Radiology Report Summarization

Radiology report summarization (RRS) focuses on automatically generating concise and accurate summaries of complex radiology reports, primarily targeting the "Impression" section, to improve communication and efficiency in healthcare. Current research heavily utilizes large language models (LLMs), often adapting pre-trained models like BART or BERT through techniques such as fine-tuning, prompt engineering (including multi-modal approaches incorporating radiographic images), and contrastive learning to enhance accuracy and faithfulness to the original report. These advancements aim to reduce radiologists' workload, improve the consistency of reporting, and facilitate better patient care by providing more accessible summaries for both clinicians and patients.