119 posters, 6 topics, 524 authors, 243 institutions
ePostersLive by SciGen Technologies S.A. All rights reserved.
RCR Global AI Conference 2026
29-30 June, 2026 | QEII Centre, Westminster
80
<p>Large Language Models to Improve Understanding of Radiology Reports -</p> <p>A Systematic Review and Meta-Analysis of Patient, Public, and Clinician Evaluations</p>
Part of Topic
Transforming practice and leadership: pilot or test data on implementation of AI into clinical practice, clinical feedback or patient perspectives
Large Language Models to Improve Understanding of Radiology Reports
A Systematic Review and Meta-Analysis of Patient, Public, and Clinician Evaluations
Background
Radiology reports are typically written for clinicians and are often difficult for patients to understand. Large language models (LLMs) can simplify complex medical language and may improve patient comprehension of imaging reports.
Methods
•Systematic review and meta-analysis of studies evaluating LLM-simplified radiology reports.
•Systematic review and meta-analysis of studies evaluating LLM-simplified radiology reports.
•MEDLINE, Embase, CENTRAL, medRxiv, bioRxiv, and arXiv were searched in Nov 2025.
•Outcomes included patient/public understanding, clinician-rated quality, readability metrics, and error rates.
Results
•38 studies including 12,922 simplified reports and 508 evaluators were included.
•Patients rated LLM-simplified reports as substantially more understandable than original radiology reports (Likert score 4.04 vs 2.16).
•Clinicians rated reports highly for accuracy (4.45/5) and completeness (4.53/5).
•Readability improved markedly across CT, MRI, and x-ray reports, reducing reading level from university to school-age level.
•Error rate was 7%, with clinically significant errors occurring in 0.9% of reports.
•No study included patient preferences, was UK-based, or used advanced prompting techniques.
Conclusion
•LLM-simplified radiology reports improve patient-perceived understanding while maintaining high clinician-rated accuracy.
•Clinician oversight remains important to minimise clinically significant errors.
•Future studies should be patient co-designed, NHS-based and safety-engineered.