Extraction of Cancer Affected Regions in Mammogram Images by Clustering and Classification Algorithms

Abstract

Objectives/Backgrounds: The breast cancer has increased significantly in the last few years. It is one type of cancer and is the second deadliest disease in the world wide. Recently, cancer is diagnosed by various test such as mammography, ultrasound, etc. Mammography is used to breast imaging to help in detecting breast cancer. Methods/Statistical Analysis: The Mammogram images are taken for the analysis to find the tumor affected regions by data mining techniques in this research work. This work uses the Median filter method for noise removal and Gaussian filter for image enhancement of preprocessing the images. The k-Means algorithm, which is easily detected and extracts tumor area by means of intensity values by segmenting the mammography images. Two types of mammography images; normal and abnormal are given as input to the algorithms. After clustering the images by k-Means algorithm, the results found are classified by J48, JRIP, Support Vector Machines (SVM), Naive Bayes and CART algorithms to verify the accuracy of the results based on its pixel values. Findings: The performance of taken classification algorithms is compared and find out the best classifier in terms of its accuracy, sensitivity and specificity. Improvements: In the future, the other classifiers and feature selection algorithms are applied to extract the mammography images. Also, it gives more than fifty images for analysis.