AI system enhances polyp detection in colonoscopy

The developed AI-assisted system significantly improves the detection and localization of colorectal polyps during colonoscopy. Out of five CNN models tested, EfficientNetV2 excelled, achieving an accuracy of 93.3% and a specificity of 94.6% on the test set. The system’s performance metrics, including a high AUC of 0.983, suggest its potential for clinical applications, as illustrated through representative case videos showing its feasibility in practice.