Abstract

Contrast-free myocardial scar segmentation from cine cardiovascular magnetic resonance (CMR) is challenging because scar appearance is subtle and associ-ated contractile dysfunction may be localized in both space and time. We pro-pose ABMP-Net, a nnU-Net backbone that combines phase-resolved bidirec-tional motion modeling with motion-residual-guided abnormality-aware aggre-gation. The input comprises 25 cine frames and automatically generated frame-wise myocardial and left-ventricular priors. The cardiac cycle is partitioned at a fixed end-systolic index, and a shared motion encoder models systolic con-traction and diastolic recovery in both temporal directions. Phase-level repre-sentations form a multi-scale motion pyramid, while latent acceleration, a damped second-order re-sidual, and directional conflict determine trajectory-specific temporal weights that retain transient motion abnormalities. The re-sulting motion evidence provides bounded residual conditioning to the segmen-tation encoder. On a fixed internal split of the 64 labeled CARE CineMyoPS cases (52 training and 12 validation), ABMP-Net achieved a Dice score of 56.31% and a Hausdorff distance of 27.08 mm. Relative to uniform temporal aggregation, abnormality-aware aggregation improved Dice by 1.49 percentage points and reduced Hausdorff distance by 1.11 mm. These results support phase-resolved, abnormality-preserving motion modeling for cine-based myo-cardial scar segmentation. The code is available at https://github.com/guantianhuang000307/ABMP-Net.

Links to Paper and Supplementary Materials

Main Paper (Open Access Version): https://papers.miccai.org/miccai-2026-sat/paper/CARE_020.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=DdOX0R46lD

BibTex

@InProceedings{HuaGua_ABMPNet_MICCAISAT2026,
        author = { Huang, Guantian AND He, Dianning AND Liu, Ting AND Cui, Xiaoyu},
        title = { { ABMP-Net: Abnormality-Aware Bidirectional Motion-Physics Network for Cine CMR-Based Myocardial Scar Segmentation } },
        booktitle = {Medical Image Computing and Computer Assisted Intervention -- MICCAI 2026 Workshops and Challenges},
        year = {2026},
        publisher = {Springer Nature Switzerland},
        volume = {LNCS 17273},
        month = {pending},
        page = {pending}
}


back to top