Abstract
Tumor Regression Grade (TRG) is a critical pathological indicator for evaluating treatment response and guiding subsequent therapy in esophageal cancer. It fundamentally relies on the descriptive assessment of the reciprocal progression between residual tumor and regression traces like fibrosis. However, current TRG assessment relies on subjective interpretation of H&E-stained slides, which is labor-intensive and prone to inter-observer variability. Although Multiple Instance Learning (MIL) has been widely adop-ted for whole-slide image analysis, existing methods rarely account for the specific pathological characteristics of TRG, often neglecting critical regressive signs and the intrinsic ordinality of the grading process. To address these limitations, we propose a novel MIL framework for esophageal cancer TRG assessment. We introduce a Dual-Branch Gated Attention architecture to disentangle malignant cellular patterns from regressive traces, emulating the diagnostic rationale of pathologists who contrast tumor residues against regression signs during assessment. Additionally, an Ordinal-Aware Regression module is incorporated to regularize the feature space to reflect ordinal continuity. Extensive experiments show that our framework significantly outperforms state-of-the-art MIL methods, highlighting its effectiveness. Our code will be made publicly available soon.
Links to Paper and Supplementary Materials
Main Paper (Open Access Version): https://papers.miccai.org/miccai-2026-sat/paper/MWM_004.pdf
SharedIt Link: Not yet available
SpringerLink (DOI): Not yet available
Supplementary Material: Not Submitted
Link to Open Review
Open Review Page: https://openreview.net/forum?id=mwNS8knZpX
BibTex
@InProceedings{ZhaLin_Pathologybased_MICCAISAT2026,
author = { Zhao, Lin AND Liu, Zhengjin AND Lin, Weiping AND Magnier, Baptiste AND Zhao, Ziqing AND Zhong, Jun AND Wang, Yong AND Li, Shouguo AND Lu, Haijie AND Wang, Liansheng},
title = { { Pathology-based Ordinal Multiple Instance Learning Framework for Esophageal Cancer Tumor Regression Grading } },
booktitle = {Medical Image Computing and Computer Assisted Intervention -- MICCAI 2026 Workshops and Challenges},
year = {2026},
publisher = {Springer Nature Switzerland},
volume = {LNCS 17262},
month = {pending},
page = {pending}
}