Large language models in interventional cardiology: current evidence and future directions

Area of Science:

• Artificial Intelligence in Medicine
• Cardiovascular Medicine
• Medical Informatics

Background:

• Large language models (LLMs) are advanced AI systems processing extensive clinical and scientific texts.
• Their application in general cardiology is known, but their specific role in interventional cardiology and catheterization laboratories (cath labs) is less defined.

Purpose of the Study:

• To review current evidence on LLM applications within the interventional cardiology workflow.
• To assess LLM capabilities in decision-making, multidisciplinary team support, and acute cath lab scenarios.

Main Methods:

• Narrative review synthesizing existing studies on LLM applications in interventional cardiology.
• Analysis of LLM performance in simulated decision-making tasks compared to expert recommendations and early-career physicians.

Main Results:

• LLMs demonstrate clinically meaningful concordance with expert recommendations for percutaneous coronary intervention vs. coronary artery bypass grafting.
• LLM outputs in simulated emergencies approach or match early-career interventional cardiologists' performance.
• Identified limitations include variable accuracy, hallucinations, lack of multimodal integration, and unresolved medicolegal issues.

Conclusions:

• LLMs show potential to support interventional cardiologists in decision-making, education, and communication.
• Future directions include multimodal LLMs and retrieval-augmented generation tools.
• Regulatory and ethical considerations are crucial for clinical adoption.