PAGET: Hierarchical Multi-Teacher Knowledge Distillation for Comprehensive Tumor Microenvironment Segmentation

Daisuke Komura, Maki Takao, Mieko Ochi, Takumi Onoyama, Hiroto Katoh, Hiroyuki Abe, Hiroyuki Sano, Teppei Konishi, Toshio Kumasaka, Tomoyuki Yokose, Yohei Miyagi, Tetsuo Ushiku, Shumpei Ishikawa
Proceedings of The 9th International Conference on Medical Imaging with Deep Learning, PMLR 315:256-279, 2026.

Abstract

Comprehensive characterization of the tumor microenvironment (TME) from H&E-stained histopathology images remains challenging due to the diversity of cellular components and limitations of current segmentation methods. We present PAGET (Pathological image segmentation via AGgrEgated Teachers), a multi-teacher knowledge distillation framework that enables simultaneous segmentation of 13 TME components from a single efficient model. Our key insight is that teacher predictions should be aggregated following the biological taxonomy of cell types—from tissue-level context through major cell categories to specific subtypes—rather than simple voting. By training specialized teachers on immunohistochemical restaining data and distilling their aggregated knowledge, the resulting student model not only matches but consistently outperforms the teacher ensemble on external datasets. We provide two complementary variants: PAGET-S for rapid semantic segmentation and PAGET-H for detailed panoptic segmentation. Extensive evaluation across three external datasets demonstrates robust generalization.

Cite this Paper

BibTeX
@InProceedings{pmlr-v315-komura26a, title = {PAGET: Hierarchical Multi-Teacher Knowledge Distillation for Comprehensive Tumor Microenvironment Segmentation}, author = {Komura, Daisuke and Takao, Maki and Ochi, Mieko and Onoyama, Takumi and Katoh, Hiroto and Abe, Hiroyuki and Sano, Hiroyuki and Konishi, Teppei and Kumasaka, Toshio and Yokose, Tomoyuki and Miyagi, Yohei and Ushiku, Tetsuo and Ishikawa, Shumpei}, booktitle = {Proceedings of The 9th International Conference on Medical Imaging with Deep Learning}, pages = {256--279}, 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/komura26a/komura26a.pdf}, url = {https://proceedings.mlr.press/v315/komura26a.html}, abstract = {Comprehensive characterization of the tumor microenvironment (TME) from H&E-stained histopathology images remains challenging due to the diversity of cellular components and limitations of current segmentation methods. We present PAGET (Pathological image segmentation via AGgrEgated Teachers), a multi-teacher knowledge distillation framework that enables simultaneous segmentation of 13 TME components from a single efficient model. Our key insight is that teacher predictions should be aggregated following the biological taxonomy of cell types—from tissue-level context through major cell categories to specific subtypes—rather than simple voting. By training specialized teachers on immunohistochemical restaining data and distilling their aggregated knowledge, the resulting student model not only matches but consistently outperforms the teacher ensemble on external datasets. We provide two complementary variants: PAGET-S for rapid semantic segmentation and PAGET-H for detailed panoptic segmentation. Extensive evaluation across three external datasets demonstrates robust generalization.} }
Endnote
%0 Conference Paper %T PAGET: Hierarchical Multi-Teacher Knowledge Distillation for Comprehensive Tumor Microenvironment Segmentation %A Daisuke Komura %A Maki Takao %A Mieko Ochi %A Takumi Onoyama %A Hiroto Katoh %A Hiroyuki Abe %A Hiroyuki Sano %A Teppei Konishi %A Toshio Kumasaka %A Tomoyuki Yokose %A Yohei Miyagi %A Tetsuo Ushiku %A Shumpei Ishikawa %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-komura26a %I PMLR %P 256--279 %U https://proceedings.mlr.press/v315/komura26a.html %V 315 %X Comprehensive characterization of the tumor microenvironment (TME) from H&E-stained histopathology images remains challenging due to the diversity of cellular components and limitations of current segmentation methods. We present PAGET (Pathological image segmentation via AGgrEgated Teachers), a multi-teacher knowledge distillation framework that enables simultaneous segmentation of 13 TME components from a single efficient model. Our key insight is that teacher predictions should be aggregated following the biological taxonomy of cell types—from tissue-level context through major cell categories to specific subtypes—rather than simple voting. By training specialized teachers on immunohistochemical restaining data and distilling their aggregated knowledge, the resulting student model not only matches but consistently outperforms the teacher ensemble on external datasets. We provide two complementary variants: PAGET-S for rapid semantic segmentation and PAGET-H for detailed panoptic segmentation. Extensive evaluation across three external datasets demonstrates robust generalization.
APA
Komura, D., Takao, M., Ochi, M., Onoyama, T., Katoh, H., Abe, H., Sano, H., Konishi, T., Kumasaka, T., Yokose, T., Miyagi, Y., Ushiku, T. & Ishikawa, S.. (2026). PAGET: Hierarchical Multi-Teacher Knowledge Distillation for Comprehensive Tumor Microenvironment Segmentation. Proceedings of The 9th International Conference on Medical Imaging with Deep Learning, in Proceedings of Machine Learning Research 315:256-279 Available from https://proceedings.mlr.press/v315/komura26a.html.

Related Material