Abstract
Coronary artery motion observed in X-ray angiography reflects complex physiological processes such as myocardial contraction, vessel compliance, and hemodynamic forces, yet it remains underused as a quantitative signal of cardiac dynamics because extracting it at scale from routine clinical data is technically demanding. This paper presents a scalable, segmentation-free, data-driven deep learning framework for large-scale analysis of coronary artery motion based on dense optical flow. Motion is extracted directly from routine angiographic sequences using Frangi-based vessel enhancement and deep optical flow estimation. Lacking ground-truth motion annotations, we adopt a dual validation strategy: synthetic data with exactly known flow, and real sequences assessed through reconstruction-based metrics and qualitative visualization. The extracted flow fields drive a video-based deep neural networkthat learns motion representations, evaluated for association with clinical risk factors using patient level stratified train/validation splitting to prevent data leakage. From a cohort of 1,824 patients, projection angle filtering and clinical label availability yielded analysis cohorts ranging from 124–316 patients (288–757 sequences) per parameter. Learned motion-derived representations predicted several cardiovascular risk factors above majority-class baselines, including ejection fraction, cholesterol level, hypertension, and nicotine use. These results position coronary artery motion from routine angiography as a scalable, clinically relevant signal warranting further investigation as a cardiovascular risk association tool.
Links to Paper and Supplementary Materials
Main Paper (Open Access Version): https://papers.miccai.org/miccai-2026-sat/paper/AMAI_068.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=%7EYasamin_Moghbelan1
BibTex
@InProceedings{MogYas_Association_MICCAISAT2026,
author = { Moghbelan, Yasamin AND Breucker, Daniel AND von Scheidt, Moritz AND Hayden, Oliver AND Bescos, Javier Olivan AND Ruijters, Danny},
title = { { Association of Cardiovascular Risk Factors with Segmentation-Free Motion-Derived Coronary Artery Dynamics } },
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}
}