Abstract
CMR-MULTI combines multi-view cine cardiac magnetic resonance segmentation, late gadolinium enhancement (LGE) segmentation, ejection fraction (EF) prediction, and scar-mass prediction in a single submission protocol. The task is constrained not only by segmentation accuracy, but also by clinical-output calibration and archive validity: a candidate can improve one view while degrading surface distances, or improve EF without changing any mask. We present a view-specialized nnU-Net-style system in which cine and LGE views are treated as separately validated branches, and candidate replacements are accepted only after a joint metric gate and a submission-structure audit. The method is built around three principles: preserve the strongest validated mask anchor, use view-specific branches only where they improve the relevant failure mode, and calibrate JSON clinical outputs after the mask package has been fixed. The paper also emphasizes evidence status: platform-confirmed segmentation results are separated from internal JSON-only calibration estimates. The platform-confirmed R100 package achieved an overall validation score of 0.7559, with Task 1 score 0.7052 and Task 2 score 0.8066. An internal R104 candidate keeps all R100 masks fixed and improves EF PCC from 0.9530 to 0.9614, giving an estimated overall score of 0.7571. Because R104 has not been platform-tested, it is reported as an EF-calibration candidate rather than as an official segmentation result.
Links to Paper and Supplementary Materials
Main Paper (Open Access Version): https://papers.miccai.org/miccai-2026-sat/paper/CMRSeg_025.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=oFyqH1NIaI
BibTex
@InProceedings{YanXin_ViewSpecialized_MICCAISAT2026,
author = { Yang, Xinyu},
title = { { View-Specialized CMR-MULTI Segmentation with Metric-Gated Clinical Output Calibration } },
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}
}