A large meta-analysis from NYU Langone Health shows that visual estimation of colorectal polyp size is frequently wrong, differing from standardized measurements by an average of nearly 1.7 mm and achieving only 60% accuracy overall. Errors worsened with larger polyps — which were often underestimated — raising the risk of inappropriate surveillance intervals and missed cancers.
Notably, AI-based measurement tools dramatically outperformed human estimation, improving accuracy more than sevenfold, pointing to a clear opportunity to standardize polyp sizing and reduce downstream screening errors.