Implementing discard strategies for diminutive polyps using autonomous CADx in clinical practice

Abstract

The strategy of discarding diminutive colonic polyps (≤5 mm) has been discussed for many years. However, despite their exceedingly low risk of malignancy, implementation has remained limited, with minimal impact on clinical practice. More recently, computer-assisted optical diagnosis (CADx) has been introduced to support and potentially enable broader adoption of this approach. We prospectively assessed patient acceptance and examiner-based quality control in 95 out of 102 patients (93.1%) who consented to this approach, involving a total of 149 polyps. Of these, 143 polyps were reviewed by three examiners blinded to CADx diagnoses and the results were compared with CADx diagnoses. Resect-and-discard accuracy was 83.5% excluding sessile serrated lesions (SSLs), while treating SSLs as neoplastic or hyperplastic yielded 75.9% and 84.3%, respectively. Excluding SSLs, the diagnose-and-leave negative predictive value was 92.3%. Surveillance interval recommendations based on CADx and expert review, aligned with current guidelines, achieved 100% concordance, ensuring appropriate follow-up. The rationale behind replacing histopathology for diminutive polyps using optical polyp diagnosis is rooted in the extremely low risk of advanced pathology in diminutive polyps, supporting the safe practice of discarding or leaving polyps in situ when accurately optically diagnosed. Adopting this approach would significantly reduce pathology-associated costs and offer immediate patient management decisions. This prospective clinical pilot study included patients aged 45–80 years undergoing elective colonoscopy at the Montreal University Hospital Center from August 2025 to November 2024. Patients were invited to a study replacing histopathological evaluations of diminutive polyps with autonomous computer-assisted optical (CADx) diagnosis. Exclusion criteria were inflammatory bowel disease, active colitis, familial polyposis syndromes, coagulopathy or severe medical comorbidities. Patients received detailed written and verbal explanations regarding autonomous use of CADx technology, and optical diagnosis polyp management strategies ( resect-and-discard and diagnose-and-leave ). All participants gave written informed consent. Patients declining participation completed a voluntary survey regarding their reasons. Of …