An Improved Small Region Growing Method For Segmenting 3D Liver MDCT Images

Image segmentation is an important issue in image processing. It is also an academic difficulty of the low level vision in the field of computer vision. Since the 1970s,many researchers pay a great effort on the subject of image segmentation, but there is no general method and objective criterion to judge whether segmentation succeed or not. The most important application field of image segmentation method is the segmentation of medical image. The segmentation of medical image applying in medical anatomy plays an important role in diagnosis. So the study of medical image arithmetic is very urgent and necessary.In recent years, many researchers had put forward some segment technology combined with some given theory, method and instruments on the base of ecumenical segment method. But such these technology only can handle one kind of special image. Owing to the particularity of medical image, a reasonable segmentation needs medical knowledge besides ecumenical segment method. This paper put forward an improving small region growing method to segment 3-D liver MDCT image on the base of the structure and figure of human liver. From the results we can see the reasonability of this method. And this method still fit for CT image besides MDCT image.This paper included:1. MDCT image pre-transact process.2.Develop an improving small region growing method to solve themisclassification evocated by blur boundary between liver and adjacent organs.3. Remove non-liver adopting mathematic morphology and on the base of the structure and figure of human liver.4. Edge smoothness for the result of segmentation.5. Develop 3-D image display system for showing the segmented results based onOpenGL.What is the characteristic lies in: this paper's subject is MDCT image. MDCT image is a new CT scanning technology appearing in recent years. It shows a widerapplication foreground due to its great capability in boosting the efficiency in clinicalid diagnosis. This is a hot issue in the medical image processing. This paper appliesslice-by-slice method to separate the slice into many small regions, then judge if these regions distribute seed point or not according to some rules. After that, choose the seed point as the center of a circle and make the region growing inside this circle, then conform the ROI (Region of Interesting) into a whole liver tissue. This method had achieved satisfied effects in the process of which borderlines between liver and adjacent organ are blur. And this paper puts forward an optimal algorithm to delete non-liver tissue producing in the process of segmentation based on the comparability between human body surface shape and liver shape. Both of the two methods can solve misclassification in the normal segment method. And the 3-D image display system offers a convenient and economical means to transplant between all kinds of platform.