Examination of Digital Mammogram Using Otsu's Function and Watershed Segmentation

Breast malignancy is one of dangerous illness among the women community and premature detection may facilitate to provide the appropriate treatment to diminish/eliminate breast cancer. Digital Mammogram (DM) is a commonly approved imaging scheme to record and scrutinize the breast cancer. This paper implements a novel hybrid approach based on the combination Otsu's multi-thresholding and Water Shed Segmentation (WSS) to mine the suspicious sections from the DM. Initially, the multi-level thresholding using the Bat Algorithm (BA) driven Otsu with a bi-, tri- and four-level thresholding is implemented to pre-process the DM. Afterward, a marker controlled WSS is implemented to mine the infected division of DM. The mined section is then evaluated using the Haralick texture feature in order to know the severity of the disease by examining its texture feature. In this paper, DM dataset with dense, medium, low and normal breast regions are analyzed independently with the proposed approach. The experimental result of this paper confirms that, proposed method is very proficient in extracting the breast malignancy from the considered DM database.