With the help of a mobile phone app developed in Graz, suspicious areas of skin and birthmarks can be scanned and the risk assessed – with 95 percent accuracy, as reported by the Medical University of Graz. The app was tested there in a clinical study.
The diagnostic and risk-based accuracy of the neural networks used in the app was evaluated. More than 1100 skin lesions from 238 patients who came to the clinic for a routine examination or for the removal of a skin abnormality (lesion) were examined. In addition to the experienced eyes of two dermatologists and histological findings, the algorithm based on artificial intelligence also played a role. The risk assessment of the algorithm was rated as “correct” if it corresponded to the assessment of both dermatologists or an existing histology. The result: more than 95 percent of the “skin samples” were correctly identified.