Automatic detection of malignant neoplasm from mammograms
Asim Ali Khan, M. Adil Khan, Ajat Shatru Arora · 2015
Breast cancer is one of the leading cause of death in woman worldwide both in developed and developing nations as per the records from World Health Organization (WHO). The World Health organization stated that more than 1.2 million women were found with breast cancer and more than 700,000 women lost their life every year in the world [1]. Mammograms are already known for its fuzzy nature, in addition to it a fuzzy classification characteristic between malignant and benign lesions; make the detection a challenging task. Results of proper extraction of ROIs prove the successful execution of preprocessing steps like label removal, pectoral removal and de-noising. To screen out the non-mass candidates from the mass ones, segmentation, texture based feature extraction and classification using Support Vector Machine (SVM) and Artificial Neural Network (ANN) is carried out. Maximum Sensitivity of 100% in all categories proves that zero probability of missing out Normal candidate while the screening process of Malignant from set of both, overall accuracy respectively ranges from 100% to 83.33% with an average of 98.90% when Normal and Malignant are classified, overall accuracy ranges from 92.33% to 80.00% with an average 84.75% when Normal and Benign are classified, and 100% to 85.71% with an average 94.90% when Benign and Malignant are classified using SVM. Whereas classification rate with ANN classifier is able to reach approx. of 92.60%, 87.50% and 90.00% respectively.