AI Image-Recognition Program Reads Echocardiograms Faster, Cuts Results Wait Time

An echocardiogram is a diagnostic imaging tool that provides valuable insights into heart structure and function, helping doctors to identify and treat various heart conditions. Now, a new study suggests that integrating an artificial intelligence (AI) program to analyze echocardiograms could reduce the time it takes to receive results, ultimately enabling more timely medical interventions.

In this study, researchers at Yale School of Medicine (New Haven, CT, USA) tested the ability of an AI program called PanEcho to interpret echocardiography videos independently. PanEcho expands on earlier AI applications in cardiology, which were typically limited to analyzing single heart views or focusing on specific disease indicators. The team developed a novel AI system capable of providing comprehensive reports on all major findings from any set of echocardiographic videos. To assess PanEcho's diagnostic performance, the researchers used a standard accuracy measure known as the area under the receiver operating characteristic curve (AUC).

The results showed that when PanEcho was evaluated across 18 diagnostic classification tasks, its average score was 0.91. Additionally, the program was tested for its ability to predict continuous echocardiographic parameters, using mean absolute error, which measures the average difference between predicted and actual values. Smaller differences indicate higher accuracy. In 21 tasks, PanEcho achieved a median normalized mean absolute error of 0.13. The AI system also demonstrated precision in quantifying left ventricle dimensions and function. For instance, when estimating the left ventricle ejection fraction, it showed a mean absolute error of 4.4%, 1.3 mm for intraventricular septum thickness, and 1.2 mm for posterior wall thickness. These measurements are essential for accurately evaluating left ventricular health, a key component of overall heart function. The next step for this AI system, according to the researchers, is to validate its performance in real-world clinical settings, which would provide further insights into its potential for everyday patient care.

“PanEcho has the potential to be used in simplified, AI-assisted screening echocardiograms,” said Gregory Holste, M.S.E., a researcher with the Cardiovascular Data Science (CarDS) Lab at the Yale School of Medicine. “In settings where expert readers may not be readily accessible, PanEcho could rapidly rule out abnormalities that would otherwise require urgent referral.”