Abstract
We describe our solution to the MICCAI 2026 Mitral Valve Anatomy Analysis (MVAA) challenge, which asks participants to segment the mitral valve across three heterogeneous imaging settings: (i) 3D cardiac CT under a semi-supervised regime, (ii) 3D transesophageal echocardiography (TEE) under full supervision, and (iii) 2D surgical endoscopic video under a semi-supervised regime. Rather than committing to a single architecture, we treat every task independently and combine self-configuring 3D~nnU-Net models, a matched reimplementation of its plain-convolution U-Net, and transformer-based SwinUNETR encoders through probability-level ensembling with 8-way test-time augmentation and connected-component post-processing. For TEE we exploit a non-overlapping external echocardiography corpus to enlarge the training set. For video we use a multi-architecture ensemble regularized by a Mean Teacher, refined with ensemble pseudo-labels, instrument copy-paste augmentation and instrument-free view generation. A recurring finding, quantified across many leaderboard submissions, is a weak and sometimes inverted correlation between local cross-validation and the hidden test set. Our final system attains leaderboard Dice scores of 0.863 (CT), 0.863 (TEE) and 0.823 (video), ranking at or near the top on all three tasks. Every source code and model weights will be publicly available.
Links to Paper and Supplementary Materials
Main Paper (Open Access Version): https://papers.miccai.org/miccai-2026-sat/paper/MVAA_047.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=DVGIkta2ef
BibTex
@InProceedings{KikShu_Solution_MICCAISAT2026,
author = { Kikuchi, Shunsuke},
title = { { Solution for MVAA challenge 2026: Ensemble, Tool-paste Augmentation and Tool-free View Generation } },
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}
}