This study explores the effectiveness of Large Language Models (LLMs) in assisting general cardiologists with diagnosing rare, life-threatening cardiac diseases, which often require specialized care. We curated a de-identified clinical dataset focused on suspected inherited cardiomyopathies and evaluated LLM-assisted assessments against those performed by cardiologists without support. The findings demonstrated that LLMs, particularly the AMIE system, significantly reduced clinically significant errors by 11.2% and missed important content by 19.6%, while preserving clinical reasoning quality. General cardiologists using AMIE reported improved assessment efficiency and accuracy, with 57% stating the system enhanced their evaluations. This evidence highlights LLMs’ potential to bridge the gap in cardiology, particularly amid workforce shortages. The study’s implications extend to better triage and management of complex patients, though caution is advised regarding overreliance on AI. Future research is essential to validate these findings and assess patient outcomes. Open-access datasets and evaluation rubrics are provided to facilitate further exploration.
Source link
