PanEcho was trained on a large dataset of 1.2 million videos from over 32,000 TTE studies involving more than 24,000 patients, collected between 2016 and 2022 from Yale New Haven Health.

As an open-source tool, PanEcho is accessible for researchers to utilize and experiment with in future studies.

The AI system also performs well in diagnosing right ventricular systolic dysfunction with an AUC of 0.94 and maintains strong accuracy in abbreviated and emergency scans, with median AUCs of 0.91 and 0.85 respectively.

Researchers from the University of Texas at Austin and Yale have developed PanEcho, an AI tool designed for comprehensive heart analysis through echocardiography, published on June 23.

Gregory Holste, a doctoral candidate and author of the study, explained that the model predicts 39 key measurements and labels typically included in a complete echocardiographic report.

The automation offered by PanEcho could streamline workflows, assist in diagnostics, and expand access to cardiac imaging, especially in resource-limited settings, though further clinical evaluation is needed before widespread use.

Trained on nearly one million echocardiographic videos, PanEcho analyzes ultrasound images of the heart to perform its assessments.

Validation results showed a median AUC of 0.91 across 18 diagnostic tasks, with strong performance in measurements such as ejection fraction estimation, which had a median normalized mean absolute error of 0.13.

Researchers are also exploring the adaptation of PanEcho for pediatric scans, highlighting its potential for rapid fine-tuning across different domains.

Unlike traditional human analysis limited to a single view, PanEcho synthesizes information from multiple views for a more comprehensive heart assessment.

The system can perform 39 echocardiographic tasks and accurately detect conditions like systolic dysfunction and severe aortic stenosis.

One of PanEcho’s key features is its ability to measure left ventricular ejection fraction more accurately than human experts.

PanEcho can fully automate the interpretation of transthoracic echocardiograms (TTEs), matching cardiologists in 39 diagnostic and measurement tasks with high accuracy.

In validation tests, PanEcho demonstrated exceptional diagnostic performance, with an AUC of 0.99 for detecting moderate or worse left ventricular systolic dysfunction and a perfect score of 1.00 for severe aortic stenosis.