Abstract

Lesions and its contours are prominent signatures to determine malignancy in mammograms. Detection of the masses and their spread in mammogram is important for radiologists. In this paper, Mammogram image is enhancement using homomorphic filtering and adaptive histogram equalization. The enhanced mammogram image is segmented using K means clustering and contour is extracted using morphological operations.The edges are detected by Sobel operator and extracted seven geometric features from the lesions. Lesions and its contours are prominent signatures to determine malignancy in mammograms.It is found that malignant lesions have speculated or ill-defined boundary and benign mass have smooth boundary. Classifications of malignant and benign are done by distance versus angle of signature.The average value of area in malignant and benign segmented mammogram images is 1073.6 and 316.7 square unit respectively. The value of area in malignant mammogram images is greater than benign images. The average range value of radius in malignant and benign mammogram images is 77.38 and 17.58 unit respectively. Signature value of range in malignant image is higher in comparison to benign image.

Keywords

Breast Cancer Detection – Mammography – K-Means Clustering – Boundary

Introduction

Medical images are rich in information that can be used for diagnosis and subsequent medical interventions. Information provide by medical image has become an indispensable part of today’s patient care. Cancer is the unrepressed development of unusual cells in the body which account for the most dangerous and life threatening diseases in the world.Cancer is the second leading cause of death globally, and was responsible for 8.8 million deaths in 2015. Globally, nearly 1 in 6 deaths is due to cancer[1]. Mammography is specialized medical imaging that uses a low-dose x-ray system to see inside the breasts. A mammography exam, called a mammogram, aids in the early detection and diagnosis of breast diseases in women[2].

Work has been done on segmentation of mass in past to know the spread of spiculation in the breast tissue. Mean shift algorithm and Fuzzy C-means and active contour models are used in [4] for the detection of masses. Suspicious focal areas are found for testing morphologic concentric layer (MCL) criteria, to detect mass region in mammogram [5]. Gradient vector flow (GVF)snake and multi-scale analysis using Gaussian pyramid has been proposed in[6] to segment masses in mammogram. At first they applied gaussian pyramid to make the image coarse, so that GVF snake is able to converge to the mass contour easily and quickly with less computation. Shape features like elongatedness, eccentricity, Euler number, Max Radius, Min Radius were used to distinguish four different shapes round, oval, lobular, irregular of mass by using C5.0 decision tree algorithm in [7]. Gabor filter banks are used for extracting local spatial textural properties of masses at different orientations and scales[8]. Multilevel wavelet decomposition method is proposed to extract mean, variance, standard deviation, entropy and mean of absolute deviation from wavelet components [9]. Boundary extraction of this mass is also very important, so that radiologists can judge whether the mass is benign or cancer. Rangayan et al in [10] proposed a region-based measure of image edge profile acutance by polygonal approximation and measured shape features like compactness, Fourier descriptors, central invariant moments and chord-length statistics to distinguish between circumscribed and spiculated tumors.In this article, the method to detect breast cancer is presented using k-means clustering algorithm.