Paper Title
Automatic Roi Generation For Lesion Segmentation Of Breast Ultrasound Image Using Directional Gradient

Abstract
In case of ultrasound image, lesion boundary suffers due to their peculiar nature, shadow, low resolution and the variance in shape. Automatically detecting tumors and extracting lesion boundaries become very difficult due to those problems. Many algorithms are proposed for automated segmentation of lesion from ultrasound image. Most of them are facing problem due to the diversity of whole ultrasound image. Automatic Region of Interest (ROI) selection is a good choice to execute the automated segmentation more specifically. The specific region should have lower variations and specific algorithm can be applied. In this paper we propose a novel technique to select the ROI for breast ultrasound image concentrating to the lesion boundaries. The images are prepossessed to detect the approximate boundary. On the basis of that boundary ROI is selected for automated segmentation of lesion boundaries. The proposed algorithm is also implemented to real patient data and shown as result at the later part of the paper. Keywords- Breast Ultrasound Image, Automatic Segmentation, ROI Generation