Accurate determination of colorectal polyp size is critical for surveillance recommendations and for generating reliable ground truth in artificial intelligence (AI) development. Visual estimation is imprecise, and no validated reference standard exists. We aimed to assess the accuracy and reproducibility of real-time digital microscopic measurement of fresh polypectomy specimens.

In this prospective study at the Centre hospitalier de l’Université de Montréal (CHUM), 70 polyps from 44 patients (mean age 65.4 years; 52.3% female) were measured on-site using a calibrated digital microscope. Three independent raters, blinded to each other, obtained long- and short-axis measurements. The primary outcome was inter-rater reliability for long-axis measurements. Secondary outcomes included short-axis reliability, overall agreement, and classification (≤5 mm vs. >5 mm).

Inter-rater reliability was excellent for both long-axis (ICC 0.941, 95% CI 0.914–0.961) and short-axis (ICC 0.943, 95% CI 0.917–0.962) measurements. Bland–Altman plots demonstrated mean differences near zero without systematic bias. Limits of agreement were within ±1 mm. When dichotomized at the 5 mm threshold, agreement was perfect (κ=1.0).

On-site digital microscopy provides highly reproducible and verifiable size data with photographic documentation. This method should be considered a reference standard for accurate polyp sizing in clinical research and AI model development.