Doctors are less accurate when diagnosing skin diseases in people with darker skin, according to a study by MIT researchers. The researchers found that dermatologists accurately characterized 38% of images of skin diseases, but only 34% of those images were of darker skin. The results were similar for general practitioners. The research team suggested that the use of an artificial intelligence (AI) algorithm could improve doctors’ accuracy, although improvements were greater when diagnosing lighter skin.
The study highlights a disparity in the diagnostic accuracy across skin tones and suggests that this could be due to the predominance of lighter skin tones in dermatology textbooks and training materials. Additionally, some doctors may have less experience in treating patients with darker skin.
The study team, led by Matt Groh PhD ’23 from the Northwestern University Kellogg School of Management, compiled 364 images from dermatology textbooks and other sources, representing 46 skin diseases across various shades of skin. They recruited subjects through Sermo, a social networking site for doctors, including 389 board-certified dermatologists, 116 dermatology residents, 459 general practitioners, and 154 other types of doctors.
Researchers found that dermatologists had higher accuracy rates than general practitioners. However, both groups lost about four percentage points in accuracy when diagnosing skin conditions on darker skin. Dermatologists were also less likely to refer darker skin images of CTCL, a rare form of cancer, for biopsy but more likely to refer them for biopsy for noncancerous skin conditions.
After evaluating how doctors performed on their own, the researchers also gave them additional images to analyze with the assistance of an AI algorithm. This AI algorithm, trained on around 30,000 images, had an accuracy rate of about 47%. The researchers found that using the AI assistance improved accuracy for both dermatologists and general practitioners. However, general practitioners showed greater improvement on images of lighter skin than darker skin.
The researchers hope their findings stimulate more training on patients with darker skin in medical schools and textbooks and help guide the deployment of AI assistance programs for dermatology.
