Federated AI screening for transthyretin amyloid cardiomyopathy

Overview

The TRACE-AI Network is a multi-site AI screening platform designed to address the significant underdiagnosis of transthyretin amyloid cardiomyopathy (ATTR-CM), a progressive and often misdiagnosed heart condition. By utilizing deep learning algorithms applied to routine cardiovascular diagnostic data, such as electrocardiograms (ECG), point-of-care ultrasound (POCUS), and echocardiograms, the network provides a scalable method to identify high-risk individuals earlier in the disease course. The solution enables improved patient outcomes through targeted referrals for confirmatory testing and earlier initiation of disease-modifying therapies.

Technical specifications

• Federated Learning Deployment: The network utilizes a secure, federated (non-data-sharing) model, allowing participating health systems to run diagnostic algorithms locally while retaining ownership of their patient data.
• Multimodal AI Tools: The platform integrates validated deep learning models, including ECG-Vision ATTR and AI-ECHO ATTR, which analyze standard clinical images to detect subtle patterns indicative of cardiac amyloidosis.
• Scalable Screening: Designed to function across large, diverse U.S. health systems, the system processes existing electronic health record data, ensuring broad applicability across socioeconomic and demographic groups.
• Research Integration: The platform supports ongoing observational studies and prospective clinical trials to validate performance against gold-standard technetium-99m pyrophosphate (PYP) imaging.

Technology readiness level

7. The technology is currently in active prospective clinical evaluation (TRACE-AI Yale, NCT07398950) and is being deployed across multiple U.S. health systems for large-scale observational research. Prototype models are publicly available for research use, and performance has been validated across multinational cohorts.