Exploring the Potential of Retrieval Augmented Generation for Question Answering in Radiology: Initial Findings and...

Area of Science:

• Medical Imaging and Artificial Intelligence
• Clinical Decision Support Systems

Background:

• Intelligent systems offer potential to enhance clinical decision-making in radiology.
• Retrieval-Augmented Generation (RAG) is an emerging technique for question answering.

Purpose of the Study:

• To explore the application of RAG for question answering in radiology.
• To evaluate a naive RAG setup using radiology-specific questions and a textbook.

Main Methods:

• A preliminary experiment was conducted using a naive RAG system.
• The system was tested on radiology-specific questions with a textbook as the knowledge source.

Main Results:

• The RAG system demonstrated moderate improvements compared to baseline methods.
• The study identified initial findings on RAG performance in a specialized domain.

Conclusions:

• RAG shows promise for radiology question answering, but requires further development.
• Future work should focus on enhancing RAG for handling complex medical knowledge and integration into clinical practice.