Evaluation of psoriasis skin disease classification using convolutional neural network

Abstract

Skin disease has lower impact on mortality compared to others but instead it has greater effect on quality of life because it involves symptoms such as pain, stinging and itchiness.  Psoriasis is one of the ordinary skin diseases which are relapsing, chronic and immune-mediated inflammatory disease.  It is estimated about 125 million people worldwide being infected with various types of skin infection.  Challenges arise when patients only predict the skin type disease they had without being accurately and precisely examined.  This is because as human being, they only observe and look at the diseases on the surface of the skin with their naked eye, where there are some limits, for example, human vision lacks of accuracy, reproducibility and quantification in the collection of image information.  As Plaque and Guttate are the most common Psoriasis skin disease happened among people, this paper presents an evaluation of Psoriasis skin disease classification using Convolutional Neural Network.  A total of 187 images which consist of 82 images for Plaque Psoriasis and 105 images for Guttate Psoriasis has been used which are retrieved from Psoriasis Image Library, International Psoriasis Council (IPC) and DermNet NZ.  Convolutional Neural Network (CNN) is applied in extracting features and analysing the classification of Psoriasis skin disease.  This paper showed the promising used of CNN with the accuracy rate of 82.9% and 72.4% for Plaque and Guttate Psoriasis skin disease, respectively.