Nature Communications, Published online: 29 June 2026; doi:10.1038/s41467-026-74575-y In this work, authors used unsupervised 3D diffusion autoencoders on 71k cardiac MRIs to learn latent heart representations predictive of cardiometabolic disease, identify 89 GWAS variants across 42 loci, and derive polygenic risk scores that stratify cardiovascular risk.
Hundreds of cardiac MRI traits derived using 3D diffusion autoencoders share a common genetic architecture
Craig A. Glastonbury