Radiologists Tested on AI X-Rays - Takeaway - 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

  • 1

    Radiologists identified AI-generated images with 75% accuracy.

  • 2

    Study involved 17 radiologists from six countries.

  • 3

    Three phases of image evaluation were conducted.

  • 4

    Musculoskeletal radiologists performed better (83%).

  • 5

    Common features of synthetic images included excessive symmetry.

  • 6

    Diagnostic accuracy was similar for both synthetics and real images.

  • 7

    Suggested strategies include watermarking and automated detection tools.

  • 8

    Limitations include small data set and potential bias from using GPT-4o for generation and detection.

Original Source(s)

Radiologists Tested on AI X-Rays

  • by Doug Brunk

  • April 1, 2026

Related Content

May Ultraprocessed Foods Raise Muscle Fat?

MRI study suggests higher intake is associated with greater intramuscular fat in patients at risk for knee osteoarthritis.

Radiology Groups Push Back on New Cancer Screening Guidelines

Radiology groups say the ACP's new screening recommendations conflict with nearly every major cancer organization and could cost thousands of lives annually.

Radiography Course Broadens RNs’ Understanding

Structured training reshapes how clinicians approach evidence, but real-world adoption still depends on time, culture, and leadership support.