A Feature Selection Framework for Granular Parakeratosis Lesion Detection

Abstract

The dermatoscopy images play an essential role in the identification of skin diseases such as granular parakeratosis. This disease appears as an intertrigo or as a non intertriginous rash (intertrigo is nothing but rash in the body folds) and sometimes causes itchy. Extracting the features from skin lesion image is very important for the classification of skin disease. Some features such as shape, size, and color can be extracted from the image depending upon the ABCD (asymmetry, border, color and diameter) rule of dermoscopy. These extracted features are mainly used to determine asymmetry and border irregularity. Asymmetry is one of the main attributes in the early detection of skin disease. The dermatologists inspect shape and irregular color distribution to identify the lesion asymmetry. This paper provides a large set of skin lesion feature extraction approaches for the lesion analysis. Several techniques exist for extracting the features from lesion images and the computational methods have been developed for enhancing the features and allow dermatologists in the earlier diagnosis of the disease. The methods or the procedures adopted by several researchers for extraction of features are explained and the effect of these methods on the skin lesion is evaluated using suitable metrics. Though extraction of the features from the skin lesion has been successfully addressed in several articles, still there is a requirement for the modification or development of new techniques for improving the efficiency of the diagnosis system.