Analysis of Skin Cancer using K-Means Clustering and Hybrid Classification Model Janney J Bethanney1,*, Roslin S. Emalda2 1Research Scholar, School of Electrical and Electronics, Sathyabama Institute of Science and Technology, Chennai, Tamilnadu, India 2Associate Professor, School of Electrical and Electronics, Sathyabama Institute of Science and Technology, Chennai, Tamilnadu, India *Corresponding Author: Bethanney Janney J, Research Scholar, School of Electrical and Electronics, Sathyabama Institute of Science and Technology, Chennai, Tamilnadu, India. Email: jannydoll@gmail.com
Online published on 19 August, 2019. Abstract Skin cancer has become the most common form of cancer and most well-known disease of the human. Early identification of skin cancer is important for improving prognosis, as patients with specific conditions can treat these diseases properly when distinguished in the initial stages. Therefore, computer-aided diagnostic systems must be developed to assist early detection of skin cancer. The most critical problem in skin cancer image evaluation is the process of segmentation. The basic idea of image segmentation is to consider equality as important criteria for dividing an image into important regions. Hence this proposed system aims to enhance a portion of the current strategies and new procedures to provide the correct, fast and reliable automated diagnosis of skin cancer. This paper proposed an early detection algorithm for skin cancer using k-means clustering and texture analysis of Local Binary Pattern, Red, Green, Blue Channels and Gray Level Cooccurence Matrix techniques. The image classification is carried out using the hybrid classification model. The Genetic Algorithm-Artificial Neural Network is a hybrid method used to classify the given dataset into a cancerous or non-cancerous image. This system is tested using a dermoscopic image dataset. The result of the proposed framework is compared to certain existing methods to achieve accuracy and performance. It results better in the process of segmentation and extraction of features. And combinations of extraction techniques also provide effective results for classifying the image of cancer and non-cancer. Top Keywords Pre-Processing, K means clustering, Feature Extraction, GLCM (Gray level co-event Matrix), GA-ANN (Genetic Algorithm-Artificial Neural Network). Top |