Abstract
Diffeomorphic deformable image registration is critical for the quantitative analysis of thoracic computed tomography (CT), yet recent learning-based methods promote diffeomorphisms through pairwise inverse-consistency regularization. This constraint, however, is selfreferential by construction: a transformation is validated solely against its own inverse, allowing anatomically implausible correspondences to pass as readily as correct ones. We introduce TrICON, a population-trained framework that generalizes pairwise inverse consistency to triplet-wise consistency, jointly enforcing invertibility, closed-loop cycle consistency, and transitivity across three images via a stochastic, differentiable formulation compatible with existing registration architectures. For a given triplet, our approach requires the direct transformation between two images to agree with the transformation composed through a third, auxiliary image, yielding a correctness signal decoupled from pairwise inverse consistency. On inter-patient lung CT registration, TrICON outperforms representative state of the art (SOTA) methods, achieving higher DSC and lower distance based errors (HD95 and ASSD), while maintaining competitive MCD performance, and preserving near-diffeomorphic smoothness and invertibility without any explicit spatial regularization. Code is available at https://github.com/creatis-myriad/TrICON.
Links to Paper and Supplementary Materials
Main Paper (Open Access Version): https://papers.miccai.org/miccai-2026-sat/paper/TIA_018.pdf
SharedIt Link: Not yet available
SpringerLink (DOI): Not yet available
Supplementary Material: Not Submitted
Link to Open Review
Open Review Page: Not Available
BibTex
@InProceedings{MarMar_TrICON_MICCAISAT2026,
author = { Marquez-Sosa, Maria AND Saillard, Émile AND Basarab, Adrian AND Sdika, Michaël},
title = { { TrICON: Learning Diffeomorphic Lung CT Registration through Triplet Consistency } },
booktitle = {Medical Image Computing and Computer Assisted Intervention -- MICCAI 2026 Workshops and Challenges},
year = {2026},
publisher = {Springer Nature Switzerland},
volume = {LNCS 17269},
month = {pending},
page = {pending}
}