S327 Enhancing Polyp Detection Through the Use of Computer-Aided Detection: A Single Center Experience

Introduction: Over the past 2 decades, the incidence of colonic neoplasia has risen markedly, prompting efforts by leading medical societies to enhance colorectal cancer screening. Artificial Intelligence (AI)-based computer-aided detection (CADe) has emerged as a tool to improve adenoma detection rates. Our study evaluated combined adenoma and sessile serrated polyp detection, hereafter referred to as significant polyp detection rate (sPDR), before and after CADe implementation. Methods: Six volunteer endoscopists participated in the study. Retrospective data were collected from screening and non-screening colonoscopies performed between September 2023 and February 2024, followed by prospective data collection from May 2024 to February 2025 after CADe was introduced. Non-screening cases included those with a history of premalignant polyps, positive stool-based tests, or symptoms. Inclusion criteria were age >18, no history of colonic malignancy or genetic polyposis, adequate bowel prep (BBPS ≥6), cecal intubation, and ≥6-minute withdrawal time. Significant polyps included tubular or advanced adenomas, and hyperplastic or sessile-serrated lesions proximal to the splenic flexure. sPDR was calculated for screening vs non-screening exams and stratified by CADe use. Chi-squared testing was used with significance set at P < 0.05. Results: Of 1,825 total colonoscopies, 1,170 were without CADe and 655 with CADe. Among screening exams, 592 were performed without CADe and 371 with. Adenomas were detected in 586 cases without CADe (253 screening, 333 non-screening) and 381 with CADe (194 screening, 187 non-screening). The overall sPDR improved from 50.34% (no CADe) to 58.32% (with CADe) (P= 0.001). In screening exams, sPDR increased from 42.74% to 52.29% (P= 0.003), and in non-screening exams there was a trend towards improvement from 57.61% to 65.85% (P= 0.07). Conclusion: CADe use was associated with a significant increase in sPDR overall and in screening exams specifically. Implementation of CADe led to an ∼8% absolute sPDR increase overall, and ∼10% for screening procedures. These findings align with other U.S. and European studies. Limitations include sample size, and further research is needed to clarify AI’s role in enhancing endoscopic quality and outcomes.