Robust Skin Disease Diagnosis with Deep Belief Network

Abstract

The skin is one of the first lines of defence against environmental influences such as
sunlight, bacteria and germs, which cause various skin diseases. In addition to the bad psychological and
physical impact caused by skin disease. So, in recent years, many artificial intelligence (AI) algorithms
have appeared that can recognize Images, through which skin diseases can be diagnosed, avoiding
traditional methods that rely on visual examination and self-evaluation based on experience. The paper
aims to classify a group of skin diseases according to the type of disease, such as Atopic Dermatitis,
Dyshidrotic Eczema and Nummular Dermatitis using a deep belief network algorithm (DBN) that was built
to suit the work. A global dataset was also used, obtained from the Kaggle website, and after conducting
experiments on it, the algorithm achieved high accuracy in diagnosing diseases, the percentage reached
98.773%. It is possible in the future to use the network to classify other types of diseases after providing it
with a large number of images of affected people