Geautomatiseerde analyse van huidkanker-app onbetrouwbaar
Automated analysis of skin cancer app is unreliable
Nederlands Tijdschrift voor Dermatologie en Venereologie , Volume 28 - Issue 6 p. 10- 13
Several smartphone apps claim to be able to find skin cancer early but studies have shown mainly poor results. However, those study groups consisted of patients whose investigated lesions were mainly selected by a medical professional. The SkinVision app (SVA) automatically analyses pictures of lesions and provides an instant rating as low, medium or high risk for skin cancer. Our aim was to find out whether the rating of the app is reliable in a real-life setting, in which consumers themselves selectthe lesions without interference of a health professional. Visitors of the National Skin Cancer Day in four university medical centres (in 2017) were asked to select up to two lesions they wanted to be analysed by the app. Then, the same lesions were rated by a dermatologist who was blinded to the rating of the app. Onehundred and twentyfive visitors with 199 lesions were included. In 90 cases (45%) the app was not able to perform an analysis (of which nine basal cell carcinoma, four atypical nevi and one lentigo maligna). More than two thirds of the cases with a red or yellow rating by the app were diagnosed as a benign nevus or a seborrheic keratosis by the dermatologist. The interobserver agreement between the ratings of the app and the dermatologist was poor (weighted kappa = 0.016; 95% CI -0.083-0.115; p = 0.737), a value comparable with mere chance. These results stress the urgent need for regulations for apps claiming to have an impact on public health. Up to now a reliable quality mark which could guide consumers and health professionals is still lacking.
|mobiele telefoon – vroegdiagnostiek – huidkanker – melanoom – basaalcelcarcinoom – geautomatiseerde analyse – app(licatie), smartphone application – skin cancer – melanoma – basal cell carcinoma – automated analysis|
|Nederlands Tijdschrift voor Dermatologie en Venereologie|
|Organisation||Department of Dermatology|
Chung, Y., van der Sande, A.A.J., Roos, K.P., Bekkenk, M.W, de Haas, E.R.M, Kelleners-Smeets, N.W.J, & Kukutsch, NA. (2018). Geautomatiseerde analyse van huidkanker-app onbetrouwbaar. Nederlands Tijdschrift voor Dermatologie en Venereologie, 28(6), 10–13. Retrieved from http://hdl.handle.net/1765/128336