Abstract
Cardiac magnetic resonance (CMR) segmentation is critical for quantitative assessment of ventricular function and myocardial tissue characterization, yet robust segmentation remains challenging due to variations across sequences, imaging planes, and acquisition centers. The MICCAI 2026 Universal Multi-Sequence, Multi-Center and Multi-View CMR Segmentation Challenge further requires segmentation masks and clinical measurements to be produced across cine (CINE) and late gadolinium enhancement (LGE) data with different view-specific label spaces. To address this challenge, we propose DINO-CMR, a task-specific framework that combines volumetric CINE segmentation, view-hybrid LGE segmentation, and decoupled measurement estimation. The CINE branch employs a masked-autoencoder-pretrained MedNeXt-S for volumetric segmentation with phase-aware ejection fraction estimation. The LGE branch selects the strongest segmentation source for each view and retains scar-mass estimates from a separate DINOv3-based path when segmentation and measurement objectives favor different models. Five-fold ensembling with test-time augmentation is applied before generating final masks and clinical measurements. On the challenge validation set, DINO-CMR (Team: NPM) achieves task scores of 0.6477 for CINE and 0.6937 for LGE, with an overall score of 0.6707.
Links to Paper and Supplementary Materials
Main Paper (Open Access Version): https://papers.miccai.org/miccai-2026-sat/paper/CMRSeg_041.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=PRmTJRY16k
BibTex
@InProceedings{ZuXua_DINOCMR_MICCAISAT2026,
author = { Zu, Xuan AND Xu, Qing AND He, Xiangjian AND Chen, Zhen},
title = { { DINO-CMR: DINOv3-Driven View-Hybrid Segmentation with Decoupled Measurement for Multi-Sequence CMR } },
booktitle = {Medical Image Computing and Computer Assisted Intervention -- MICCAI 2026 Workshops and Challenges},
year = {2026},
publisher = {Springer Nature Switzerland},
volume = {LNCS 17268},
month = {pending},
page = {pending}
}