International Journal of Applied Science and Engineering Special issue: The 10th International Conference on Awareness Science and Technology (iCAST 2019) https://doi.org/10.6703/IJASE.202009_17(3).311 311 OPEN ACCESS Received: July 28, 2020 Accepted: August 31, 2020 Corresponding Author: Sasikala Shanmugam sundarsasi@gmail.com, sasikala.s.ece@kct.ac.in Copyright: The Author(s). This is an open access article distributed under the terms of the Creative Commons Attribution License (CC BY 4.0), which permits unrestricted distribution provided the original author and source are cited. Publisher: Chaoyang University of Technology ISSN: 1727-2394 (Print) ISSN: 1727-7841 (Online) Analyses of statistical feature fusion techniques in breast cancer detection Sasikala Shanmugam 1* , Arun Kumar Shanmugam 1 , Bharathi Mayilswamy 1 , Ezhilarasi Muthusamy 2 1 Department of ECE, Kumaraguru College of Technology, Coimbatore, India 2 Department of EIE, Kumaraguru College of Technology, Coimbatore, India ABSTRACT Breast cancer is one of the mortal diseases amongst women with increased incidences and mortality rate in every year globally. As its symptoms are not prominently noticeable in early stage, the early detection is difficult. Over the past four decades Mammography is used for diagnosing breast diseases. Most of CAD systems use either Cranio-Caudal or Medio-Lateral Oblique mammographic views. Radiologist will look at both the view for better diagnosis. To incorporate this perception with CAD, the detection performance of various statistical feature fusion in fusing the texture features of these two mammographic views are analysed in this work. The improved performance of accuracy: 97.5%, sensitivity: 100%, specificity: 97.2%, precision: 97.1%, F1 score: 96.23%, Mathews Correlation Coefficient: 0.952% and Balanced Classification Rate: 98.74% was achieved with Local Binary Pattern features fused through Canonical Correlation Analysis. Keywords: Breast cancer, Mammogram, MLO, CC, PCA, CCA, GDA, DCA, SVM 1. INTRODUCTION Female breast cancer is a killer disease in this era and its mortality and incidence have been increased by more than 14% and 20% respectively from 2008 (Sasikala et al., 2019). It is most frequent cancer in 154 countries out of 185 countries in terms of new cases and the leading causes of cancer death in 103 countries. In 2018, 2.1 million cases i.e. 1 in 4 cases among women were newly diagnosed worldwide (Sasikala et al., 2019) and an estimate of 0.0627 million breast cancer deaths of women was occur globally (WHO, 2020). To overcome the subjective variations in diagnosis by radiologist, Computer Aided systems for Detection (CADe) and Diagnosis (CADx) were developed. CAD through various breast imaging techniques play a significant role in breast cancer diagnosis. The General flow of a CAD system is depicted in Fig. 1. As Mammography is able to predict the presence of tumour before it becomes visible clinically, images of various mammographic views are widely used for diagnosis. Inspection of both Cranio-Caudal (CC) and Medio-Lateral Oblique (MLO) views better discriminates the tumours (Sasikala et al., 2019). Improvement in performance with two view information was evidenced by many Retrospective and Prospective studies as it mimics the radiologist’s perception in diagnosis. Fig. 2 shows the general flow of two view diagnostic system. A quantity which measures the spatial arrangement of image intensities with respect to a pixel and its predefined neighbourhoods is known as texture (Tourassi, 1999). Since the characteristics of tumour are better represented by texture features, malignancy in tumours could be better discriminated when texture-based image analysis is used in