Subscribe to RSS
DOI: 10.1055/s-0045-1805783
Accuracy of real-time polyp characterization in artificial intelligence-assisted colonoscopy
Authors
Aims Artificial intelligence (AI) assistance during colonoscopy improves polyp detection rate (PDR) and adenoma detection rate (ADR) [1]. However, most colorectal polyps are diminutive and carry minimal cancer risk [2]. AI could enable diagnostic strategies such as leave-in-situ and resect-and-discard, provided it meets predefined quality benchmarks [3]. This study aimed to assess the accuracy of a computer-aided polyp characterization system (CADx) in predicting histology of diminutive rectosigmoid polyps during real-time colonoscopy in a Danish population.
Methods We conducted a prospective diagnostic accuracy study across four endoscopy centers using the GI Genius 3.0 system (Medtronic). Adults referred for colonoscopy due to a positive fecal immunochemical test (FIT) (> 100 μg/L), surveillance, or other diagnostic indications were included. Patients with inadequate bowel preparation were excluded. The procedures were carried out by a diverse group of endoscopists. Histopathology served as the reference standard, and all polyps were categorized on a binary scale as adenomas or non-adenomas. Sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and overall accuracy were calculated from a confusion matrix.
Results We included 278 patients for analysis. There was a total of 772 polypectomies. In the rectosigmoid colon, 294 polyps were removed with 221 being diminutive (1-5 mm). Out of those 221 polyps, 33 were classified as 'no prediction' by the system, and therefore excluded from the analysis. The system failed to detect one polyp, and three cases had missing data. For diminutive rectosigmoid polyps (n=184), the CADx-system achieved a sensitivity of 93% (95% CI, 87–97%) and a specificity of 33% (95% CI, 22–46%). The PPV and NPV were 70% (95% CI, 62–77%) and 74% (95% CI, 55–88%), respectively, with an overall accuracy of 71% (95% CI, 64–77%) and a diagnostic odds ratio (DOR) of 6.69. False positive rates across different polyp size categories were assessed for the entire colon. In diminutive polyps (1–5 mm), the false positive rate was 18.7%. For small polyps (6–9 mm), the rate was 16.3%, and for large polyps (≥ 10 mm), the false positive rate was 4.8%.
Conclusions The CADx-system demonstrated a high sensitivity, but a significantly lower specificity compared to prior studies [4] [5] [6] [7], driven by a high false positive rate. False positive rates were higher in smaller polyps and decreased as polyp size increased. A key limitation is the small sample size, which may partially account for the variability in our results. Nevertheless, our findings contribute to the growing body of knowledge on CADx. Inconsistent findings across studies highlight the challenges in standardizing AI-based characterization systems. Our results indicates that CADx is a promising future adjunct to colonoscopy, but its current performance is insufficient for reliable implementation in resect-and-discard or leave-in-situ strategies.
Publication History
Article published online:
27 March 2025
© 2025. European Society of Gastrointestinal Endoscopy. All rights reserved.
Georg Thieme Verlag KG
Oswald-Hesse-Straße 50, 70469 Stuttgart, Germany
-
References
- 1 Ishita B, Paulina W, Shin-ei K, Masashi M, Øyvind H, Shraddha G. et al. Real-Time Artificial Intelligence–Based Optical Diagnosis of Neoplastic Polyps during Colonoscopy NEJM Evid. 2022; 1 (6) EVIDoa2200003.
- 2 Ponugoti PL, Cummings OW, Rex DK.. Risk of cancer in small and diminutive colorectal polyps. Dig liver Dis Off J Ital Soc Gastroenterol Ital Assoc Study Liver 2017; 49 (1): 34-7
- 3 Ashat M, Klair JS, Singh D, Murali AR, Krishnamoorthi R.. Impact of real-time use of artificial intelligence in improving adenoma detection during colonoscopy: A systematic review and meta-analysis. Endosc Int open 2021; 9 (4): E513-21
- 4 Li JW, Wu CCH, Lee JWJ, Liang R, Soon GST, Wang LM. et al. Real-World Validation of a Computer-Aided Diagnosis System for Prediction of Polyp Histology in Colonoscopy: A Prospective Multicenter Study. Am J Gastroenterol 2023; 118 (8): 1353-64
- 5 Rondonotti E, Hassan C, Tamanini G, Antonelli G, Andrisani G, Leonetti G. et al. Artificial intelligence-assisted optical diagnosis for the resect-and-discard strategy in clinical practice: the Artificial intelligence BLI Characterization (ABC) study. Endoscopy. Germany 2023; 55 p 14-22
- 6 Hassan C, Balsamo G, Lorenzetti R, Zullo A, Antonelli G.. Artificial Intelligence Allows Leaving-In-Situ Colorectal Polyps. Clin Gastroenterol Hepatol Off Clin Pract J Am Gastroenterol Assoc 2022; 20 (11): 2505-2513.e4
- 7
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 (1): 88-99
Reference Ris Wihthout Link