AI-Enabled Echocardiography Identifies Prominent Epicardial Adipose Tissue Associated With an Adverse Cardiometabolic Profile

Background: Excess epicardial adipose tissue (EAT) is associated with cardiovascular-kidney-metabolic (CKM) dysfunction, but its assessment has traditionally required advanced imaging. Objectives: We investigated whether AI-enabled echocardiography could characterize epicardial adiposity and identify residual cardiometabolic risk beyond body mass index (BMI). Methods: We developed PanAdipo, a video-based deep learning model in 1,114,441 videos from 28,797 studies across the Yale-New Haven Health System (YNHHS; 2016-2022), using report-derived expert reader labels of prominent EAT (2.3% of studies). PanAdipo was evaluated in five cohorts: a temporally distinct YNHHS transthoracic echocardiography (TTE) set (n=4,588 patients contributing 5,130 studies), an emergency department point-of-care ultrasound (POCUS) cohort (n=10,957), the geographically distinct MIMIC-IV dataset (n=4,549), the community-based Multi-Ethnic Study of Atherosclerosis (MESA Exam 6, n=2,740), and in held-out YNHHS patients with paired coronary CT angiography within 12 months (n=5,167). Analyses examined discrimination of the report label and of CT-defined high indexed EAT volume, correlation with paired cardiac CT phenotypes, and BMI-adjusted associations with cardiometabolic biomarkers and newly documented metabolic disease. Results: In the held-out health system test set, PanAdipo identified prominent EAT documented in clinical reports with an area under the receiver operating characteristic curve (AUROC) of 0.86 (95% CI, 0.82-0.91) across all views and of up to 0.93 when combining apical 4-chamber and parasternal views. Against paired CT-derived measurements, PanAdipo discriminated age- and sex-specific high EAT (EAT/heart volume [≥]95th percentile) with AUROCs of 0.91 to 0.95 across age-sex groups and correlated more strongly with indexed EAT volume ({rho}=0.76) than with BMI ({rho}=0.41, all...