Abstract
Ultrasound is the primary imaging modality in cardiology. Despite the commercial availability of 3D ultrasound, 2D echocardiogra- phy remains the most widely adopted in clinical practice. To overcome the inherent loss of 3D spatial context, reconstructing the 3D heart from 2D ultrasound has become the focus of numerous recent studies. In this work, we propose a novel approach for reconstructing the 3D left ventricle by using its representation in the spherical harmonic basis to learn a shape prior via triangular transport map. During inference, this learned prior is combined with a data fidelity term to fit the reconstruction to 2D observations along the corresponding acquisition planes. We first train the prior on a cohort of healthy adults. Subsequently, using this exact prior, we evaluate our reconstruction approach on three test sets: a cohort of healthy adults, a young population cohort, and a cohort of patients diagnosed with heart failure. Our method achieves a point-to-point Mean Absolute Error (MAE) of 0.55 mm, 0.62 mm, and 1.05 mm, alongside a relative volume error of 0.018, 0.015, and 0.045 on the three cohorts, respectively. Furthermore, a Dice Score of 0.99 is consistently achieved on the 2D observation planes across all three test sets.
Links to Paper and Supplementary Materials
Main Paper (Open Access Version): https://papers.miccai.org/miccai-2026-sat/paper/STACOM2026_014.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/profile?id=~Abdelkhalak_CHETOUI1
BibTex
@InProceedings{CheAbd_TT3D_MICCAISAT2026,
author = { Chetoui, Abdelkhalak AND Lombardi, Damiano AND Morales, Hernán G. AND Hermida, Uxio AND Fernández, Miguel A.},
title = { { TT3D: Triangular Transport for 3D Left Ventricle Reconstruction } },
booktitle = {Medical Image Computing and Computer Assisted Intervention -- MICCAI 2026 Workshops and Challenges},
year = {2026},
publisher = {Springer Nature Switzerland},
volume = {LNCS 17267},
month = {pending},
page = {pending}
}