AI Echo Tool Boosts Cardiac Amyloid Detection

Jeremy Slivnick, MD, Assistant Professor of Medicine at The University of Chicago Medical Center, discussed the promise of an artificial intelligence (AI) model designed to automatically analyze echocardiograms for signs of cardiac amyloid, particularly in older patients with heart failure and increased wall thickness. He explained that such a tool could generate real-time predictions, enabling earlier diagnosis and faster referral for treatment, especially in community or rural settings with limited expertise or awareness. Dr. Slivnick emphasized the importance of validating the model across a diverse, global dataset to ensure broad applicability, while also noting key barriers to adoption, including image confidentiality, institutional logistics, and unclear reimbursement pathways. Overall, he expressed optimism that AI could help “democratize” access to diagnosis and therapy, potentially improving survival for patients with cardiac amyloid by pairing enhanced detection with timely treatment.

Transcript

Cardio Care Today: How do you see AI-based tools like this echocardiogram model transforming the diagnostic pathway for cardiac amyloidosis, especially in primary and secondary care settings?

Jeremy Slivnick: Yes. I think that the advantage of this tool is that it can be automatically deployed at health institutions. Our hope would be that patients that sort of meet those initial risk features—older patients with heart failure, increased wall thickness—that this model could be run automatically on their echoes and basically generate a real-time prediction [based] on the presence of amyloid. Patients that are considered positive by this model could then be streamlined for downstream testing and diagnosis and therapy. We hope that the deployment of a model like this could help to broaden the detection of it—particularly at health centers that have less capability or in community or rural settings where there’s less awareness for cardiac amyloidosis—that this tool could help to democratize the diagnosis and access to therapies.

From a practical standpoint, what do you foresee as the main challenges and facilitators to integrating this FDA-cleared AI tool into routine echocardiography workflows?

There has been an influx of many AI tools that have shown the potential to improve clinical practice in echocardiography, but implementation is the biggest barrier. For one, there are medical images and the potential for loss of confidentiality. There are a lot of logistics that need to be approved at every institution to ensure that the image sharing is safe and that the images remain de-identified. That’s part of it. Additionally, the reimbursement model is something that needs to be clarified too. Typically, the way that these have been reimbursed is through patient diagnosis. So when you have a patient who’s detected by the algorithm, there’s payment from the insurance companies to pick up these patients. But that’s something that … [is] a new area and something that we need to see how it works as it gets deployed in clinical practice.

The study reported high accuracy across a diverse, multiethnic population—how important is this generalizability for clinical adoption, and where do you think further validation is needed?

It’s critical. When we deploy these models, we deploy them potentially across the world. One of the concerns is that if you only validate a study like this in a specific population or a specific demographic group, you don’t know whether it can be generalized to the entire population. So we felt that was important, and that’s why we created the multi-ethnic dataset utilizing a global population that had a high population of different sociodemographic groups to ensure that this model could be deployed. Now it’s important to understand this is a retrospective study, and we still need to understand how this works in clinical practice in the real world, particularly as our understanding of amyloidosis evolves and patients are getting diagnosed earlier.

With emerging treatments offering significant survival benefits, how do you envision the combination of AI screening and early therapeutic intervention changing the long-term prognosis for patients with cardiac amyloidosis?

We’re already seeing that this combination of improved awareness and access to therapies has improved survival significantly for patients with cardiac amyloidosis. AI has the potential to really supercharge that kind of symbiotic relationship between diagnosis and therapy. We can augment the detection, particularly in areas where there’s less access to diagnosis. So, potentially in rural or in community settings, this could be leveraged to really broaden our detection of this disease and then streamline them [patients] to places where they can receive treatment.