A novel algorithm for locating region of interest in breast ultra sound images

Breast Cancer is a common cancer found in Indian women. The main objective is to automate the process of Breast Cancer detection by identifying the Region of interest. In this paper a novel approach to identify and initialize the region of interest automatically in a Breast Ultrasound image is proposed. We propose Hybrid Improved Speckle reducing anisotropic diffusion filter in the pre-processing stage followed by a novel algorithm to extract the region of interest. Hybrid filter in preprocessing is designed by combining Gaussian blur filter and Modified SRAD filter. To extract the region of interest Black Hat Transform and Simple threshold algorithm is proposed. Modified SRDA is applied on the Gaussian blur image which reduces the speckle and enhances the image to find the ROI. We deploy 2 methods to find the ROI, black hat-transform and Simple Threshold. In this work 180 different BUS images where used out of which 78 where malignant and 102 was benign. Proposed work was applied on these images and got an accuracy of 96%. The proposed work makes the existing CAD system automatic by finding the Region of interest in BUS images.