3D Active Shape Model for CT-scan liver segmentation

Abstract

This paper present an automatic 3D liver segmentation based on Active Shape Model. It allows us to introduce a 3D modeling feature for the target organ to lead the segmentation. This method is tested on the dataset IRCAD which containe a 20 Computed tomography exams. These exams are obtained with different scanning protocol. Thence, we used two algorithms. First, we employed the Shape Context based Corresponding Point Model with a B-spline registration to normalize the 3D dataset with the landmarks mean distance equal to 95%. Then, we applied the active shape model. The experiments demonstrate that this algorithm is efficient and it have a tolerate value of Modified Hausdorff Distance of 3D matching between surface mesh using the iso-surface reconstruction and the Active Shape Model. Its range equal to 28.95mm.