A Novel Algorithm to Select a Seed Point Automatically In Breast Ultrasound Image

摘要

In automatic segmentation of breast ultrasound images, seed point is to be selected automatically for contour evolution to proceed. The total computation task involved in the initialization of seed point, occupies most of the system time and resource. In this paper, the proposed method computes and identifies the seed point automatically by which automatic contour initialization is done. In pre-processing stage Improved SRAD filter is developed and method for selection of seed point automatically is based on the changes in gray level intensities or texture features, which serves as an initialization for level set segmentation. The seed point that is automatically plotted on to Ultrasound B-scan images also plotted on its corresponding elastogram pair.Proposed method is applied on to 50 US B scan images of benign solid mass, 80 malignant solid massed and 30 images of complex and simple cysts. For validation process, the seed point obtained by the approach mapped on the image pairs. The above approach shows us that the work presented will successfully and seed point using texture values of US B scan Images with an accuracy of 87%. With this proposed method the computational time and resource used for segmentation of Breast legions will be minimized to maximum extent and is more accurate.