Skin Diseases Classification using Deep Convolutional Neural Network

Abstract

Dermatological diseases are considered as the fourth cause of the problem of non-toxic diseases in recent years. These skin abnormalities are not affected by the physical human body as they also affect mental health. The tremendous development in medical technology has great potential in diagnosing skin diseases more quickly and accurately, but the cost of this diagnosis is still limited and expensive in addition to the high rate of diagnosis error for some types of the skin diseases. This study proposes a method that is based on the Convolution Neural Network( to detect and classify ten types of the skin diseases. The proposed CNN architecture in general nine convolution layers and two fully connected layers. The contribution of the current method is its ability to detect many visual similar diseases and some of them aren't detected by computerized programs previously. Nine diseases are detected by a certain program while most of the previous works work on a few diseases. All the parameters of this proposed CNN are determined experimentally. The total accuracy of the proposed model is 91.07%. This proposed model works well with many issues of the detection and classification of the skin diseases. It is an encouraging method compared with other methods.