Abstract
Mitral valve repair is planned and guided on three very different images of the same anatomy: preoperative cardiac computed tomography (CT), intraoperative 3D transesophageal echocardiography (3D TEE), and the surgeon’s view of the valve. Delineating the valve consistently in all three is the perception layer a patient-specific model of that valve rests on, yet annotations are scarce in every modality and the gap between them is wide. We introduce BEAT, a boundary-aware efficient anatomy-transfer recipe, instantiated once per modality from four ideas. It transfers strong pretrained backbones, a CT-pretrained STUNet for the volumetric tasks and an ImageNet-pretrained EfficientNet for video. It supervises anatomical boundaries directly with train-only distance and topology losses, since two of the three official metrics measure surface error. It bridges the small-labeled cross-domain gap with augmentation tailored to the underlying physical shift: a reconstruction-kernel model for CT and a photometric-plus-Fourier model for video. And it confines every computationally expensive component to training, leaving a lightweight single-pass deployment model. On the MICCAI 2026 MVAA challenge, BEAT achieves Dice scores of 0.85 on CT, 0.83 on 3D TEE, and 0.77 on surgical video, and ablations isolate the contribution of each component
Links to Paper and Supplementary Materials
Main Paper (Open Access Version): https://papers.miccai.org/miccai-2026-sat/paper/MVAA_060.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=yRMVmTyr3S
BibTex
@InProceedings{RegAth_BEAT_MICCAISAT2026,
author = { Rege, Atharva Atul AND Dukre, Adinath Madhavrao AND Shah, Sarth Santosh AND Razzak, Imran},
title = { { BEAT: Boundary-aware Efficient Anatomy-Transfer for Multimodal Mitral Valve Segmentation } },
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}
}