Dermatology Study: AI Improves Diagnostic Accuracy at Point of Care
A new study published in the Journal of Investigative Dermatology (JID) found that an artificial intelligence solution has the ability to analyze skin conditions with the same degree of accuracy as primary care physicians (PCPs) referencing a visual aid.
The technology, called DermExpert from VisualDx—a company that makes clinical decision support software—was found to be equally effective when identifying diseases in light and dark skin types. Researchers believe this suggests that clinical decision support tools that are built on diverse data sets can augment physician decision-making and help to reduce medical errors and improve patient outcomes—particularly for patients of color.
They additionally noted that diagnosing the skin remains challenging as dermatologic disease presentation varies widely, and most medical schools offer 10 hours or less of dermatology-specific instruction. To contextualize general practitioners’ ability to diagnose skin conditions, the research team compared the accuracy of board-certified internal medicine physicians to DermExpert. When presented with a series of clinical images, PCPs accurately diagnosed skin conditions 36 percent of the time; with the assistance of a visual aid, their diagnostic accuracy increased to 68 percent. Similarly, DermExpert achieved 68 percent accuracy, indicating that PCPs can leverage VisualDx’s solution at the point of care to diagnose skin conditions with confidence and speed.
“Research shows that over a quarter of all patient visits involve a skin-related problem, which means all physicians must be able to identify dermatologic conditions, no matter their specialty,” said Steve Xu, M.D., a board-certified dermatologist, assistant professor of dermatology at Northwestern University Feinberg School of Medicine, and the study’s primary investigator. “Our research demonstrates a clear need for diagnostic tools like VisualDx in the exam room to make specialist knowledge readily accessible for the benefit of both the patient and the provider.
Given that skin of color remains significantly underrepresented in medical education and training, researchers content, Xu’s team also examined the AI algorithm’s ability to accurately identify diseases across a range of skin types. When VisualDx Plus DermExpert was applied to a wide set of clinical images, it achieved nearly equal efficacy in analyzing lighter (70 percent) and darker skin (68 percent). As the medical community continues to seek ways to eliminate racial bias in healthcare, this study indicates that AI can help reduce health disparities for traditionally marginalized patient populations.
Art Papier, M.D., co-founder and CEO at VisualDx, added, “Skin of color remains extremely underrepresented in medical literature and training, which has huge implications for dermatology outcomes.” He noted that because of that, VisualDx has deliberately collected hundreds of thousands of medical images that display a wide spectrum of disease presentation across all skin types. “Our mission at VisualDx is not to replace providers, but to give them that expert ‘second opinion’ needed to make the right call with confidence,” said Papier.
