Abstract
Mitral valve disease is assessed and treated along a pathway producing several kinds of image: cardiac computed tomography before the operation, transesophageal echocardiography during it, and video from the surgical field. A method built for one modality covers only one step, so a single pipeline spanning all three is needed to follow a patient from diagnosis to intervention. We describe such a pipeline: a self-configuring three-dimensional network for the volumetric modalities, a mean-teacher ensemble of two-dimensional networks with ImageNet-pretrained encoders for the video, a label audit and cleaning step, and post-processing that removes small isolated regions and replaces empty predictions with a fallback mask. It was developed on the Mitral Valve Anatomy Analysis (MVAA) 2026 challenge data: 27 labelled and 1040 unlabelled computed tomography volumes, 105 labelled echocardiography volumes, and 180 labelled and 1379 unlabelled video frames. On the held-out test set it reached Dice 0.839, Hausdorff distance 5.43 mm and average surface distance 0.43 mm for computed tomography; 0.853, 9.19mm and 0.56mm for echocardiography; and 0.755, 236.29 pixels and 63.76 pixels for video. The largest gains came not from the network but from correcting noisy training labels and handling frames where the model returned an empty mask, cutting the video Hausdorff distance from 520.44 to 236.29 pixels.
Links to Paper and Supplementary Materials
Main Paper (Open Access Version): https://papers.miccai.org/miccai-2026-sat/paper/MVAA_112.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=Ff5JMEWYF4
BibTex
@InProceedings{KacSan_AMultimodal_MICCAISAT2026,
author = { Kachole, Sanket AND Bakas, Spyridon},
title = { { A Multimodal Segmentation Pipeline for Mitral Valve Anatomy Analysis in Cardiac CT, 3D Transesophageal Echocardiography, and Surgical Video } },
booktitle = {Medical Image Computing and Computer Assisted Intervention -- MICCAI 2026 Workshops and Challenges},
year = {2026},
publisher = {Springer Nature Switzerland},
volume = {LNCS 17269},
month = {pending},
page = {pending}
}