Can you spot a deepfake X-ray? Neither can your radiologist.
Researchers tested 17 radiologists to see if they could tell the difference between real X-rays and those generated by ChatGPT, with only 41% noticing anything was wrong initially. The study highlights the potential risk of deepfake technology in medical care, with even the models that generated the images unable to reliably detect them.
A recent study published in Radiology found that deepfake X-rays generated by ChatGPT can be highly convincing, even to experienced radiologists. The researchers used simple prompts to generate X-rays with specified anatomical locations, disorders, and levels of noise. The radiologists were initially asked to diagnose patients based on the synthetic images, with only 41% noticing anything was wrong. Even when they knew to look out for deepfakes, they only accurately differentiated them 75% of the time. The study's findings raise concerns about the potential disruption of medical care by deepfake technology. The researchers used four multimodal models to detect the deepfakes, but they were only able to distinguish them accurately 57-85% of the time.
This content was automatically generated and/or translated by AI. It may contain inaccuracies. Please refer to the original sources for verification.