Radiologists Tested on AI X-Rays - Takeaways - MDSpire

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.

  • By

  • Doug Brunk

  • April 1, 2026

  • 4 min

Share

  • 1

    Radiologists identified AI-generated radiographs with 75% accuracy, highlighting challenges in detecting realistic synthetic medical images.

  • 2

    The study involved 17 radiologists from six countries and assessed their performance across three phases using 264 radiographs.

  • 3

    Diagnostic accuracy for identifying abnormalities was high, reaching 92% for synthetic and 91% for real radiographs.

  • 4

    Experience did not significantly enhance detection performance, although musculoskeletal radiologists outperformed others at 83% accuracy.

  • 5

    The findings underscore the need for training and safeguards against the misuse of synthetic images in clinical and legal settings.

Original Source(s)

Radiologists Tested on AI X-Rays

  • by Doug Brunk

  • April 1, 2026

Related Content

Case Report: C-arm–assisted peri-hip endoscopic debridement for calcific tendinitis at the deep insertion of the gluteus maximus

Multimodal imaging features of primary breast epithelial–myoepithelial carcinoma: a case report and literature review

Image-domain deep learning denoising for low-dose chest CT on a single 128-slice CT platform: a retrospective image-quality assessment