Objective
To measure the performance of smart phone applications which evaluate photographs of skin lesions and provide the user feedback as to their likelihood of malignancy.
Design
Case-control diagnostic accuracy study
Setting
Academic dermatology department
Participants
Digital clinical images of pigmented cutaneous lesions (60 melanoma cases and 128 benign lesion controls), all with histologic diagnosis rendered by a board-certified dermatopathologist, obtained prior to biopsy in patients undergoing lesion removal as part of routine care.
Main Outcome Measures
Sensitivity, specificity, and positive and negative predictive values of four smart phone applications designed to aid non-clinician users in determining if their skin lesion is benign or malignant.
Results
Sensitivity of the four tested applications ranged from 6.8% to 98.1%. Specificity ranged from 30.4% to 93.7%. Positive predictive value ranged from 33.3% to 42.1%, and negative predictive value ranged from 65.4% to 97.0%. The highest sensitivity for melanoma diagnosis was observed for an application that sends the image directly to a board-certified dermatologist for analysis and the lowest sensitivity was observed for applications that use automated algorithms to analyze images.
Conclusions
The performance of smart phone applications in assessing melanoma risk is highly variable, and 3 out of 4 smart phone applications incorrectly classified 30% or more of melanomas as unconcerning. Reliance on these applications, which are not subject to regulatory oversight, in lieu of medical consultation, has the potential to delay the diagnosis of melanoma and to harm users.