Supplemental Material for: Research article The diagnostic performance of a novel computer-aided characterization system capable of differentiating sessile serrated lesions as neoplasms in a clinical practice setting

Le résumé fourni par la source

Introduction A new narrow-band imaging computer-aided diagnosis (NBI-CAD) has been developed that can differentiate sessile serrated lesions (SSLs) as a neoplastic lesion from hyperplastic polyps. However, its diagnostic performance remains unclear. We aimed to investigate the diagnostic performance of the NBI-CAD system in routine colonoscopy practice. Methods This single-center retrospective study included patients who underwent colonoscopy with NBI-CAD between March and May 2024. Exclusions were cases with no lesions, inadequate bowel preparation, lesions indicated for surgery, no NBI-CAD diagnosis, left in situ for endoscopic hyperplastic polyp, or pathological normal tissue. The diagnostic performance of NBI-CAD alone and endoscopists with NBI-CAD was compared. The diagnosis of NBI-CAD alone was exploratorily defined as the more frequent of “neoplastic” or “non-neoplastic.” Endoscopists diagnosed lesions in real time with reference to the NBI-CAD result. For pathological diagnosis, adenomatous lesions or SSLs were defined as neoplastic, and others as non-neoplastic. We analyzed factors associated with misdiagnosis by NBI-CAD alone. Results A total of 169 patients with 598 lesions were analyzed. Sensitivity did not significantly differ between NBI-CAD alone and endoscopists with NBI-CAD (0.980 vs. 0.987, P = 0.343). Specificity and accuracy were significantly improved by endoscopists with NBI-CAD compared to NBI-CAD alone (0.467 vs. 0.317, P = 0.0044 and 0.935 vs. 0.913, P = 0.018). In multivariate analysis, rectosigmoid location [Odds ratio (OR) 2.8, P = 0.01] and JNET type 1 (OR 17.3, P < 0.001) were identified as independent risk factors for misdiagnosis. Conclusion In our exploratory study, NBI-CAD showed favorable sensitivity for colorectal neoplasms, and endoscopists improved specificity and accuracy without reducing sensitivity.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.