Abstract

The MVAA2026 challenge targets multi-modal mitral valve analysis across cardiac CT, 3D transesophageal echocardiography (TEE), and surgical video frames, requiring robust segmentation under limited annotations, modality-speciffc artifacts, and large anatomical variability. We present a validation-guided medical world modeling framework tai-lored to the three tasks. For Task 1, we develop a semi-supervised 3D CT segmentation pipeline that combines reffned pseudo-label learning, 3D CubeMix consistency, tri-view pseudo-label fusion, and volumetric masked anatomy modeling to exploit both labeled and unlabeled CT volumes. For Task 2, we address fully supervised 3D TEE leaffet seg-mentation using global-local modeling, boundary-aware shape guidance, and a ffve-fold SegResNet ensemble to improve robustness for thin valve structures. For Task 3, we build a frame-wise surgical image segmen-tation framework based on ImageNet-initialized Mean Teacher learning, SAM-guided pseudo-label reffnement, and prototype-guided pseudo-label denoising to reduce false positives around hollow cardiac chambers. On the offfcial validation platform, our ffnal submissions achieved Dice scores of 0.8418, 0.8047, and 0.7983 for Tasks 1–3, respectively. The correspond-ing HD and ASD were 5.6920 and 0.3122 for Task 1, 16.9773 and 0.9108 for Task 2, and 175.1968 and 20.3188 for Task 3. These results demon-strate the effectiveness of adapting medical world modeling principles to modality-speciffc anatomical segmentation across CT, ultrasound, and surgical imaging.

Links to Paper and Supplementary Materials

Main Paper (Open Access Version): https://papers.miccai.org/miccai-2026-sat/paper/MVAA_071.pdf

SharedIt Link: Not yet available

SpringerLink (DOI): Not yet available

Supplementary Material: https://papers.miccai.org/miccai-2026-sat/supp/MVAA_071_supp.pdf

Link to Open Review

Open Review Page: https://openreview.net/forum?id=84diNRzMLv

BibTex

@InProceedings{CheSih_TaskSpecific_MICCAISAT2026,
        author = { Chen, Sihan AND Chi, Jianning},
        title = { { Task-Specific Segmentation Frameworks for Multi-Modal Mitral Valve Tasks in MVAA2026 } },
        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