Abstract
For some biomedical purpose, a specific region in a target 3-D image needs to be extracted. The 3-D image usually consists of many 2-D slices. In this thesis, a semi-automatic process will be proposed. First, template-driven segmentation is presented to find an approximate outline and position of the specific region on each slice of the target 3-D image based on a set of template 3-D image data. After modifying the outline and position by hand, a proper initial position of the contour for snake algorithm on each slice is obtained. This is the only manual work in this process. The snake algorithm will extract the specific region on each slice by deforming the contour. Thus, some energy functions at each point on the contour are defined, and the contour will be deformed depending on the energy functions. There are five energy functions presented in this thesis: (i) Continuity energy function is used to maintain the continuity of the contour. (ii) Curve energy function is used to maintain the smoothness of the contour. (iii) Image energy function is used to find a possible boundary of the specific region. (iv) Balloon energy function is used to expand or shrink the contour. (v) Additional energy function is used to cope with the problem of the blurred block.The process is more efficient than extracting the specific region on each slice by hand. Moreover, the problem of the blurred blocks in the specific region will be addressed.