Study Shows AI Can Improve Early Detection of Cardiac Amyloidosis

A new Ultromics study on artificial intelligence (AI) in echocardiography is revealing the technology’s increasing role in early detection of cardiac amyloidosis.

Once diagnosed only after years of unexplained heart failure symptoms, cardiac amyloidosis is now at the center of cardiology. With its ability to spot the disease on routine heart ultrasounds, AI certainly could support earlier intervention in the disease course, when treatment may offer greater benefit.

Drawing on 4,815 patient cases from 17 hospitals in the United States and United Kingdom, Ultromics modelled the ways in which EchoGo Amyloidosis could improve referral decisions in real-world practice. AI detected cardiac amyloidosis earlier and more accurately than traditional methods, finding patients who would otherwise have been missed while reducing unnecessary testing. Published as an abstract in the Journal of the American Society of Echocardiography (JASE), the results held true across both low- and high-prevalence settings, showing the potential impact of AI in everyday clinical practice. Major findings included:

• In low-prevalence scenarios, referral decisions based on wall thickness alone correctly identified ~ 65% of patients with cardiac amyloidosis. Incorporating AI increased correct referral rates to ~ 76% to 80%, meaning more patients could be identified earlier while avoiding unnecessary referrals.¹
• In higher-prevalence scenarios, AI could reduce unnecessary referrals by up to 18% while maintaining high detection rates.¹
• The findings were consistent across hospitals in both the United States and United Kingdom, underscoring the technology’s potential for broad clinical use.¹

Cardiac amyloidosis is increasingly recognized as a common driver of heart failure. Newly available therapies such as tafamidis and acoramidis can slow disease progression and reduce mortality, but they are effective only when patients are identified early. Up to 66% of cases are undiagnosed in clinical practice.^2-4 

“Too often, patients with cardiac amyloidosis are diagnosed only after years of unexplained symptoms and irreversible damage,” Ultromics Clinical Sciences Director and lead study author Dr. Ashley Akerman said. “Our findings suggest that using EchoGo Amyloidosis to  enhance routine heart scans, doctors could better identify at-risk patients, reduce unnecessary testing, and ensure those who need confirmatory diagnosis and treatment, receive it sooner.”

EchoGo Amyloidosis is designed to help close this diagnostic gap by analyzing echocardiograms at the pixel level to detect subtle patterns often missed by the human eye. Trained and validated on 7,174 patients (9,700-plus echo videos) from 15 international sites, and tested on more than 2,700 additional patients across 18 sites, the model achieved high accuracy (AUC 0.93) across multi-ethnic, real-world populations. Its cardiac amyloidosis model provides consistent, automated assessments that help clinicians identify at-risk patients sooner, improve referral decisions for confirmatory testing and connect more patients to life-prolonging therapies.⁵

This study adds to the growing clinical validation of Ultromics’ EchoGo platform, the first U.S. Food and Drug Administration (FDA)-cleared and Medicare-reimbursed AI system for echocardiography. With results documented in more than 25 peer-reviewed studies, EchoGo is used at such U.S. hospitals as UChicago Medicine, Northwestern, and City of Hope, where it supports earlier detection of complex cardiovascular conditions and more precise patient management.

Founded out of the University of Oxford, Ultromics is redefining cardiovascular care with FDA-cleared, AI-powered tools that enhance echocardiographic diagnosis. The company is backed by leading investors and U.S. healthcare systems and aims to transform the way heart disease is diagnosed and treated.

References
¹ Akerman AP, et al. J Am Soc Echocardiogr. 2025;38(9S):Axxx.
² González-López E, et al. Eur Heart J. 2015;36:2585–94.
³ Hahn VS, et al. JACC Heart Fail. 2020;8:712–24.
⁴ AbouEzzeddine OF, et al. JAMA Cardiol. 2021;6:1267–74.
⁵ Slivnick JA, Hawkes W, et al. Eur Heart J. 2025;ehaf387.