Abstract
Mitral valve segmentation across cardiac computed tomography, three-dimensional transesophageal echocardiography, and echocardiographic videos remains challenging because of substantial modality differences, limited annotations, complex anatomical boundaries, and distant false-positive regions. This study presents a task-adaptive segmentation framework for multi-modal and multi-dimensional mitral valve analysis. For Task 1, an exponential moving average teacher-based semi-supervised strategy is employed for cardiac CT segmentation. Composite segmentation losses, confidence-based pseudo-label filtering, sliding-window inference, and test-time augmentation are integrated to improve robustness under limited supervision. For Task 2, a three-dimensional full-resolution nnU-Net v2 is developed for multi-structure segmentation in transesophageal echocardiography, while selective multi-fold ensembling is used to enhance the delineation of complex anatomical structures and boundaries. For Task 3, U-Net++ with an EfficientNet-B4 encoder is adopted for frame-wise echocardiographic video segmentation. Dice-Focal loss, video-level data partitioning, multi-view test-time augmentation, and distant false-positive suppression are combined to improve segmentation reliability. Experimental results demonstrate that the proposed framework effectively accommodates different imaging modalities, spatial dimensions, and supervision settings, achieving stable and competitive performance across all three mitral valve segmentation tasks.
Links to Paper and Supplementary Materials
Main Paper (Open Access Version): https://papers.miccai.org/miccai-2026-sat/paper/MVAA_073.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=XTkvc1WkRv
BibTex
@InProceedings{YanDon_Comprehensive_MICCAISAT2026,
author = { Yang, Donglin AND Cui, Yankun AND Zhang, Yi AND Hu, Wei AND Zhang, Wenfeng AND Zeng, Pan AND Yang, Wen},
title = { { Comprehensive Multi-Modal Mitral Valve Segmentation across Cardiac CT and Echocardiography } },
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}
}