Identification of Psoriasis Disease In Dermatology Using Machine Learning Technique

Abstract

Accuracy and reliability of classification technique are vital components for any Skin Lesion Detection System. This paper represents a dermatology based Skin Lesion Detection System (SLDS) to do classification automatically of dermatology images into psoriasis disease. The uniqueness of the detection system is an exploration of the various features related to the classification technique in Support Vector Machine (SVM) model. The features consist of texture, color space, redness of psoriatic disease and local binary pattern. The proposed SLDS detection system accompanied with the various technique like pre-preprocessing, data augmentation, feature selection, feature extraction and classification component which make the system robust. This system is trained on skin dataset that consists of psoriatic affected skin disease and healthy skin images. The training procedure mainly deal with machine learning parameters. At first, we have trained our system with 140 images and then tested with test set consist of 60 images, then, at last, the system performance was evaluated based on the various classifiers that we used like CART, SVM, and LR. The proposed SLDS system achieve classification accuracy with 100%, 98.36%, and 98.36% and respectively. We can be shown that the consistency factor and reliability has improved with the use of novel local binary feature through the analysis and research we found that the accuracy is improving with increasing the size of data.