In this section
@inproceedings{Hu:2027:10.1007/978-3-032-38186-6_51,
author = {Hu, L and Wang, Z and Wu, Y and Zhang, Z and Wang, A and Yuan, R and Wang, F and Yap, CH and Yang, G},
doi = {10.1007/978-3-032-38186-6_51},
pages = {538--548},
title = {Geometry-Aware Self-supervised Learning for Whole-Heart 3D+t Representation},
url = {http://dx.doi.org/10.1007/978-3-032-38186-6_51},
year = {2027}
}
TY - CPAPER
AB - Cardiac magnetic resonance (CMR) is a clinical reference standard for assessing cardiac morphology and function. Whole-heart assessment relies on sparse multi-plane cine acquisitions from complementary short-axis (SAX) and 2/3/4-chamber long-axis (LAX) views across the cardiac cycle. Although large-scale cohorts such as UK Biobank provide tens of thousands of such studies, most scans remain unlabeled, and learning compact whole-heart representations from the resulting high-dimensional 3D+t data is still challenging under practical memory constraints. To address this, we introduce a geometry-aware self-supervised learning framework for whole-heart representation learning from sparse multi-plane cine CMR. First, we inject patient-space coordinates into spatio-temporal tubelet tokens, which preserves explicit spatial identity and enables shared per-plane encoding. Next, we explicitly model cross-plane structure by aggregating multi-plane features into a latent 3D volume via voxel-query cross-attention. To ensure geometric consistency while keeping supervision efficient, we reconstruct masked tubelets by coordinate-constrained trilinear reprojection from the latent volume. Finally, we adopt hierarchical masking that combines intra-plane tubelet masking with inter-plane plane dropping, turning reconstruction into cross-plane completion and strengthening global 3D reasoning. Experiments show that coherent 3D anatomy can be decoded from the learned latent volume using a lightweight linear probe, and that masked-plane reconstruction recovers missing views from remaining planes. Moreover, our method improves downstream multi-plane segmentation over state-of-the-art methods, while maintaining favorable efficiency under the same input setting. Code is available at this URL.
AU - Hu,L
AU - Wang,Z
AU - Wu,Y
AU - Zhang,Z
AU - Wang,A
AU - Yuan,R
AU - Wang,F
AU - Yap,CH
AU - Yang,G
DO - 10.1007/978-3-032-38186-6_51
EP - 548
PY - 2027///
SN - 0302-9743
SP - 538
TI - Geometry-Aware Self-supervised Learning for Whole-Heart 3D+t Representation
UR - http://dx.doi.org/10.1007/978-3-032-38186-6_51
ER -
For enquiries about the MRI Physics Collective, please contact:
Mary Finnegan
Senior MR Physicist at the ImperiaCollege Healthcare NHS Trust
Pete Lally
Assistant Professor in Magnetic Resonance (MR) Physics at ImperiaCollege
Jan Sedlacik
MR Physicist at the Robert Steiner MR Unit, Hammersmith Hospital Campus