Unsupervised Method to Localize Masses in Mammograms

Abstract

Breast cancer is one of the most common and prevalent type of cancer that mainly affects the women population. chances of effective treatment increase with early diagnosis. Mammography is considered one of the effective and proven technique for early diagnosis of breast cancer. Tissues around masses look identical in mammogram, which makes automatic detection process a very challenging task; they are indistinguishable from the surrounding parenchyma. In this paper, we present an efficient and automated approach to segment masses in mammograms. The proposed method uses hierarchical clustering to isolate the salient area followed by extraction of features to reject false detection. We applied our method on two popular publicly available data-sets (mini-MIAS and DDSM). A total of 56 images from minimias database, and 76 images from DDSM were randomly selected. Results are explained in-terms of ROC (Receiver Operating Characteristics) curves and compared with other state of the art techniques. Experimental results demonstrate the efficiency and advantages of the proposed system in automatic mass identification in mammograms.

A Hybrid Textural and Geometrical Feature Extraction to Reveal Hidden Information From Suspicious Regions on Mammograms

“…The feature extraction stage of any CAD system directly affects the system performance by the way how data is represented. Besides utilizing statistical features [9][10][11][12][13][14][15][16][17][18] for breast cancer diagnosis, use of mammographic features [17,[19][20][21][22] like shape, spicule index, contour, size, density, and brightness would be more consistent through a radiologist's evaluation. Although pixel intensity is a sufficient measure for brightness, additional textural information [11,16,17,[23][24][25] is required for density determination as well as geometrical techniques [16,[26][27][28][29][30] should be used for shape and contour definition.…”