Background and aims: The diagnosis of transthyretin amyloid cardiomyopathy (ATTR-CM) requires advanced imaging, precluding large-scale testing for pre-clinical disease. We examined an application of artificial intelligence (AI) to transthoracic echocardiography (TTE) and electrocardiography (ECG) as a scalable risk stratification strategy for pre-clinical ATTR-CM.
Methods: In 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 studies of individuals referred for nuclear cardiac amyloid testing in an independent population at YNHHS and an external population from Houston Methodist Hospitals (HMH). We evaluated longitudinal trends in AI-defined probabilities of ATTR-CM using age/sex-adjusted linear mixed models and their ability to stratify the risk of ATTR-CM across pre-clinical stages.
Results: Among 984 participants 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 cohorts and modalities, AI-defined ATTR-CM probabilities derived from 7,423 TTEs and 32,205 ECGs diverged as early as 3 years before clinical diagnosis in cases versus controls (p time(x)group interaction≤0.004). One-to-three years before referral for ATTR-CM testing, a double-negative screen (AI-Echo(-)/AI-ECG(-)) had sensitivity of 0.98 (95%CI:0.96-0.99) and 0.89 (95%CI:0.86-0.92), whereas a double-positive screen (AI-Echo(+)/AI-ECG(+)) yielded specificity of 0.72 (95%CI:0.69-0.74) and 0.91 (95%CI:0.90-0.91) in YNHHS and HMH, respectively.
Conclusions: AI applied to echocardiographic videos and ECG images may enable scalable risk stratification of ATTR-CM during its early pre-clinical course.
Keywords: artificial intelligence; cardiac amyloidosis; echocardiography; electrocardiography; screening; transthyretin.