Deep Learning Models Based on Artificial Intelligence Can Now Identify a Person’s Race from Their X-Ray Film.
Artificial Intelligence (AI) continues to demonstrate astonishing capabilities. Recently, a team of researchers from the United States, Canada, and Taiwan has trained their AI to achieve a level of performance that a “real-life” doctor cannot replicate.
AI can accurately predict a patient’s race based solely on their X-ray.
Specifically, the AI can predict a patient’s race with an accuracy of up to 90% after being trained for a certain period.
“We discovered that the AI’s ability to predict race is accurate even from damaged, cropped, or edited medical images,” the research team stated. “In such cases, even clinical experts cannot achieve this, leading to risks in implementing treatment methods.”
To accomplish this, the researchers utilized hundreds of thousands of existing X-ray images that were meticulously labeled with the patient’s race to “train” the AI.
However, they also admitted that “it is unclear why the AI system is capable of identifying race so effectively from images that inherently do not contain accompanying information,” something that even a well-trained doctor cannot do.
Even when provided with limited information, such as by removing clues about bone density or focusing on a small part of the body, the AI still performs remarkably well in image-based diagnosis.
It is known that this research builds upon the results of a previous study, which involved scanning X-ray images using AI to look for signs of disease in Black individuals. Essentially, the AI mimics human thought processes to quickly detect characteristics within the data.
This study has been published in the journal The Lancet Digital Health.
