Can AI Predict the Histology of Colonic Polyps? IA-COLO Study: A Prospective Multicenter Study Evaluating the Histological Prediction of the CAD-EYE System in Screening Colonoscopy

Aims The “diagnose-and-leave” and “resect-and-discard” strategies would reduce the environmental impact and the costs associated with screening colonoscopies, but requires a reliable optical diagnosis capability. Currently, the performance thresholds set by the American Society of Gastrointestinal Endoscopy (ASGE) and the European Society of Gastrointestinal Endoscopy (ESGE) are not being met due to variability in endoscopists’ optical diagnosis performance. Artificial Intelligence (AI) systems that predict the histology of colorectal polyps in real-time could provide a promising solution to reduce this variability. This study aimed to assess the diagnosis performance of the CAD-EYE system for predicting the neoplastic histology of colonic polyps [1] [2] [3] [4] [5] [6] [7].

Methods We conducted a cross-sectional, multi-center study to evaluate the CAD-EYE system in a real-life setting, focusing on patients undergoing screening colonoscopies at five centers in France. The AI-generated predictions (hyperplastic vs. neoplastic) were compared to histopathology results (gold standard). We classified low grade and high-grade dysplasia adenoma, sessile serrated lesions with dysplasia and cancers as ‘neoplastic’, and hyperplastic polyps and sessile serrated lesions without dysplasia as ‘non-neoplastic’. We performed a sensitivity analysis by classifying sessile serrated adenomas without dysplasia as ‘neoplastic’. The primary outcome was the sensitivity of the CAD-EYE system in predicting the neoplastic nature of colorectal polyps. The secondary outcomes were the specificity, PPV and NPV of the CAD-EYE system, the diagnosis performance of physicians, and the polyp detection performance of the CAD-EYE system.

Results In total, 398 colorectal polyps from 139 patients were sent for histological analysis, with 343 polyps included in the primary outcome analysis. Characterization using the CAD-EYE system was feasible in 96% of cases. The sensitivity of the CAD-EYE system for predicting the neoplastic nature of polyps was 80% (95% CI, 74%-85%) not different from 85% (p=0.0642). The specificity, NPV, and PPV were 79%, 64%, and 90%, respectively. The diagnostic performance of the CAD-EYE system tended to be higher for diminutive rectosigmoid polyps (sensitivity 80%, specificity 97%, NPV 86%). The sensitivity of the endoscopists was 90% (95% CI, 86%-94%), significantly higher from the sensitivity of the CAD-EYE system (p=0.014). The specificity and NPV of the endoscopists were 71% and 77%, respectively. Mean number of polyps detected per colonoscopy was higher if colonoscopies were assisted by CAD-EYE (3.3 versus 2.3, p<0.01).

Conclusions In this study, the sensitivity of endoscopists in predicting the neoplastic nature of colonic polyps was higher than sensitivity of the CADx. Diagnosis performance of the CADx were better for diminutive rectosigmoid polyps. The quality thresholds required by learned societies are still not met. Artificial intelligence appears to be beneficial in polyps’ detection.

Publication History

Article published online:
27 March 2025

• References

• 1 Vu HT, Sayuk GS, Hollander TG, Clebanoff J, Edmundowicz SA, Gyawali CP. et al. Resect and Discard Approach to Colon Polyps: Real-World Applicability Among Academic and Community Gastroenterologists. Dig Dis Sci 2015; 60 (02): 502-8
• 2 Mori Y, Kudo SE, East JE, Rastogi A, Bretthauer M, Misawa M. et al. Cost savings in colonoscopy with artificial intelligence-aided polyp diagnosis: an add-on analysis of a clinical trial (with video). Gastrointest Endosc 2020; 92 (04): 905-911.e1
• 3 Hassan C, Pickhardt PJ, Rex DK.. A Resect and Discard Strategy Would Improve Cost-Effectiveness of Colorectal Cancer Screening. Clin Gastroenterol Hepatol 2010; 8 (10): 865-869.e3
• 4 Houwen BBSL, Hassan C, Coupé VMH, Greuter MJE, Hazewinkel Y, Vleugels JLA. et al. Definition of competence standards for optical diagnosis of diminutive colorectal polyps: European Society of Gastrointestinal Endoscopy (ESGE) Position Statement. Endoscopy 2022; 54 (01): 88-99
• 5 Rex DK, Kahi C, O’Brien M, Levin TR, Pohl H, Rastogi A. et al. The American Society for Gastrointestinal Endoscopy PIVI (Preservation and Incorporation of Valuable Endoscopic Innovations) on real-time endoscopic assessment of the histology of diminutive colorectal polyps. Gastrointestinal Endoscopy 2011; 73 (03): 419-22
• 6 Hassan C, Sharma P, Mori Y, Bretthauer M, Rex DK, Repici A. et al. Comparative Performance of Artificial Intelligence Optical Diagnosis Systems for Leaving in Situ Colorectal Polyps. Gastroenterology 2023; 164 (03): 467-469.e4
• 7 Byrne MF, Chapados N, Soudan F, Oertel C, Linares Pérez M, Kelly R. et al. Real-time differentiation of adenomatous and hyperplastic diminutive colorectal polyps during analysis of unaltered videos of standard colonoscopy using a deep learning model. Gut 2019; 68 (01): 94-100