Automated Diagnosis of Skin Diseases Using Gray Level Co-occurrence Matrix with Back-Propagation Neural Network Algorithm

Abstract

Various aspects of skin diseases inspired us to develop an automated diagnosis system of the disease and its affecting area(s).  The objective of this work is to analyze the methods to auto-detect skin diseases and to find out an optimal approach taking both diseases’ nature and their variety in consideration. Gray Level Co-occurrence Matrix (GLCM) defines the probability of occurrence of a gray level within its neighbors at a specific distance. Five GLCM features (Contrast, Homogeneity, Dissimilarity, Correlation and Entropy) have been extracted from the source images. A feed-forward neural network with Back Propagation Neural Network (BPNN) algorithm as the learning method is chosen. Mean Square Error (MSE) is selected as the error calculation method. Six skin disease types (Acne, Eczema, Erythema, Ichthyosis, Psoriasis and Scabies) have been classified. The average accuracy of the reference method is 79%, and it has been improved to 83.3%.

Keywords: Gray Level Co-occurrence Matrix (GLCM), medical imaging, skin diseases screening, Back Propagation Neural Network (BPNN)

Cite this Article

Kibria G, Islam MR. Automated Diagnosis of Skin Diseases Using Gray Level Co-occurrence Matrix with Back-Propagation Neural Network Algorithm. Research and Reviews: Journal of Computational Biology (RRJoCB). 2015; 4(1): 22–27p.