Deep-learning analysis of 3D microarchitectural remodeling in hypertrophic cardiomyopathy

Abstract

Hypertrophic cardiomyopathy (HCM), a genetic heart disease defined by unexplained cardiac wall thickening, is a leading cause of sudden death worldwide. However, the three-dimensional organization of cardiac tissue underlying left ventricular hypertrophy remains poorly understood. We developed CaMVIA-3D, a deep-learning volumetric imaging and analysis pipeline to characterize cardiac microarchitecture. Analysis of tissues from HCM hearts revealed genotype-specific differences in cardiomyocyte volume, morphology, and extracellular volume, with pathogenic variants exhibiting greater concentric cellular hypertrophy and disarray and variant-negative cases showing predominant fibrosis. Longitudinal profiling of a pig HCM model revealed early-onset fibrosis preceding cardiomyocyte hypertrophy. Integrating transcriptomic and morphologic changes, we identified genes associated with cellular and extracellular remodeling. These findings define genotype-specific microstructural differences in HCM, offering insights to improve diagnostics and targeted therapies.