LLM based summarization of radiology reports based on readers' knowledge level

What is it about?

Radiology reports, even though readily available to patients, are hard to read to read and understand by laymen patients if they do not have any background clinical knowledge. While significant research effort has been directed towards automatic generation of reports, relatively little attention has been paid to building automated tools to explain these reports to patients with varying level of clinical knowledge and understanding. We built an end-to-end joint framework for summarizing radiology reports based on readers' level of clinical knowledge (expert-level and laymen level) as described in the input prompt. Our framework is built through task-specific finetuning of popular LLM.

Featured Image

Why is it important?

This work is important as the proposed framework can aid patient education by making radiology reports comprehensible to laymen patients. Inability to understand clinical information leads to missed follow-up opportunities and ultimately poor clinical outcome. The proposed framework can alleviate the barrier of education and knowledge level for patients when it comes to understanding their clinical status described in radiology reports, potentially improving follow-up and clinical outcomes.