Generative AI Model Can Help Explain Echocardiogram Reports

THURSDAY, Aug. 8, 2024 (HealthDay News) — Generative artificial intelligence (AI) models can help clinicians explain echocardiogram results to patients, according to a research letter published online July 31 in JACC: Cardiovascular Imaging.

Jacob A. Martin, M.D., from the NYU Grossman School of Medicine in New York City, and colleagues examined the performance of ChatGPT, a generative AI model, for creating patient-oriented echocardiogram reports from 100 patients at an academic health system. AI-generated explanations were assessed by five echocardiographers for acceptance, accuracy, relevance, understandability, and representation of quantitative information.

The researchers found that the echocardiographers agreed or strongly agreed that 73 percent of the AI-generated explanations could be sent to patients without modifications. All explanations were rated as all true or mostly correct (84 and 16 percent, respectively). In terms of relevance, 76 percent of explanations were categorized as containing all of the important information, while 15, 7, and 2 percent contained most, about half, and less than half, respectively. Of the explanations with incorrect or missing information, none were rated as potentially dangerous.

“Most AI-generated explanations were accurate, relevant, and readily communicable to patients, highlighting the potential of AI as a helpful tool, while human oversight remains essential for refinement,” the authors write.

Two authors disclosed ties to the biopharmaceutical and medical device industries.

Abstract/Full Text (subscription or payment may be required)