Benchmarking Deep Learning and Vision Foundation Models for Atypical vs. Normal Mitosis Classification with Cross-Dataset Evaluation

Banerjee S, Weiss V, Donovan TA, Fick RHJ, Conrad T, Ammeling J, Porsche N, Klopfleisch R, Kaltenecker CC, Breininger K, Aubreville M, Bertram CA (2026)

Publication Status: Submitted

Publication Type: Unpublished / Preprint

Future Publication Type: Journal article

Publication year: 2026

DOI: 10.48550/arXiv.2506.21444

Abstract

Atypical mitosis marks a deviation in the cell division process that has been shown be an independent prognostic marker for tumor malignancy. However, atypical mitosis classification remains challenging due to low prevalence, at times subtle morphological differences from normal mitotic figures, low inter-rater agreement among pathologists, and class imbalance in datasets. Building on the Atypical Mitosis dataset for Breast Cancer (AMi-Br), this study presents a comprehensive benchmark comparing deep learning approaches for automated atypical mitotic figure (AMF) classification, including end-to-end trained deep learning models, foundation models with linear probing, and foundation models fine-tuned with low-rank adaptation (LoRA). For rigorous evaluation, we further introduce two new held-out AMF datasets - AtNorM-Br, a dataset of mitotic figures from the TCGA breast cancer cohort, and AtNorM-MD, a multi-domain dataset of mitotic figures from a subset of the MIDOG++ training set. We found average balanced accuracy values of up to 0.8135, 0.7788, and 0.7723 on the in-domain AMi-Br and the out-of-domain AtNorm-Br and AtNorM-MD datasets, respectively. Our work shows that atypical mitotic figure classification, while being a challenging problem, can be effectively addressed through the use of recent advances in transfer learning and model fine-tuning techniques

Additional Organisation(s)

Sonderforschungsbereich 1540/1 - Erforschung der Mechanik des Gehirns (EBM): Verständnis, Engineering und Nutzung mechanischer Eigenschaften und Signale in der Entwicklung, Physiologie und Pathologie des zentralen Nervensystems

Involved external institutions

Freie Universität Berlin DE Germany (DE) Medizinische Universität Wien AT Austria (AT) Fachhochschule Flensburg DE Germany (DE) Veterinärmedizinische Universität Wien (Vetmeduni Vienna) / University of Veterinary Medicine, Vienna AT Austria (AT) Julius-Maximilians-Universität Würzburg DE Germany (DE) Diffusely FR France (FR) Technische Hochschule Ingolstadt DE Germany (DE) The Schwarzman Animal Medical Center (AMC) US United States (USA) (US)

How to cite

APA:

Banerjee, S., Weiss, V., Donovan, T.A., Fick, R.H.J., Conrad, T., Ammeling, J.,... Bertram, C.A. (2026). Benchmarking Deep Learning and Vision Foundation Models for Atypical vs. Normal Mitosis Classification with Cross-Dataset Evaluation. (Unpublished, Submitted).

MLA:

Banerjee, Sweta, et al. Benchmarking Deep Learning and Vision Foundation Models for Atypical vs. Normal Mitosis Classification with Cross-Dataset Evaluation. Unpublished, Submitted. 2026.

BibTeX: Download