From Pixels to Histopathology: A Graph-Based Framework for Interpretable Whole Slide Image Analysis

Alexander Weers, Alexander H. Berger, Laurin Lux, Peter Schüffler, Daniel Rueckert, Johannes C. Paetzold
Proceedings of The 9th International Conference on Medical Imaging with Deep Learning, PMLR 315:1153-1177, 2026.

Abstract

The histopathological analysis of whole-slide images (WSIs) is fundamental to cancer diagnosis but is a time-consuming and expert-driven process. While deep learning methods show promising results, dominant patch-based methods artificially fragment tissue, ignore biological boundaries, and produce black-box predictions. We overcome these limitations with a novel framework that transforms gigapixel WSIs into tissue-boundary aligned graph representations and is interpretable by design. Our approach builds graph nodes from tissue regions that respect natural structures, not arbitrary grids. We introduce an adaptive graph coarsening technique, guided by learned embeddings, to efficiently merge homogeneous regions while preserving diagnostically critical details in heterogeneous areas. Each node is enriched with a compact, interpretable feature set capturing clinically-motivated priors. A graph attention network then performs diagnosis on this compact representation. We demonstrate strong performance on cancer staging and survival prediction, outperforming methods with similar data requirements. Crucially, our data-efficient model (requiring $>300\times$ less training data) achieves results competitive with a massive foundation model, while offering full interpretability through feature attribution.

Cite this Paper

BibTeX
@InProceedings{pmlr-v315-weers26a, title = {From Pixels to Histopathology: A Graph-Based Framework for Interpretable Whole Slide Image Analysis}, author = {Weers, Alexander and Berger, Alexander H. and Lux, Laurin and Sch\"uffler, Peter and Rueckert, Daniel and Paetzold, Johannes C.}, booktitle = {Proceedings of The 9th International Conference on Medical Imaging with Deep Learning}, pages = {1153--1177}, year = {2026}, editor = {Huo, Yuankai and Gao, Mingchen and Kuo, Chang-Fu and Jin, Yueming and Deng, Ruining}, volume = {315}, series = {Proceedings of Machine Learning Research}, month = {08--10 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v315/main/assets/weers26a/weers26a.pdf}, url = {https://proceedings.mlr.press/v315/weers26a.html}, abstract = {The histopathological analysis of whole-slide images (WSIs) is fundamental to cancer diagnosis but is a time-consuming and expert-driven process. While deep learning methods show promising results, dominant patch-based methods artificially fragment tissue, ignore biological boundaries, and produce black-box predictions. We overcome these limitations with a novel framework that transforms gigapixel WSIs into tissue-boundary aligned graph representations and is interpretable by design. Our approach builds graph nodes from tissue regions that respect natural structures, not arbitrary grids. We introduce an adaptive graph coarsening technique, guided by learned embeddings, to efficiently merge homogeneous regions while preserving diagnostically critical details in heterogeneous areas. Each node is enriched with a compact, interpretable feature set capturing clinically-motivated priors. A graph attention network then performs diagnosis on this compact representation. We demonstrate strong performance on cancer staging and survival prediction, outperforming methods with similar data requirements. Crucially, our data-efficient model (requiring $>300\times$ less training data) achieves results competitive with a massive foundation model, while offering full interpretability through feature attribution.} }
Endnote
%0 Conference Paper %T From Pixels to Histopathology: A Graph-Based Framework for Interpretable Whole Slide Image Analysis %A Alexander Weers %A Alexander H. Berger %A Laurin Lux %A Peter Schüffler %A Daniel Rueckert %A Johannes C. Paetzold %B Proceedings of The 9th International Conference on Medical Imaging with Deep Learning %C Proceedings of Machine Learning Research %D 2026 %E Yuankai Huo %E Mingchen Gao %E Chang-Fu Kuo %E Yueming Jin %E Ruining Deng %F pmlr-v315-weers26a %I PMLR %P 1153--1177 %U https://proceedings.mlr.press/v315/weers26a.html %V 315 %X The histopathological analysis of whole-slide images (WSIs) is fundamental to cancer diagnosis but is a time-consuming and expert-driven process. While deep learning methods show promising results, dominant patch-based methods artificially fragment tissue, ignore biological boundaries, and produce black-box predictions. We overcome these limitations with a novel framework that transforms gigapixel WSIs into tissue-boundary aligned graph representations and is interpretable by design. Our approach builds graph nodes from tissue regions that respect natural structures, not arbitrary grids. We introduce an adaptive graph coarsening technique, guided by learned embeddings, to efficiently merge homogeneous regions while preserving diagnostically critical details in heterogeneous areas. Each node is enriched with a compact, interpretable feature set capturing clinically-motivated priors. A graph attention network then performs diagnosis on this compact representation. We demonstrate strong performance on cancer staging and survival prediction, outperforming methods with similar data requirements. Crucially, our data-efficient model (requiring $>300\times$ less training data) achieves results competitive with a massive foundation model, while offering full interpretability through feature attribution.
APA
Weers, A., Berger, A.H., Lux, L., Schüffler, P., Rueckert, D. & Paetzold, J.C.. (2026). From Pixels to Histopathology: A Graph-Based Framework for Interpretable Whole Slide Image Analysis. Proceedings of The 9th International Conference on Medical Imaging with Deep Learning, in Proceedings of Machine Learning Research 315:1153-1177 Available from https://proceedings.mlr.press/v315/weers26a.html.

Related Material