Advancements in Radiology Report Generation: A Comprehensive Analysis.

Area of Science:

• Medical Imaging and Radiology
• Artificial Intelligence in Healthcare
• Natural Language Processing

Background:

• Increasing demand for radiological services and radiologist shortages create workload challenges.
• Ensuring accuracy and timeliness of radiological reports is critical for clinical decision-making.
• Artificial intelligence (AI) presents potential solutions for radiology report generation (RRG).

Purpose of the Study:

• To provide a comprehensive overview of AI-driven RRG developments from 2021 to 2025.
• To focus on emerging transformer-based and vision-language models (VLMs) in RRG.
• To analyze datasets, evaluation metrics, and leading model performance in RRG.

Main Methods:

• Review of transformer models, VLMs, and Large Language Models (LLMs) for RRG.
• Examination of datasets and evaluation metrics for RRG applications.
• Analysis of leading AI model performance, strengths, and limitations in RRG.

Main Results:

• AI, particularly transformer and VLMs, shows promise in automating and improving RRG.
• Key methods, architectures, and techniques in recent RRG advancements are highlighted.
• Leading models demonstrate varying performance, with identified strengths and weaknesses.

Conclusions:

• AI-powered RRG systems can enhance diagnostic speed and reduce radiologist workload.
• Further research is needed to improve existing AI systems and explore new avenues in RRG.
• Advancing AI capabilities in RRG can lead to better clinical decision-making and patient outcomes.