Uncertainty-Aware Ensemble Segmentation of Breast Cancer Tissue Microarrays

Lucia Schmidt-Santiago, Roman Kinakh, Sergio Carreras-Salinas, Sara Guerrero-Aspizcua, Gonzalo R. Ríos-Muñoz, Arrate Muñoz-Barrutia
Proceedings of the MICCAI Workshop on Computational Pathology, PMLR 316:39-51, 2026.

Abstract

Breast cancer Tissue Microarrays (TMAs) offer a high-throughput platform for studying tumor morphology and biomarker expression. We present an automated deep learning pipeline for semantic segmentation of Hematoxylin and Eosin (H&E)-stained breast cancer TMAs, integrating ensemble U-Net architectures with ResNet encoders and Monte Carlo Dropout (MCDO) for uncertainty estimation. A robust pre-processing workflow addresses illumination artifacts, staining variability, and tissue detection. Multiple U-Net models were trained using distinct loss functions to address class imbalance and feature iversity. Predictions were combined via soft voting, emulating consensus among pathologists. Uncertainty was quantified using MCDO across ensemble outputs, enhancing reliability and interpretability. Our pipeline outperforms similar methods such as WeGleNet (mIoU = 0.4368) and HistoSegNet (mIoU = 0.5505), achieving a mean IoU of 0.58 $\pm$ 0.11 and Dice Score of 0.66 $\pm$ 0.10. Calibration analysis shows superior alignment of standard deviation–based uncertainty estimates with actual prediction errors (UCE = 0.085 $\pm$ 0.033). This pipeline effectively segments complex histopathological structures and flags ambiguous regions for review, supporting downstream biomarker discovery and clinical interpretation.

Cite this Paper

BibTeX
@InProceedings{pmlr-v316-schmidt-santiago26a, title = {Uncertainty-Aware Ensemble Segmentation of Breast Cancer Tissue Microarrays}, author = {Schmidt-Santiago, Lucia and Kinakh, Roman and Carreras-Salinas, Sergio and Guerrero-Aspizcua, Sara and R\'{i}os-Mu\~{n}oz, Gonzalo R. and Mu\~{n}oz-Barrutia, Arrate}, booktitle = {Proceedings of the MICCAI Workshop on Computational Pathology}, pages = {39--51}, 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/schmidt-santiago26a/schmidt-santiago26a.pdf}, url = {https://proceedings.mlr.press/v316/schmidt-santiago26a.html}, abstract = {Breast cancer Tissue Microarrays (TMAs) offer a high-throughput platform for studying tumor morphology and biomarker expression. We present an automated deep learning pipeline for semantic segmentation of Hematoxylin and Eosin (H&E)-stained breast cancer TMAs, integrating ensemble U-Net architectures with ResNet encoders and Monte Carlo Dropout (MCDO) for uncertainty estimation. A robust pre-processing workflow addresses illumination artifacts, staining variability, and tissue detection. Multiple U-Net models were trained using distinct loss functions to address class imbalance and feature iversity. Predictions were combined via soft voting, emulating consensus among pathologists. Uncertainty was quantified using MCDO across ensemble outputs, enhancing reliability and interpretability. Our pipeline outperforms similar methods such as WeGleNet (mIoU = 0.4368) and HistoSegNet (mIoU = 0.5505), achieving a mean IoU of 0.58 $\pm$ 0.11 and Dice Score of 0.66 $\pm$ 0.10. Calibration analysis shows superior alignment of standard deviation–based uncertainty estimates with actual prediction errors (UCE = 0.085 $\pm$ 0.033). This pipeline effectively segments complex histopathological structures and flags ambiguous regions for review, supporting downstream biomarker discovery and clinical interpretation.} }
Endnote
%0 Conference Paper %T Uncertainty-Aware Ensemble Segmentation of Breast Cancer Tissue Microarrays %A Lucia Schmidt-Santiago %A Roman Kinakh %A Sergio Carreras-Salinas %A Sara Guerrero-Aspizcua %A Gonzalo R. Ríos-Muñoz %A Arrate Muñoz-Barrutia %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-schmidt-santiago26a %I PMLR %P 39--51 %U https://proceedings.mlr.press/v316/schmidt-santiago26a.html %V 316 %X Breast cancer Tissue Microarrays (TMAs) offer a high-throughput platform for studying tumor morphology and biomarker expression. We present an automated deep learning pipeline for semantic segmentation of Hematoxylin and Eosin (H&E)-stained breast cancer TMAs, integrating ensemble U-Net architectures with ResNet encoders and Monte Carlo Dropout (MCDO) for uncertainty estimation. A robust pre-processing workflow addresses illumination artifacts, staining variability, and tissue detection. Multiple U-Net models were trained using distinct loss functions to address class imbalance and feature iversity. Predictions were combined via soft voting, emulating consensus among pathologists. Uncertainty was quantified using MCDO across ensemble outputs, enhancing reliability and interpretability. Our pipeline outperforms similar methods such as WeGleNet (mIoU = 0.4368) and HistoSegNet (mIoU = 0.5505), achieving a mean IoU of 0.58 $\pm$ 0.11 and Dice Score of 0.66 $\pm$ 0.10. Calibration analysis shows superior alignment of standard deviation–based uncertainty estimates with actual prediction errors (UCE = 0.085 $\pm$ 0.033). This pipeline effectively segments complex histopathological structures and flags ambiguous regions for review, supporting downstream biomarker discovery and clinical interpretation.
APA
Schmidt-Santiago, L., Kinakh, R., Carreras-Salinas, S., Guerrero-Aspizcua, S., Ríos-Muñoz, G.R. & Muñoz-Barrutia, A.. (2026). Uncertainty-Aware Ensemble Segmentation of Breast Cancer Tissue Microarrays. Proceedings of the MICCAI Workshop on Computational Pathology, in Proceedings of Machine Learning Research 316:39-51 Available from https://proceedings.mlr.press/v316/schmidt-santiago26a.html.

Related Material