Abstract
Accurate prediction of 1p/19q status is important for glioma diagnosis and prognosis. Existing deep learning methods usually treat 1p/19q codeletion prediction as a binary classification task, ignoring the individual deletion status of 1p and 19q. In this paper, we propose D2TMIL, a disentangled dual-task multiple instance learning framework for 1p/19q status prediction from whole-slide histopathological images. The proposed method decomposes codeletion prediction into two subtasks, 1p deletion prediction and 19q deletion prediction, and estimates the final codeletion probability through their joint distribution. A dual-task cross-attention module and a mutual-information dual-task coupling module are introduced to enhance cross-task feature interaction and model task correlations. Experiments on a glioma dataset show that D2TMIL outperforms baseline methods in both single-arm deletion prediction and 1p/19q codeletion prediction.
Links to Paper and Supplementary Materials
Main Paper (Open Access Version): https://papers.miccai.org/miccai-2026-sat/paper/CMMCA2026_005.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=I35FOaiLzd
BibTex
@InProceedings{ZhaXia_D2TMIL_MICCAISAT2026,
author = { Zhao, Xiangyu AND Hou, Jiaxin AND Zhang, Ranran AND Hu, Hao AND Lu, Yanhua AND Jin, Ben AND He, Miaoxia AND Qin, Wenjian},
title = { { D2TMIL: Disentangled Dual-Task Multiple Instance Learning for 1p/19q Status Prediction from Whole Slide Images } },
booktitle = {Medical Image Computing and Computer Assisted Intervention -- MICCAI 2026 Workshops and Challenges},
year = {2026},
publisher = {Springer Nature Switzerland},
volume = {LNCS 17278},
month = {pending},
page = {pending}
}