磁共振脑部图像壳核区域有监督蚁群分割算法研究电子与通信工程专业论文.docx

哈尔滨工业大学工程硕士学位论文 哈尔滨工业大学工程硕士学位论文 - - II - Abstract The image segmentation research for putamen region in brain MRI is based on ant colony optimization theory in swarm intelligence. The putamen region is the main region which causes Parkinsons?Disease(PD). The refined segmentation for the region is a challenging matter, which has not only a strong scientific value on brain science, and can help extracting characteristic parameters for pos t PD discrimination and diagnosis, but also can create a theoretic foundation for other more smaller regions refined segmentation in brain MRI. A supervised segmentation algorithm based on ant colony for putamen region in brain MRI is studied. Since the variance of the putamen template and the searching contour is adopted as the object function, the solution process for the supervised ant colony algorithm model proposed is transformed as the process of the minimum of the object function, or as the optimal searc hing path problem in the searching space. A new scheme for finding searching space is adopted. The normals of the points on the putamen template contour are given, and a searching space along the corresponding normal direction for a point on the template c ontour can be decided in the proposed scheme. The comparison between the proposed scheme and the scheme based on the centroid of the template contour proposed by Wang Xiaonian et al is described. The experimental results show that the proposed scheme can perfectly approach the shape of the putamen, and the implement is relative easy, and the segmentation effect is better. For the scheme adopted by Wang Xiaonian et al, since the putamen has a special shape, and the centroid of the putamen is near the most contour of the putamen, so that some points in the searching space are centralized and overlapped, and in some section, the fluctuant tendency of the searching space does not accord with the rough shape of the putamen, the scheme can result in a different segementation. By a