Abstract
Accurate automatic myocardial pathology segmentation in multi-sequence cardiac magnetic resonance remains a challenging problem due to the high domain shift between different sequences, often requiring alignment and joint processing, the small and irregular shape of pathological regions, and occasional missing sequences. Current proposed strategies mainly aim to segment pathology directly from multi-sequence data or use a two-stage process in which the myocardium is segmented first to guide pathology detection. While effective, these strategies may still suffer from domain shift and often depend on complete multi-sequence data. To address these challenges, we propose a three-stage segmentation pipeline that combines causality-inspired data augmentation strategies with polar transformation for pathology segmentation, leveraging the common circular shape of the myocardium to improve performance. We train domain-invariant models that achieve consistent performance across all tested domains, yielding accurate ventricles segmentation even on unseen sequences. We evaluate our approach on the CARE 2026 MyoPS dataset, achieving Dice values of 0.72 for both scar and edema regions, surpassing standard state-of-the-art Cartesian-based baselines. These results highlight the potential of employing polar transformation strategies for myocardial pathology segmentation.
Links to Paper and Supplementary Materials
Main Paper (Open Access Version): https://papers.miccai.org/miccai-2026-sat/paper/CARE_046.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=HtqSxdEKxV
BibTex
@InProceedings{RibMat_Causalityinspired_MICCAISAT2026,
author = { Ribeiro, Matheus A. O. AND Punithakumar, Kumaradevan AND Nunes, Fátima L. S.},
title = { { Causality-inspired Polar Approach for Myocardial Pathology Segmentation in Multi-sequence CMR } },
booktitle = {Medical Image Computing and Computer Assisted Intervention -- MICCAI 2026 Workshops and Challenges},
year = {2026},
publisher = {Springer Nature Switzerland},
volume = {LNCS 17273},
month = {pending},
page = {pending}
}