基于MC算法的高质量脊柱CT图像三维重建  

HIGH-QUALITY 3D RECONSTRUCTION OF SPINE CT IMAGES BASED ON MC ALGORITHM

作　　者：许婉露[1] 李彬[1] 田联房[1] 

机构地区：[1]华南理工大学自动化科学与工程学院广东广州510640

出　　处：《计算机应用与软件》2013年第6期79-82,92,共5页Computer Applications and Software

基　　金：教育部高等学校博士学科点专项科研基金项目(200805610018);粤港关键领域重点突破项目(佛山2010Z11);广东省教育部产学研结合项目(2009B090300057);广东省自然科学基金项目(S2011010005811);华南理工大学国家人体组织功能重建工程技术研究中心以及广东省生物医学工程重点实验室课题

摘　　要：从脊柱CT图像中重建出脊柱的三维模型以提供直观的术前病灶信息,能够有效辅助高难度的脊柱畸形矫正手术。针对传统MC(Marching Cubes)算法存在的重建表面不平滑、结构拓扑歧义的局限以及人体脊柱重构碎片过多的特点,提出一种基于保边局部高斯滤波与三维区域增长的改进型MC算法。该算法采用保边滤波去噪并增强边缘,局部高斯滤波平滑待重建区域以改变原有体素类型,减少二义性体素对数,有效地解决了重建表面不平滑与结构拓扑歧义问题;采用基于三维区域增长的双阈值分割算法,大大减少碎骨重建的数量。实验证明,采用高质量重建算法重建的脊柱三维模型能够满足医学三维可视化的要求。Reconstructing 3D model of spine from its CT images for providing intuitive preoperative lesion information can effectively assist the high-difficulty spine deformity corrective surgery. As traditional marching cubes （MC） algorithm has the limitations in roughness on reconstruction surface and topological ambiguity, as well as too many fragments in human spine reconstruction, in this paper we propose an improved MC algorithm which is based on edge-preserving local Gaussian filtering and 3D region growing. The algorithm adopts the edge-preserving filtering to eliminate the noises and enhance the edges, and uses the local Gaussian filtering to smooth the pending reconstruction areas for changing original cube types and reducing the number of ambiguous voxels, these effectively solve the problems of roughness on reconstruction surface and topological ambiguity. The dual-threshold segmentation algorithm based on 3D region growing is applied, which can significantly reduce the number of bone fragments reconstruction. Experimental results demonstrate that the 3D spine model reconstructed on this high-quality reconstruction algorithm can serve well the purpose of medical 3D visualisation.

关 键 词：脊柱CT图像 MC算法 三维重建 保边局部高斯滤波 三维区域增长 

分 类 号：TP301[自动化与计算机技术—计算机系统结构]