Tracking the pre-clinical progression of transthyretin amyloid cardiomyopathy using artificial intelligence-enabled electrocardiography and echocardiography.

Abstract

Diagnosing transthyretin amyloid cardiomyopathy (ATTR-CM) requires advanced imaging, precluding large-scale testing for pre-clinical disease. We examined the application of artificial intelligence (AI) to echocardiography (TTE) and electrocardiography (ECG) as a scalable strategy to quantify pre-clinical trends in ATTR-CM. Across age/sex-matched case-control datasets in the Yale-New Haven Health System (YNHHS) we trained deep learning models to identify ATTR-CM-specific signatures on TTE videos and ECG images (area under the curve of 0.93 and 0.91, respectively). We deployed these across all studies of individuals referred for cardiac nuclear amyloid imaging in an independent population at YNHHS and an external population from the Houston Methodist Hospitals (HMH) to define longitudinal trends in AI-defined probabilities for ATTR-CM using age/sex-adjusted linear mixed models, and describe discrimination metrics during the early pre-clinical stage. Among 984 participants referred for cardiac nuclear amyloid imaging at YNHHS (median age 74 years, 44.3% female) and 806 at HMH (69 years, 34.5% female), 112 (11.4%) and 174 (21.6%) tested positive for ATTR-CM, respectively. Across both cohorts and modalities, AI-defined ATTR-CM probabilities derived from 7,423 TTEs and 32,205 ECGs showed significantly faster progression rates in the years before clinical diagnosis in cases versus controls (p time × group interaction ≤0.004). In the one-to-three-year window before cardiac nuclear amyloid imaging sensitivity/specificity metrics were estimated at 86.2%/44.2% [YNHHS] vs 65.7%/65.5% [HMH] for AI-Echo, and 89.8%/40.6% [YNHHS] vs 88.5%/35.1% [HMH] for AI-ECG. We demonstrate that AI tools for echocardiographic videos and ECG images can enable scalable identification of pre-clinical ATTR-CM, flagging individuals who may benefit from risk-modifying therapies.