Cross-sectional study assesses ChatGPT-5's diagnostic accuracy in evaluating dental radiographs, indicating limitations in sensitivity.
Key Points
This study aims to assess the diagnostic accuracy of ChatGPT-5 in interpreting periapical radiographs related to root canal treatments and identifying pathosis.
Cross-sectional study analyzing 271 anonymized periapical radiographs classified as straightforward or complex
Used standardized prompts for ChatGPT-5 and compared results with general dentists and endodontic specialists
Calculated sensitivity, specificity, and overall accuracy using the McNemar test.
ChatGPT-5 showed high specificity (up to 99.3%) for normal or adequately treated findings
Sensitivity for short obturations was 13.7%, and for voids, it ranged from 9.0% to 22.7%
Overall accuracy was significantly lower (54.0%-63.2%) than that of general dentists (76.0%-85.6%) with p < 0.001.