{"entity":{"id":"ai-mitosis-detection","kind":"technology","name":"AI mitosis detection and grading","aka":["mitotic figure detection","AI mitotic count","automated mitosis counting","mitosis detection AI"],"tldr":"Software that finds dividing cells on a digitised biopsy slide and counts them. The count is part of how many cancers are graded, and people counting by eye often disagree.","summary":"The mitotic count is a component of the Nottingham grade of breast cancer and of risk scores for tumours such as gastrointestinal stromal tumours, and pathologists agree on it poorly. Automated detection on whole-slide images has been driven by open challenges. TUPAC16 (Veta and colleagues, Medical Image Analysis 2019) asked teams to predict mitotic scores from 821 breast cancer whole-slide images; the best method reached a quadratic-weighted kappa of 0.567 against the reference. The MIDOG 2021 challenge (Aubreville and colleagues, Medical Image Analysis 2023) targeted the drop in accuracy when a model meets a new slide scanner: trained on 200 cases from four scanners and tested on 100 cases including two unseen scanners, the winning algorithm reached an F1 score of 0.748 (95 per cent CI 0.704 to 0.781), above six experts on the same task.\n\nMitosis counting is one task within broader AI pathology scoring. The barrier to use is generalisation across scanners, stains and laboratories, and validation of grades rather than counts against patient outcomes.","status":"emerging","asOf":"2026-11-03","links":[{"label":"Veta et al. 2019, Medical Image Analysis: predicting breast tumour proliferation from whole-slide images, the TUPAC16 challenge","url":"https://doi.org/10.1016/j.media.2019.02.012"},{"label":"Aubreville et al. 2023, Medical Image Analysis: mitosis domain generalisation in histopathology images, the MIDOG challenge","url":"https://doi.org/10.1016/j.media.2022.102699"}],"tags":[],"related":[],"cancers":["breast-cancer"],"sections":["diagnostics","ai-computation"],"technologies":["ai-pathology-scoring","digital-pathology-ai"],"targets":[],"drugs":[],"companies":[],"institutions":[],"pathways":["mitotic-spindle-checkpoint"],"terms":["nottingham-grade","gist-risk-stratification","tumour-grade","digital-pathology-wsi"],"trials":[],"people":[],"bottlenecks":[],"keyPapers":[],"journals":[],"dependsOn":[],"notes":[],"principle":"Object detection networks scan high-power fields of a whole-slide image for mitotic figures, distinguish them from look-alikes (apoptotic and pyknotic nuclei), and aggregate counts per area into a score or grade component.","strengths":["Reduces inter-observer variability in a key grading input","Counts across the whole slide rather than chosen fields","Open benchmarks allow comparison"],"limitations":["Accuracy falls on unseen scanners and stains","Few validations against patient outcomes","Pathologist remains responsible for the grade"],"since":2016},"route":"/technologies/ai-mitosis-detection/","neighbours":{"cancer":[{"id":"breast-cancer","kind":"cancer","name":"Breast cancer (all types)","route":"/cancers/breast-cancer/"}],"section":[{"id":"ai-computation","kind":"section","name":"AI & Computation","route":"/fronts/ai-computation/"},{"id":"diagnostics","kind":"section","name":"Diagnostics & Biomarkers","route":"/fronts/diagnostics/"}],"technology":[{"id":"ai-pathology-scoring","kind":"technology","name":"AI scoring of biomarkers and grade on pathology slides","route":"/technologies/ai-pathology-scoring/"},{"id":"digital-pathology-ai","kind":"technology","name":"Digital pathology & AI","route":"/technologies/digital-pathology-ai/"}],"pathway":[{"id":"mitotic-spindle-checkpoint","kind":"pathway","name":"Mitosis & the spindle assembly checkpoint","route":"/pathways/mitotic-spindle-checkpoint/"}],"term":[{"id":"digital-pathology-wsi","kind":"term","name":"Digital pathology and whole-slide images (WSI)","route":"/terms/digital-pathology-wsi/"},{"id":"gist-risk-stratification","kind":"term","name":"GIST risk stratification (mitotic count, size, site; Miettinen and modified NIH criteria)","route":"/terms/gist-risk-stratification/"},{"id":"tumour-grade","kind":"term","name":"Grade","route":"/terms/tumour-grade/"},{"id":"nottingham-grade","kind":"term","name":"Nottingham grade (breast cancer grade 1, 2 and 3)","route":"/terms/nottingham-grade/"}]}}