Chicago – Smartphone applications that use algorithms to detect skin cancer may be unreliable, finds a new study from the University of Pittsburgh Medical Center.
Researchers used four apps to assess 188 images of skin lesions, 60 of which were melanoma. Three of the four apps wrongly diagnosed 30 percent or more of the melanoma cases as “unconcerning,” according to a press release from JAMA Network, which published the study.
The app that was most sensitive to melanoma provided images to a board-certified dermatologist for analysis. Conversely, sensitivity was lowest in apps that tested the images using automated algorithms.
Researchers cautioned against relying on the apps because doing so – rather than seeking a doctor – could delay proper diagnosis and treatment.
The study was published online Jan. 16 in JAMA Dermatology.