In a multireader diagnostic accuracy study, radiologists correctly distinguished AI-generated radiographs from real images about 75% of the time, with similar performance across GPT-4o–generated multiregion images and RoentGen-generated chest radiographs. Diagnostic accuracy for abnormalities and perceived image quality were comparable between synthetic and authentic images, and neither radiologists nor multimodal large language models reliably identified all synthetic radiographs.

Source: Radiology