Radiologists Tested on AI X-Rays - Summary - MDSpire
From the Journals

Radiologists Tested on AI X-Rays

Study found synthetic radiographs were often difficult to distinguish from real images, with physicians identifying them correctly about 75% of the time.

Share

A study revealed that radiologists identified AI-generated radiographs with 75% accuracy, highlighting the challenge of distinguishing realistic synthetic images from actual ones. Led by Dr. Mickael Tordjman and Dr. Murat Yuce from the Icahn School of Medicine, the research involved 17 radiologists evaluating 264 images across three phases using different large language models (LLMs). While diagnostic accuracy for identifying abnormalities was similar for both synthetic and real images, the findings underscore the risks associated with artificial images in clinical practice, calling for enhanced training and detection strategies.

Original Source(s)

Radiologists Tested on AI X-Rays

  • by Doug Brunk

  • April 1, 2026

Related Content

Patients With Lung Cancer May Safely Receive Adequate SBRT Dosage in One Treatment

“This randomized phase II study demonstrated that 30 Gy in one fraction was equivalent to 60 Gy in three fractions in terms of toxicity, local failure, freedom from distal failure, progression-free su...

MRI May Improve Prostate Cancer Monitoring

Volume tracking and longitudinal imaging may enhance assessment of progression during active surveillance.

Lymphatic Imaging Tests a New Technique

A modified interstitial approach uses intradermal contrast injection with rapid digital acquisition.