We developed a new method for liver segmentation and modeling in single phase CT angiographic images, which employs boundary feature extraction and shape reconstruction based on radial basis function (RBF). In the method, segmentation is performed in several steps; initial shape definition, surface voxel extraction based on voxel features such as curvature, and shape reconstruction by RBF. In this paper, we describe our method and present our experimental results for validation of our method regarding segmentation performance by using several clinical dada sets.