Artificial Intelligence Significantly Improves Adenoma Detection Rate but Does Not Affect Polyp Detection Rate in Colonoscopy: A Propensity Score Matching Study

BackgroundColorectal cancer (CRC) remains a major cause of cancer-related morbidity and mortality worldwide. Endoscopy and adenoma removal are effective in reducing the incidence of CRC. Recent advances in artificial intelligence (AI)-assisted endoscopy have demonstrated the potential to improve detection outcomes. This study aimed to evaluate the effectiveness of AI-assisted endoscopy in improving adenoma detection rate (ADR) and polyp detection rate (PDR) during colonoscopy in a real-world clinical setting. MethodsA total of 824 colonoscopies between August 2022 and February 2024 at Inje University Busan Paik Hospital were included in the study. Patients were divided into two groups: AI CAD-assisted colonoscopy (N = 393) and conventional colonoscopy (N = 431). Propensity score matching was then performed using a 1:1 nearest-neighbor algorithm, balancing key covariates, including age, sex, BMI, ASA score, bowel preparation quality, and the ratio of expert endoscopists. Ultimately, 786 patients (393 per group) were included in the final comparative analysis. Logistic regression analyses were used to evaluate the association between AI CAD-assisted colonoscopy and ADR and PDR, adjusting for potential confounders. ResultsADR was significantly higher in the AI CAD-assisted group (41.5%) compared to the No-AI CAD group (34.4%) (adjusted OR = 1.380; 95% CI: 1.012-1.885; P = 0.042). PDR was higher in the AI CAD-assisted group (53.2% vs. 46.1%), but the difference was not statistically significant (OR = 1.312; 95% CI: 0.971-1.774; P = 0.077). Older age and higher BMI were positively associated with ADR, while male sex and higher ASA scores were negatively associated. ConclusionsAI-assisted CAD colonoscopy was independently associated with improved ADR after adjustment for potential confounders. While the increase in PDR was not statistically significant, the findings support the clinical utility of AI CAD. Larger multicenter prospective studies are warranted to validate these findings and guide the integration of AI tools into routine endoscopic practice.