A new study from MIT researchers suggests that doctors are less accurate in diagnosing skin diseases on patients with darker skin through images alone. The study analyzed the accuracy of over 1,000 dermatologists and general practitioners and found that dermatologists correctly diagnosed about 38% of the images they saw, but only 34% for those of darker skin tones. General practitioners, who were less accurate overall, also demonstrated a similar decrease in accuracy with darker skin.
However, the study also found that assistance from artificial intelligence algorithms could improve the doctors’ accuracy. These improvements were more prominent for patients with lighter skin tones. The researchers believed that the discrepancy could be due to the fact that dermatology textbooks and training materials predominantly feature lighter skin tones. Doctors might also have less experience with treating patients with darker skin.
The researchers used 364 images from dermatology textbooks and other sources to test the doctors’ diagnostic accuracy. The diseases featured included inflammatory skin diseases like atopic dermatitis, Lyme disease, and secondary syphilis, and a rare form of cancer called cutaneous T-cell lymphoma (CTCL), which could look similar to an inflammatory skin condition.
The research team invited a group of 389 board-certified dermatologists, 116 dermatology residents, 459 general practitioners, and 154 other types of doctors to the study. Each participant was shown 10 images and asked to provide their top three predictions for the disease each image could represent. They were also asked whether they would recommend a biopsy for the patient.
The results showed that specialists in dermatology had higher accuracy rates, correctly diagnosing 38% of the images compared to 19% for general practitioners. Both groups, however, demonstrated a decrease in accuracy when diagnosing skin conditions based on images of darker skin.
Additionally, the researchers developed an AI algorithm and found that it improved diagnostic accuracy for both dermatologists and general practitioners. However, general practitioners showed greater improvement on images of lighter skin than darker skin. The researchers hope these findings will encourage medical schools and textbooks to incorporate more training on treating patients with darker skin. The findings could also help in developing AI assistance programs for dermatology.
