PulseExploreJournal ClubResearchersJournals
Instagram
HomeJournal ClubExplore
Synapse
⌘+K
Synapse
May 3, 2026SHILAP Revista de lepidopterologíaOpen Access

Assessing the Realism of Artificial Intelligence-Generated Colorectal Polyp Images: An International Multicenter Blinded Reader Study

View Full Paper
Ask AI
Bookmark
Share

Authors

PSPhilipp SodmannRWRonja WeberVWValentin Wettstein

Discussion

Loading...

Member takes

Overview

Randomized trial evaluates sensitivity and accuracy of AI-generated polyp images in endoscopists, suggesting high realism.

Key Points

  • The study aims to assess the realism of synthetic colorectal polyp images generated by an AI model.
  • 53 endoscopists evaluated 20 real and 20 AI-generated polyp images in random order.
  • Evaluation was performed using a web platform, classifying images as real or synthetic.
  • The trial was registered under the identifier NCT07108569.
  • Sensitivity for detecting AI-generated images was 66% (95% CI, 65 - 67%).
  • Specificity for detecting real images was 80% (95% CI, 79 - 81%).
  • Overall accuracy of the assessments was 73% (95% CI, 72 - 73%).

Cite This Study

Sodmann et al. (2026) studied this question.

synapsesocial.com/papers/69f6e5308071d4f1bdfc5e76https://doi.org/10.1055/a-2865-3480
View Full Paper
Ask AI
Bookmark
Share