Abstract
Spatiotemporal cardiac magnetic resonance imaging (CMR) analysis and motion tracking suffer from out-of-slice patient motion, respiratory misalignment, and the lack of physical constraints on latent representations. In this work, we present NIMOSEF-R, a novel coordinate-based neural implicit representation (INR) that incorporates Riemannian manifold constraints and a slice alignment module for cardiac shape analysis and motion estimation. Unlike classical Euclidean coordinate networks (NIMOSEF-E), NIMOSEF-R restricts the shape latent codes to a hyperbolic Poincar'e ball to model hierarchical anatomical relationships, and the motion trajectories to a unit hypersphere to model cardiac cycle periodicity. Furthermore, a learnable slice correction layer computes rigid in-plane transformations for each slice along the Z-axis to correct slice misalignment during data acquisition. Evaluated on a cohort of 105 subjects from the ACDC dataset, NIMOSEF-R achieves significant improvements in 3D stack continuity ($p < 0.001$), halving CoM Z-continuity second differences and improving adjacent contour overlap compared to both original segmentation and rigid-aligned baselines. Furthermore, NIMOSEF-R yields smooth, physically plausible latent motion trajectories with a path length of $1.97$\,rad (vs $29.95$\,rad for NIMOSEF-E), capturing the physiological contraction-relaxation cycle (evidenced by a strain-like motion distance profile), and maintains high pathology classification accuracy under the hyperbolic metric ($0.55$ vs $0.43$ for NIMOSEF-E).
Links to Paper and Supplementary Materials
Main Paper (Open Access Version): https://papers.miccai.org/miccai-2026-sat/paper/Off_Grid_025.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=CfdGyqjgBj
BibTex
@InProceedings{CanGon_NIMOSEFR_MICCAISAT2026,
author = { Canastra, Gonçalo AND Gordaliza, Pedro M. AND van Heeswijk, Ruud B. AND Richiardi, Jonas AND Banus, Jaume},
title = { { NIMOSEF-R: Neural Implicit Motion and Segmentation Functions with Riemannian Embedding Priors and Breath-Motion Correction } },
booktitle = {Medical Image Computing and Computer Assisted Intervention -- MICCAI 2026 Workshops and Challenges},
year = {2026},
publisher = {Springer Nature Switzerland},
volume = {LNCS 17277},
month = {pending},
page = {pending}
}