Chromosome Mask-Conditioned Generative Inpainting for Atypical Mitosis Classification

Sweta Banerjee, Viktoria Weiss, Thomas Conrad, Taryn A. Donovan, Jonas Ammeling, Rutger H.J. Fick, Jonas Utz, Robert Klopfleisch, Christopher Kaltenecker, Christof A. Bertram, Katharina Breininger, Marc Aubreville
Proceedings of the MICCAI Workshop on Computational Pathology, PMLR 316:266-277, 2026.

Abstract

Atypical mitoses are critical prognostic markers for tumor proliferation, yet classification efforts are compromised by class imbalance, data scarcity, and noisy labels. Our work focuses on hematoxylin and eosin (H&E)-stained histopathology images, where identifying such mitoses is particularly challenging due to overlapping morphological features and stain variability. We address these challenges with a novel approach for biologically informed inpainting, conditioned on a histological context patch, an inpainting mask, and a chromosome segmentation mask. This triple-conditioned generative strategy allows disentanglement of the mitotic figure shape information from the cellular context and enables the utilization of large-scale datasets that do not contain atypical sub-classification for training classification models. We evaluate both adversarial and denoising diffusion-based inpainting strategies.Our approach mitigates the lack of data diversity and label noise, thereby substantially improving classification performance for atypical vs. normal mitoses - as demonstrated by downstream classification with EfficientNet-B0 and Low-rank adaptation (LoRA) finetuned foundation models. We provide the complete source code, including all our methods, at our github repository: https://github.com/DeepMicroscopy/ChroMa-GI.

Cite this Paper

BibTeX
@InProceedings{pmlr-v316-banerjee26a, title = {Chromosome Mask-Conditioned Generative Inpainting for Atypical Mitosis Classification}, author = {Banerjee, Sweta and Weiss, Viktoria and Conrad, Thomas and Donovan, Taryn A. and Ammeling, Jonas and Fick, Rutger H.J. and Utz, Jonas and Klopfleisch, Robert and Kaltenecker, Christopher and Bertram, Christof A. and Breininger, Katharina and Aubreville, Marc}, booktitle = {Proceedings of the MICCAI Workshop on Computational Pathology}, pages = {266--277}, year = {2026}, editor = {Studer, Linda and Ciompi, Francesco and Khalili, Nadieh and Faryna, Khrystyna and Faryna, Khrystyna and Yeong, Joe and Lau, Mai Chan and Chen, Hao and Liu, Ziyi and Brattoli, Biagio}, volume = {316}, series = {Proceedings of Machine Learning Research}, month = {27 Sep}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v316/main/assets/banerjee26a/banerjee26a.pdf}, url = {https://proceedings.mlr.press/v316/banerjee26a.html}, abstract = {Atypical mitoses are critical prognostic markers for tumor proliferation, yet classification efforts are compromised by class imbalance, data scarcity, and noisy labels. Our work focuses on hematoxylin and eosin (H&E)-stained histopathology images, where identifying such mitoses is particularly challenging due to overlapping morphological features and stain variability. We address these challenges with a novel approach for biologically informed inpainting, conditioned on a histological context patch, an inpainting mask, and a chromosome segmentation mask. This triple-conditioned generative strategy allows disentanglement of the mitotic figure shape information from the cellular context and enables the utilization of large-scale datasets that do not contain atypical sub-classification for training classification models. We evaluate both adversarial and denoising diffusion-based inpainting strategies.Our approach mitigates the lack of data diversity and label noise, thereby substantially improving classification performance for atypical vs. normal mitoses - as demonstrated by downstream classification with EfficientNet-B0 and Low-rank adaptation (LoRA) finetuned foundation models. We provide the complete source code, including all our methods, at our github repository: https://github.com/DeepMicroscopy/ChroMa-GI.} }
Endnote
%0 Conference Paper %T Chromosome Mask-Conditioned Generative Inpainting for Atypical Mitosis Classification %A Sweta Banerjee %A Viktoria Weiss %A Thomas Conrad %A Taryn A. Donovan %A Jonas Ammeling %A Rutger H.J. Fick %A Jonas Utz %A Robert Klopfleisch %A Christopher Kaltenecker %A Christof A. Bertram %A Katharina Breininger %A Marc Aubreville %B Proceedings of the MICCAI Workshop on Computational Pathology %C Proceedings of Machine Learning Research %D 2026 %E Linda Studer %E Francesco Ciompi %E Nadieh Khalili %E Khrystyna Faryna %E Khrystyna Faryna %E Joe Yeong %E Mai Chan Lau %E Hao Chen %E Ziyi Liu %E Biagio Brattoli %F pmlr-v316-banerjee26a %I PMLR %P 266--277 %U https://proceedings.mlr.press/v316/banerjee26a.html %V 316 %X Atypical mitoses are critical prognostic markers for tumor proliferation, yet classification efforts are compromised by class imbalance, data scarcity, and noisy labels. Our work focuses on hematoxylin and eosin (H&E)-stained histopathology images, where identifying such mitoses is particularly challenging due to overlapping morphological features and stain variability. We address these challenges with a novel approach for biologically informed inpainting, conditioned on a histological context patch, an inpainting mask, and a chromosome segmentation mask. This triple-conditioned generative strategy allows disentanglement of the mitotic figure shape information from the cellular context and enables the utilization of large-scale datasets that do not contain atypical sub-classification for training classification models. We evaluate both adversarial and denoising diffusion-based inpainting strategies.Our approach mitigates the lack of data diversity and label noise, thereby substantially improving classification performance for atypical vs. normal mitoses - as demonstrated by downstream classification with EfficientNet-B0 and Low-rank adaptation (LoRA) finetuned foundation models. We provide the complete source code, including all our methods, at our github repository: https://github.com/DeepMicroscopy/ChroMa-GI.
APA
Banerjee, S., Weiss, V., Conrad, T., Donovan, T.A., Ammeling, J., Fick, R.H., Utz, J., Klopfleisch, R., Kaltenecker, C., Bertram, C.A., Breininger, K. & Aubreville, M.. (2026). Chromosome Mask-Conditioned Generative Inpainting for Atypical Mitosis Classification. Proceedings of the MICCAI Workshop on Computational Pathology, in Proceedings of Machine Learning Research 316:266-277 Available from https://proceedings.mlr.press/v316/banerjee26a.html.

Related Material