Texture features neural classifier of some skin diseases

This paper presents a texture classification system for three skin diseases. In this study 162 images were involved where each disease group consisted of 54 images (34 for training and 20 for testing). The set of features were preprocessed by using principal component analysis (PCA) method for features reduction. The best results were obtained by using a back propagation neural network of 17 nodes at the input layer, 11 nodes at the hidden layer, and 3 nodes at the output layer. The network correctly succeeded to classify the input images to the mentioned groups with 70% for the first disease, 70% for the second one and 100% for the last disease