Certain Investigations of Melanoma in Dermoscopy Images Using Texture Features and Machine Learning approach

Abstract

Melanoma is a serious form of skin cancer, which is developed in some pigment-producing cells called as melanocytes. Melanoma can develop anywhere on the skin, but certain parts are more likely to be than others. In men, it affects the chest and the back. In women, the legs are the most common site, present in neck and faces too. Early diagnosis can promote early treatment of melanoma, thereby can save many lives. Computer Aided Diagnostic (CAD) systems can be employed for early and accurate diagnosis of melanoma. The CAD system can assist the doctors in the recognition of skin cancer as benign and malignant by providing better results. newlineTo detect skin lesion a capable robust system is necessary and the process of classification is essential for early detection and prognosis to save human life with minimum effort and time. In this work, an improved robust technique is developed for classifying melanocytic tumors either as Melanoma and NonMelanoma on consideration with the digital Dermoscopy image s texture features. newlineThe proposed system contains the following phases: Image Acquisition, Pre-Processing, Image Segmentation, Feature Extraction and Classification. Image acquisition is an initial step in any image processing system. Dermoscopic images collected from dermatology service of Chalmeda Anand Rao Institute of Medical Sciences, Karimnagar, Telangana, India is used in this work. Image pre-processing is an essential step in melanoma detection as it removes noises, artifacts and enhances the quality of original image. The input image may have thick hairs surrounding the lesion in skin cancer images, which can be removed using Dull Razor approach. Also the images may be affected by noise during acquisition. These noises are reduced by employing a median filter. The hair removed, noise reduced image is enhanced by Min-Max Linear Contrast Stretch enhancement technique. The preprocessed images are given to the next phase of segmentation. newline newline