Cardiovascular measures from abdominal MRI provide insights into abdominal vessel genetic architecture

Cardiovascular disease remains a major source of morbidity and mortality, and population imaging studies have yielded insights into disease etiology and risk. In this study, we segment the heart, aorta, and vena cava from abdominal magnetic resonance imaging (MRI) scans using deep learning. We generate six image-derived phenotypes (IDP): heart volume, four aortic and one vena cava cross-sectional areas (CSA), from 44,541 UK Biobank participants, and explore their associations with disease outcomes, as well as genetic and environmental factors. Here we show concordance between our IDPs and related IDPs from cardiac magnetic resonance imaging, the current gold standard, and replicate previous findings related to sex differences and age-related changes in heart and vessel dimensions. We identify a significant association between infrarenal descending aorta CSA and incident abdominal aortic aneurysm, and between heart volume and several cardiovascular disorders. In a genome-wide association study, we identify 72 associations at 59 loci (15 novel). We derive a polygenic risk score for each trait and demonstrated an association with thoracic aneurysm diagnosis, pointing to a potential screening opportunity. We demonstrate substantial genetic correlation with cardiovascular traits including aneurysms, varicose veins, dysrhythmia, and cardiac failure. Finally, heritability enrichment analysis implicates vascular tissue in the heritability of these traits. This study illustrates the value of non-specific abdominal MRI for exploring cardiovascular disease risk in cohort studies, and identifies novel genetic associations with vascular dimensions. Heart disease is a major health problem. This study explored whether whole-body MRI scans, medical scans that use magnetic fields and radio waves to create detailed images of the inside of the body, could provide useful insights into heart and blood vessel health. Currently MRI is not typically used for heart analysis, By analyzing data from over 44,000 people, we found links between heart size, blood vessel health, and disease risk. We confirmed our results with other types of heart scan and identified inherited (genetic) factors influencing heart and vessel size. Larger aortas were linked to aneurysms (swelling of the artery wall), and heart volume was connected to heart disease. We also developed a genetic risk score that could help predict certain conditions. Our study shows how artificial intelligence (AI) and MRI scans can improve heart disease research and possibly be used to identify and hence prevent heart disease. Basty and Sorokin et al. use deep learning to analyze abdominal MRI from 44,541 UK Biobank participants. They define six cardiovascular traits linked to aneurysm and broader cardiovascular risk, with genome-wide association studies revealing 72 associations at 59 loci.