of the other outcomes, no statistically significance was reached for SDR (15 % [18/120] vs 12.1 % [13/140]; p = 0.2) and patients with advanced adenomas (12.5 % [15/120] vs 9.2 % [13/140]; p = 0.2).

Conclusions This preliminary results suggest that AI can be a useful tool during screening colonoscopy, since it can improve the neoplasia yield.

Citation: Milluzzo SM, Cesaro P, Hassan C et al. OP122 INCREMENTAL YIELD OF ARTIFICIAL INTELLIGENCE IN FOLLOW-UP SCREENING COLONOSCOPIES – AN INTERIM ANALYSIS. Endoscopy 2021; 53: S51.

Publikationsverlauf

Artikel online veröffentlicht:
19. März 2021