Related Experiment Video
Updated: May 30, 2025
07:47
Author Spotlight: 3D Scanning and Augmented Reality for Enhanced Cancer Surgery Communication
Published on: December 15, 2023
601
Provision of Radiology Reports Simplified With Large Language Models to Patients With Cancer: Impact on Patient
Amit Gupta1, Swarndeep Singh2, Hema Malhotra3
1Department of Radiodiagnosis, All India Institute of Medical Sciences, New Delhi, India.
JCO Clinical Cancer Informatics
|January 29, 2025
Summary
Simplified radiology reports, generated using large language models (LLMs), significantly improve oncology patients' understanding of their disease. LLMs show promise for creating these patient-friendly reports with necessary human oversight.
Area of Science:
- Radiology
- Medical Informatics
- Artificial Intelligence
Background:
- Radiology reports can be complex for patients.
- Improved patient comprehension is crucial for informed decision-making in oncology.
Purpose of the Study:
- To assess the utility of simplified radiology reports for oncology patients.
- To evaluate the feasibility of using large language models (LLMs) for report simplification.
Main Methods:
- Five LLMs were tested for simplifying 50 oncology CT report impressions.
- Readability indices and qualitative assessments selected the best LLM-prompt combination.
- 100 oncology patients received either original or simplified reports to assess knowledge and utility.
Main Results:
- Claude Opus-Prompt 3 showed slight superiority in simplification, with other top LLMs performing comparably.
- Patients receiving simplified reports demonstrated significantly better knowledge and confidence.
- Radiologist oversight was minimal, with only 3 of 50 simplified reports requiring corrections.
Conclusions:
- Simplified radiology reports substantially enhance patient understanding and confidence.
- LLMs are effective tools for generating simplified radiology reports.
- Human oversight remains essential for ensuring accuracy in LLM-generated reports.
