Abstract
Breast cancer is a leading cause of cancer-related mortality worldwide, and early detection through magnetic resonance imaging (MRI) is critical for improving patient outcomes. Nevertheless, accurate interpretation of breast MRI remains highly challenging, particularly in multi-center settings where variability in scanners, imaging protocols, and patient populations introduces substantial domain shifts. Such heterogeneity frequently results in inconsistent diagnoses and limits the generalizability of deep learning models trained on single-institution data. To address these limitations, we propose DAMST, a hybrid framework that integrates a DINOv2-based Medical Slice Transformer (MST) with adversarial domain adaptation for robust multi-center breast MRI analysis. We introduce a domain classifier that is trained to detect specific acquisition-site patterns and gradient reversal mechanism to learn general discriminative information across multi-center data. In addition, a progressive scheduling scheme is introduced to make adversarial training more stable and help the model adapt better to images from different hospitals by increasing the adversarial effect step by step during training. Experiments on five independent multi-center breast MRI datasets show that DAMST consistently surpasses state-of-the-art baselines in accuracy and interpretability. This highlights the promise of domain-aware architectures for reliable breast cancer detection across heterogeneous clinical environments.
Links to Paper and Supplementary Materials
Main Paper (Open Access Version): https://papers.miccai.org/miccai-2026-sat/paper/Deep_Brea3th_031.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=1WNH5byvQB
BibTex
@InProceedings{Abdull_DAMST_MICCAISAT2026,
author = { Abdullah AND Huang, Tao AND Lee, Ickjai AND Ahn, Euijoon},
title = { { DAMST: Domain-Adaptive Medical Slice Transformer for Multi-Center Breast MRI Classification } },
booktitle = {Medical Image Computing and Computer Assisted Intervention -- MICCAI 2026 Workshops and Challenges},
year = {2026},
publisher = {Springer Nature Switzerland},
volume = {LNCS 17256},
month = {pending},
page = {pending}
}