Related Experiment Video
Updated: Sep 11, 2025
05:33
Introduction of an Integrated Pathology Image Management, Artificial Intelligence, and Reporting System
Published on: July 11, 2025
244
Exploring Radiologists' Use of AI Chatbots for Assistance in Image Interpretation: Patterns of Use and Trust
1Department of Radiological Sciences, College of Applied Medical Sciences, King Saud University, P.O.Box 10219, Riyadh, 11433, Kingdom of Saudi Arabia. mohalarifi@ksu.edu.sa.
Journal of Imaging Informatics in Medicine
|August 14, 2025
Summary
Radiologists found AI-generated patient-friendly radiology reports acceptable, particularly for mammography and CT scans. While generally supportive, they noted MRI reports require more clarity and emotional consideration for improved patient understanding.
Area of Science:
- Radiology
- Artificial Intelligence
- Medical Communication
Background:
- Patient comprehension of radiology reports is often limited by complex medical terminology.
- There is a growing need for patient-friendly medical communication to improve health literacy.
- Artificial intelligence (AI) offers potential solutions for simplifying complex medical information.
Purpose of the Study:
- To evaluate radiologists' perceptions of AI-generated, patient-friendly radiology reports.
- To assess the correctness, completeness, terminology complexity, and emotional impact of AI-simplified reports across MRI, CT, and mammogram/ultrasound modalities.
- To determine the influence of structured prompts on the quality of AI-generated reports.
Main Methods:
- Seventy-nine radiologists from Saudi Arabian hospitals reviewed AI-simplified radiology reports (MRI, CT, mammogram/ultrasound) generated by ChatGPT-4 using a detailed prompt.
- Participants completed questionnaires assessing factual correctness, completeness, terminology complexity, and emotional impact.
- Statistical analyses included descriptive statistics, Friedman tests, and Pearson correlations.
Main Results:
- Radiologists rated mammogram reports highest for correctness (4.22) and completeness, followed by CT and MRI.
- Statistically significant differences were observed in correctness (p<0.001) and completeness (p=0.001) across modalities.
- Moderate anxiety and complexity ratings were noted, with MRI reports eliciting slightly higher concern; a weak positive correlation (r=0.235, p=0.037) existed between radiologist experience and mammogram correctness.
Conclusions:
- Radiologists generally support AI-generated simplified radiology reports when structured prompts ensure clarity, summaries, and glossaries.
- Mammography and CT reports were perceived favorably, but MRI reports require enhanced clarity and emotional support.
- AI holds promise for improving patient understanding of radiology reports, with modality-specific refinements needed.
