TRUE-Colon: Exposing a Consistent Transfer Asymmetry in Real-Time Polyp Detection

2026-08-13

AI Summary

This study addresses clinical performance degradation in colonoscopy CADe systems caused by data bias by constructing the True-Colon benchmark and proposing a standardized full-procedure evaluation protocol. By comparing mainstream detectors on curated datasets versus 60 unedited full-length videos, this work quantitatively reveals transfer asymmetry in real-time polyp detection for the first time. Results demonstrate that full-procedure training significantly reduces false positives while maintaining precision, with Transformer architectures achieving optimal sensitivity and CNNs offering superior throughput. Ultimately, this research provides a critical benchmark and methodological foundation for bridging the gap between algorithmic development and real-world clinical deployment in automated colonoscopy.

Abstract

Computer-aided detection (CADe) systems for colonoscopy promise to reduce clinical miss rates, yet reliable real-world deployment remains elusive. This translational gap stems in part from a structural flaw in model development: the reliance on curated datasets that under-represent the long negative stretches and procedure-related artifacts characteristic of routine examinations. Training and evaluating architectures strictly on these lesion-centric benchmarks creates an illusion of success, since such benchmarks cannot capture clinically crucial metrics. To expose this gap, we establish TRUE-Colon, a standardized benchmarking protocol that measures key deployment characteristics alongside localization accuracy, and evaluate four real-time architectures (Faster R-CNN, YOLOv8, YOLOv11, RT-DETR) across curated benchmarks (SUN, PICCOLO) and 60 unedited, full-length procedures (REAL-Colon). We observe a consistent transfer asymmetry: models trained strictly on curated clips suffer a severe performance collapse when evaluated on full procedures, whereas procedure-trained models substantially improve rejection of non-polyp content on REAL-Colon, and largely retain their accuracy on curated benchmarks. Beyond transferability, we find that the Transformer detector attains the strongest sensitivity and the earliest, most persistent detections, while the convolutional detectors stay competitive at a higher throughput. Together, these results indicate that both training and benchmarking for deployable CADe should shift from curated, lesion-centric clips toward full-procedure data and deployment-relevant operating points. Source code is available at https://github.com/sdoerrich97/true-colon.

Problem

• Colonoscopy CADe
• Real-world deployment
• Transfer asymmetry
• Curated datasets
• Benchmarking protocol

Innovation

• Transfer Asymmetry
• Full-procedure Benchmarking
• Real-time Polyp Detection
• TRUE-Colon
• Deployment-relevant Evaluation