Cardiac amyloidosis (CA) is an increasingly recognized cause of heart failure, but delayed diagnosis remains common despite the availability of disease-modifying therapies.  In a study published in the Circulation: Cardiovascular Imaging, investigators developed and evaluated a multiparametric AI echo-clinical model (AI-ECM) that integrates clinical characteristics, laboratory biomarkers, and transthoracic echocardiography (TTE) parameters with an existing validated TTE-based deep learning model, Us2.Ca.

The study utilized data from the Amyloidosis Imaging International Consortium, a multinational registry comprising 727 patients with cardiac amyloidosis and 316 control subjects. Controls included individuals evaluated for suspected transthyretin cardiac amyloidosis with negative diagnostic workups and patients with biopsy-confirmed extracardiac light-chain amyloidosis without cardiac involvement. Model performance was assessed using area under the curve (AUC), accuracy, sensitivity, and specificity.

Findings

• The AI-ECM achieved an AUC of 0.94 compared with 0.89 for the TTE-only Us2.Ca model (P=0.006).
• Overall diagnostic accuracy improved from 80% with Us2.Ca to 90% with AI-ECM.
• Sensitivity increased from 76% to 93%, while specificity was 85% with AI-ECM compared with 91% for the TTE-only model.
• The original Us2.Ca model produced indeterminate results in 9% of cases, whereas the AI-ECM successfully classified all patients and improved detection of light-chain cardiac amyloidosis.

The findings suggest that combining echocardiographic imaging with routinely available clinical and laboratory information can enhance AI-based detection of cardiac amyloidosis.