Large Language Models in Cardiology: Systematic Review

Area of Science:

• Artificial Intelligence in Medicine
• Cardiology Applications
• Natural Language Processing

Background:

• Large language models (LLMs) are increasingly adopted in healthcare.
• Their specific utility and limitations in cardiology remain unevaluated.

Purpose of the Study:

• To systematically review the applications and performance of LLMs in cardiology.
• To assess LLM capabilities across various cardiology tasks, including chronic conditions, acute events, education, and diagnostics.

Main Methods:

• A systematic literature search was performed in PubMed and Scopus.
• Studies evaluating LLM performance in cardiology tasks were included and assessed for risk of bias.

Main Results:

• LLMs demonstrated high accuracy in heart failure (91%) and ECG interpretation (91%), outperforming physicians in the latter.
• Performance varied, with issues in emergency guidance accuracy and readability across different LLM versions and tasks.
• ChatGPT-4 showed improved accuracy over ChatGPT-3.5 in physician education tasks.

Conclusions:

• LLMs offer significant potential in cardiology, particularly for ECG interpretation and medical education.
• Current limitations in emergency scenarios and readability require attention.
• Further research with multimodal models and prospective validation is essential.