Large language models (LLMs) in radiography research: A narrative review.

Area of Science:

• Radiology and Medical Imaging
• Artificial Intelligence in Healthcare
• Computational Science

Background:

• Artificial intelligence (AI) is increasingly integrated into radiography research and practice.
• Generative AI (GenAI), especially Large Language Models (LLMs), has emerging applications in radiography.
• These applications extend beyond clinical support to research methodologies.

Purpose of the Study:

• To explore the opportunities and challenges of integrating LLMs in radiography research.
• To review the current landscape of LLM applications in the field.
• To discuss the implications for future radiography research and practice.

Main Methods:

• A narrative review approach was utilized.
• Relevant publications were identified and synthesized.
• Author-led examples and case studies were integrated.

Main Results:

• LLMs accelerate research by aiding literature retrieval, survey development, synthetic data generation, and data analysis.
• Case studies show benefits like improved CT image quality and reduced examination times.
• Challenges include output inaccuracies (hallucinations), data bias, privacy concerns, regulatory hurdles, and environmental costs.

Conclusions:

• GenAI and LLMs present transformative potential for radiography research.
• Integration requires rigorous validation against expert data, transparent reporting, and robust ethical frameworks.
• A 'trust but verify' approach, coupled with training and governance, is crucial for safe and effective implementation.