Abstract

MVAA 2026 evaluates mitral-valve segmentation in cardiac computed tomography (CT), three-dimensional transesophageal echocardiography (3D TEE), and surgical video. These modalities differ in spatial sampling, image appearance, target definition, and annotation scale; we therefore develop three task-specific segmentation pipelines instead of imposing a shared architecture. For CT, a semi-supervised SE-Residual U-Net, a large-kernel MedNeXt branch, and a larger-context distance-aware branch are combined by calibrated foreground-probability fusion. For 3D TEE, an anisotropy-aware residual encoder-decoder is trained in five folds, followed by logit-level fusion and class-aware connected-component refinement. For surgical video, a ReSurgSAM2-based video pipeline uses three fold-specific models, segmented bidirectional inference, and majority fusion. The final official hidden-test evaluation reports DSC/HD/ASD values of 0.8216/6.2431/0.5738 for CT, 0.8518/9.3992/0.5736 for 3D TEE, and 0.7291/630.8260/347.6523 for video under the official task-specific distance conventions. These results demonstrate that modality-matched spatial modeling and task-appropriate ensembling offer a practical approach to multimodal mitral-valve analysis.

Links to Paper and Supplementary Materials

Main Paper (Open Access Version): https://papers.miccai.org/miccai-2026-sat/paper/MVAA_059.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=GBP07tVNSb

BibTex

@InProceedings{YanHan_ModalityAware_MICCAISAT2026,
        author = { Yang, Hang AND Zhang, Xiaoxia AND Pan, Xing AND Lyu, Daoming},
        title = { { Modality-Aware Segmentation of Mitral Valve Anatomy in CT, 3D TEE, and Surgical Video } },
        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}
}


back to top