Abstract Body (Do not enter title and authors here): Background
Early diagnosis of cardiac transthyretin amyloidosis (ATTR) facilitates disease-modifying therapies. EchoNet-LVH is a deep learning model trained on 16062 echocardiograms which quantifies the likelihood of cardiac amyloidosis. Patients with ATTR polyneuropathy are at risk for cardiac involvement, highlighting the systemic nature of ATTR.

Aim:
To evaluate whether the EchoNet-LVH prediction of cardiac amyloidosis identifies patients in an ATTR polyneuropathy cohort who have known amyloid cardiomyopathy.

Hypothesis:
The EchoNet-LVH cardiac amyloidosis score will be associated with amyloid cardiomyopathy, cardiovascular symptoms, and N-terminal pro-B-type natriuretic peptide (NT-proBNP).

Methods
The NEURO-TTRansform trial enrolled patients with hereditary ATTR polyneuropathy, regardless of cardiac involvement. ATTR cardiomyopathy at baseline was defined by a clinical diagnosis of cardiac amyloidosis or IVSd≥13mm in the absence of hypertension. We applied Echonet-LVH to echocardiograms performed at baseline. The association of EchoNet-LVH score with amyloid cardiomyopathy, abnormal cardiovascular symptoms (NYHA functional class II or greater) and NT-proBNP (>125 pg/mL) was estimated by logistic regression, adjusted for age and sex.

Results
EchoNet-LVH score could be calculated in 130 out of 204 patients (64%). Mean age was 55 years, 28% were female, and 52 were defined as having ATTR cardiomyopathy. The EchoNet-LVH categorized 62 patients (48%) as not predicted to have cardiac involvement, 28 (22%) as intermediate risk, and 40 (31%) as positive. Higher EchoNet-LVH category was associated with higher odds of amyloid cardiomyopathy (OR 4.5 [95% CI 2.6-7.67] per level, p<0.001). Among patients with predicted cardiac involvement 85% had cardiomyopathy, compared to 20% in the predicted negative and intermediate categories. Higher score was associated with greater likelihood NYHA functional class II or greater and elevated NT-proBNP.

Conclusion
These data show the feasibility of utilizing the EchoNet-LVH model to identify patients with a higher likelihood of amyloid cardiomyopathy. Automated deep learning tools like this may aid in the early diagnosis of cardiac amyloidosis.