Abstract
Mitral valve segmentation is clinically important for quantitative valve assessment, intervention planning, and computer-assisted mitral valve repair. This challenge addresses mitral valve segmentation across three heterogeneous imaging settings: semi-supervised 3D cardiac CT segmentation, fully supervised 3D transesophageal echocardiography (TEE) segmentation, and semi-supervised frame-level surgical video segmentation. In this work, we present a practical competition solution tailored to these three tracks. For the semi-supervised tracks, we adopt an offline pseudo-labeling strategy to provide stable supervision for unlabeled data. Supervised models are first trained on limited annotations and then ensembled to generate pseudo labels. To reduce the effect of noisy predictions, we further apply uncertainty-based sample selection and task-specific pseudo-label refinement before semi-supervised retraining. For the fully supervised 3D TEE track, we build a robust Auto3DSeg-based pipeline with five-fold cross-validation, ensemble inference, test-time augmentation, and connected-component post-processing. Across all tracks, our design focuses on maximizing supervision from limited annotations while effectively exploiting unlabeled data when available. At the time of submission, our solution achieved the top performance on the validation leaderboard.
Links to Paper and Supplementary Materials
Main Paper (Open Access Version): https://papers.miccai.org/miccai-2026-sat/paper/MVAA_022.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=pvacHzocId
BibTex
@InProceedings{XuJie_From_MICCAISAT2026,
author = { Xu, Jie AND Guo, Bowen AND Li, Wutong AND Li, Shangkun AND Guo, Yi AND Wang, Yuanyuan AND Li, Zeju},
title = { { From CT to Surgical Video: Robust Mitral Valve Segmentation across Multimodal Clinical Imaging } },
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}
}
