In this post
A new JACC Advances study compared four approaches for detecting transthyretin cardiac amyloidosis in a real-world heart failure population, finding that deep learning echocardiography models outperformed traditional risk-score and claims-based approaches, with EchoGo Amyloidosis achieving the best overall diagnostic performance.
Jonathan Hourmozdi, Nicholas Easton, Simon Benigeri, James D. Thomas, Akhil Narang, David Ouyang, Grant Duffy, Ross Upton, Will Hawkes, Ashley Akerman, Ike Okwuosa, Adrienne Kline, Abel N. Kho, Yuan Luo, Sanjiv J. Shah, and Faraz S. Ahmad.
Key findings
- EchoGo Amyloidosis achieved the highest AUC of all evaluated models at 0.92.
- EchoGo Amyloidosis achieved the highest sensitivity at 85%, meaning fewer patients with disease were missed.
- Deep learning echocardiography models outperformed traditional risk scores and claims-based machine learning approaches.
- EchoGo Amyloidosis met predefined fairness criteria for equal opportunity among Black patients.
- The findings support AI-enabled echocardiography as a scalable approach to improving cardiac amyloidosis detection.
EchoGo Amyloidosis performance
A new independent study published in JACC Advances found that EchoGo Amyloidosis achieved the highest overall performance among four evaluated approaches for detecting ATTR-CM, outperforming both traditional risk scores and claims-based machine learning models in a real-world heart failure population.
The study evaluated 3,368 patients from a large integrated health system, including 176 confirmed cases of ATTR-CM, and directly compared four different screening approaches. EchoGo Amyloidosis achieved an AUC of 0.92, exceeding both the Mayo ATTR-CM Score (0.79) and EchoNet-LVH (0.88), demonstrating strong discrimination for identifying patients with ATTR-CM.
Importantly, EchoGo Amyloidosis also achieved the highest sensitivity of all evaluated models at 85%, meaning fewer patients with disease were missed. This is particularly relevant in ATTR-CM, where diagnostic delays commonly exceed several years and can prevent patients from accessing disease-modifying therapies in a timely manner.
Supporting early identification of ATTR-CM
The findings add to the growing body of evidence supporting AI-enabled echocardiography as a scalable approach to improving identification of cardiac amyloidosis.
Unlike traditional scoring systems that rely on manually collected clinical variables, EchoGo Amyloidosis analyses routinely acquired echocardiographic images and appears to capture disease characteristics beyond conventional measurements.
Performance across diverse patient populations
The investigators also conducted a fairness audit to evaluate potential racial bias. EchoGo Amyloidosis met predefined fairness criteria for equal opportunity among Black patients, a population disproportionately affected by delayed diagnosis of ATTR-CM.
The authors concluded that deep learning echocardiography models demonstrated the best overall performance while showing a low risk of harm due to racial bias.
About the study
The retrospective study included 3,368 patients from a large integrated health system, including 176 confirmed cases of ATTR-CM and 3,192 matched heart failure controls.
The authors concluded that deep learning, echo-based models demonstrated the strongest overall discrimination for ATTR-CM detection and may help support earlier identification of patients who could benefit from further diagnostic evaluation and treatment.
Publication
Hourmozdi J, Easton N, Benigeri S, et al. Evaluating the Performance and Potential Bias of Predictive Models for Detection of Transthyretin Cardiac Amyloidosis. JACC Advances. 2025;4(8):101901. 