Clinical Report: Image Conditioning May Lower AI Detectability
Overview
A blinded reader study found that image-conditioned generative AI produced synthetic chest radiographs that radiologists were less likely to identify as artificial compared to text-only generated images. The detection rate for image-conditioned images was 34%, while it was 56% for text-only images.
Background
The ability to accurately distinguish between real and AI-generated medical images is crucial for maintaining diagnostic integrity in radiology. This study provides insights into how image conditioning affects the identification of synthetic radiographs.
Data Highlights
| Model | Text-Only Detection Rate | Image-Conditioned Detection Rate | Absolute Reduction |
|---|---|---|---|
| GPT-image | 49% | 30% | 19 percentage points |
| Gemini-image | 64% | 38% | 26 percentage points |
Key Findings
- Radiologists identified 34% of image-conditioned synthetic radiographs as artificial compared to 56% for text-only synthetic radiographs.
- For the GPT-image model, the detection rate decreased from 49% to 30% with image conditioning.
- For the Gemini-image model, detection decreased from 64% to 38% with image conditioning.
- Pneumothorax had the highest pooled AI detection rate at 55% among disease categories.
- Detection rates varied among readers, ranging from 61% to 98% for GPT-image and 69% to 97% for Gemini-image.
Clinical Implications
Radiologists should be aware of the potential for synthetic images to appear more realistic.
Conclusion
This study highlights the challenges in detecting AI-generated radiographs and the influence of image conditioning on radiologist identification.
Related Resources & Content
- Frontiers in Medicine, 2026 -- Detectability and healthcare implications of generative AI–synthesized chest radiographs: a blinded radiologist reader study
- the ophthalmologist, 2026 -- AI Enhances Meibomian Gland Detection
- MDSpire News, 2026 -- Radiologists Tested on AI X-Rays
- npj Digital Medicine, 2025 -- Unique Visual Preferences Influence Medical Imaging Diagnoses in Humans and AI
- ACR Approves First Practice Parameter for Imaging Artificial Intelligence, 2026
- the pathologist — Beyond Image Analysis: How AI is Reshaping the Pathology Workflow
- ACR Approves First Practice Parameter for Imaging Artificial Intelligence
- Detectability and healthcare implications of generative AI–synthesized chest radiographs: a blinded radiologist reader study - PMC
- A Systematic Review on Synthetic Medical Images Generation-Recent Trends and Future Opportunities - PubMed
This content is an AI-generated, fully rewritten summary based on a published scholarly article. It does not reproduce the original text and is not a substitute for the original publication. Readers are encouraged to consult the source for full context, data, and methodology.
