Font Size: a A A

Study On Segmentation Of Breast Tumor In Ultrasonogram

Posted on:2009-11-03Degree:MasterType:Thesis
Country:ChinaCandidate:T ZhaoFull Text:PDF
GTID:2198360272960967Subject:Computer software and theory
Abstract/Summary:PDF Full Text Request
Breast cancer continues to be a significant public health problem in the world. Ultrasound is widely used on diagnosing mammary cancer for it has no radiation to body and the price is relatively lower than others. Being an assistant to doctors, CAD applied to early breast cancer detection improves image qualities as well as brings diagnosis enhancement greatly. This paper adapts the technology of median filtering combined with linear transformation of gray level to denoising and enhancing ultrasonogram.This paper researches the ultrasonogram segmentation technique used in early diagnosing mammary cancer. According to the main points of each technique and their advantages and disadvantages, this paper proposes an improvement algorithm based on region growing of ultrasonogram segmentation. This algorithm improves the method choosing seed. First, to segment the image roughly by way of threshold value. And then, taking the core of black pixel of binary image as the center, to choose a 7×7 window. To count the average grey level of all the pixels in the window. To choose the pixel whose grey level most close to average level as the seed point for region growing. According to statistics near the seed and the region already growed, confirm the growing threshold value. The edge of growing region combined the both advantage is shot of burrs and more smooth. And it is accurate to extract the edge of tumor relatively.In order to extract tumor ultimate boundary from the obtained intial tumor region, this paper proposes a method extracting tumor boundary based on morphology and boundary following. Through this method, morphology is used to filter the initial tumor region, and improved boundary following is used to extract tumor boundary. Improved boundary following method gives a definition of position code. According to the position of the current point contrasting to the former boundary point, the starting point is decided for searching the next target point. So, the target point can be found by searching most 5 times, not necessary to search all the points of the eight regions. This method save the time for boundary following while find exactly tumor boundary.
Keywords/Search Tags:Mammary Tumor, Ultrasound, Image Segmentation, Region Growing, Boundary Following
PDF Full Text Request
Related items