Detection of Breast Cancer Using Fusion of MLO and CC View Features Through a Hybrid Technique Based on Binary Firefly Algorithm and Optimum-Path Forest Classifier

Abstract

Breast cancer is a leading killer disease among women of the new era. As per the GLOBOCAN project, the breast cancer incidences showed an increase from 22.2 to 27% globally from 2008 to 2012. Many times, no obvious symptoms were identified in breast cancer patients. Accurate detection of breast cancer at the earliest stage is very much essential to reduce mortality. Mammography has been used as a gold standard for over 40 years in diagnosing breast diseases. Computer-Aided Detection (CAD) systems have been developed to avoid the subjective analysis of screening mammograms made by radiologist. Craniocaudal (CC) view and Mediolateral Oblique (MLO) view are commonly used for breast cancer detection and diagnosis. Detection accuracy of breast cancers can be improved as the number of views is increased. This work is focused to improve the detection performance by fusing Local Binary Pattern (LBP) features extracted from MLO and CC view mammograms through a hybrid feature fusion technique based on Firefly algorithm and Optimum-Path Forest classifier. Seven performance metrics such as accuracy, sensitivity, specificity, precision, F1 score, Mathews Correlation Coefficient (MCC) and Balanced Classification Rate (BCR) were used to analyse the detection performance. The proposed work shows better performance when compared to existing work in literature.