Radiologists correctly identify AI-generated images 75% of the time, with CT and MRI images being particularly challenging to spot.
Key Details
- 1Study published July 16 in Clinical Radiology assessed 182 radiologists.
- 2Radiologists correctly identified AI-generated images 75% of the time; real images at 83.4%.
- 3AI-generated CT and MRI images were hardest to distinguish; x-ray and ultrasound fakes detected more easily.
- 4Radiologist experience or AI familiarity did not affect accuracy; specialty interest did (+3.8%).
- 5Synthetic images may benefit training and data privacy but also risk misuse (e.g., in research or patient records).
Why It Matters
This study highlights both the utility and the risks of high-fidelity synthetic radiology images. The difficulty in distinguishing AI fakes raises concerns for research integrity, clinical safety, and stresses the need for detection standards and awareness.
Source
AuntMinnie
Related News
•Radiology Business
Meta Funds Ohio State’s Radiology AI Innovation Projects
Meta will fund Ohio State researchers to explore AI-driven advancements in radiology, including chest X-ray reporting and disease prediction.
•AuntMinnie
ChatGPT-5 Performs Poorly at Detecting Distal Radius Fractures on X-rays
ChatGPT-5 missed about half of distal radius fractures on x-rays, showing limited diagnostic capability in a real-world scenario.
•AuntMinnie
AI Triage Tool Speeds Up ICH Head CT Turnaround in ED
AI-based triage for intracranial hemorrhage on head CT reduced emergency department report turnaround by 13 minutes.