Usage of Deep Learning in skin cancer detection

Image recognition is an important topic when it comes to skin cancer detection, since the incidence of skin cancer is rising sharply, worldwide, and the detection of pigmented lesions is still a challenge for dermatologists. The field constitutes an interface between dermatology and computer science and combines conventional detection methods with KI systems. Studies have shown that convolutional networks cause good outcomes in skin cancer detection due to its handling with big data, an AUC at 99% with a ResNet-152 was measured. The diagnostic accuracy is impressive and can be compared to a dermatologysts decision. However, there are some limitations, e.g. restricted training data, that need to be considered and also further explored, therefore at the moment it can be only used as an assisting tool.