结合形变优化与三角剖分的CT图像肝脏分割

Liver Segmentation from CT Image Combined with Deformation and Triangle Subdivision Model

针对传统形变模型在形变过程中易陷入局部最优,导致难以实现肝脏凹陷区域准确分割的问题,提出一种形变模型,并将其应用于CT图像的肝脏分割.首先建立一种单纯型的表面模型,并利用此模型表示肝脏的初始边界;然后利用模型中顶点与其3个邻域顶点的关系构建一组新的内力及其约束模型;再引入气球力模型,将其与图像的Gabor特征结合构建形变模型的外力及其约束方法,促使模型在内力和外力的作用下更快速地逼近肝脏边界;为了使模型能够进入肝脏的凹陷区域,构建一种三角剖分优化算法,使模型在形变过程中可以插入新的顶点.将文中方法应用于医学图像计算和计算机辅助介入国际会议提供的肝脏数据中进行实验的结果表明,该模型对肝脏分割具有较好的适用性和鲁棒性,并获得了较高的分割精度.

As the traditional deformable model is easy to fall into local optimum, high precision segmentation for the liver depression area is difficult. In order to solve this problem, a novel algorithm based on deformation and triangle subdivision model is proposed for the liver segmentation from CT images. First, a simplex mesh model is established to represent the initial boundary of the liver. Second, a new set of internal force and constraint model are constructed based on the relationship between the vertex and its three neighborhoods vertices. Third, the external force and the constraint method are established by using the combination of the balloon force model and the Gabor feature of images to push the model more quickly approaching the liver boundary under the action of internal force and external fore. Forth, in the process of the model deformation, new vertices can be inserted to ensure the smoothness of the model and the accuracy of segmentation results for sag area of liver. Finally, the data provided by international conference on medical image computing and computer assisted Intervention are used to investigate segmentation results. The experimental results show that the proposed method for liver segmentation has better applicability and robustness, as well as gets higher division accuracy.