A new study by researchers from Massachusetts Institute of Technology (MIT) has found that doctors are less accurate in diagnosing skin diseases based on images when the patient has darker skin. The study included over 1,000 dermatologists and general practitioners; dermatologists accurately diagnosed 38 percent of images overall, but only 34 percent of darker skin images, while general practitioners showed a similar decrease in accuracy for darker skin.
The researchers found that assistance from AI could help to improve doctors’ diagnostic accuracy, but these improvements were more pronounced with lighter-skinned patients. These findings highlight the need for further research and policy changes around dermatology education to address these disparities.
The study also builds upon previous research which found that the images used in dermatology textbooks and training materials are predominantly of lighter skin tones. While this may contribute to the noted discrepancies in diagnosis, the researchers also suggested that some doctors may simply have less experience diagnosing patients with darker skin.
AI tools were shown to assist in improving doctors’ diagnostic accuracy, with an AI algorithm developed by the MIT research team correctly classifying skin disease images 47 percent of the time. By using AI, the accuracy of diagnosis was increased to 60 percent for dermatologists and 47 percent for general practitioners.
Despite these improvements, general practitioners demonstrated a larger improvement in diagnosing lighter skin than darker skin with the help of AI. These discrepancies could potentially be due to a lack of experience or in-depth knowledge about how different skin diseases may appear on different skin tones.
Overall, the study advocates that more training on darker skin patients should be integrated within medical school curriculums and textbooks to potentially bridge this diagnostic gap. Furthermore, the study underscored the potential of AI in enhancing diagnostic accuracy for dermatologists and general practitioners.