Abstract
<title>Abstract</title> <p>Mitotic count per 2 mm² is a key criterion for grading well-differentiated neuroendocrine tumors (NETs). In conventional light microscopy, this is approximated by counting mitoses across approximately 10 high-power fields (HPFs); however, HPFs are microscope-dependent constructs with no inherent equivalent in whole slide images (WSIs). To navigate this challenge, a pilot study was conducted, using 20 gastrointestinal and pancreatic NETs, comparing conventional manual mitotic counting with two digitally defined 2 mm² approaches; a calibrated grid overlay and a freehand calibrated annotation. Digital methods produced modest numeric increases in mitotic counts (mean: manual 1.75, grid 2.05, annotation 2.05) without causing grade upgrades in any case. Wilcoxon signed-rank testing demonstrated consistent directional increases that did not reach statistical significance (p = 0.29 and p = 0.91, respectively). Intraclass correlation coefficients (ICC) showed good-to-excellent intermethod agreement across all method pairs (0.82–0.97). Mean inter-method coefficient of variation was 47%, with greater proportional variability observed in low-count cases. Digitally defined approaches required more than twice the assessment time of conventional counting (7.3–8.1 vs 3.1 minutes per slide). These findings underscore the need for digitally native operational standards and robust automated mitosis detection to support reproducible grading in the digital era.</p>