A Statistical Assembled Model for Segmentation of Entire 3D Vasculature

We introduce a novel statistical deformable model called SAMTUS for the segmentation of soft tissue tubular structures. The model is composed of an assembly of statistically deformable tubular segments whereby the junctions of the tubular branches are used as landmarks for constructing the underlying point distribution model. The flexibility of SAMTUS is governed by two independent statistical models that describe the axis variation (Statistical Axis Model, or SAM) and the cross-sectional radius variation (Statistical Surface Model, or SSM) respectively. We also propose a SAMTUS based segmentation algorithm for an entire tubular structure. The approach has been applied to the segmentation of the three-dimensional vasculature of zebrafish embryo. The efficiency and robustness of this method is evaluated through quantification results on both sectional level and volumetric level.

• Type: Conference
• Year: 2006
• Where: ICPR
• Authors: Judy Jun Feng, Horace Ho-Shing Ip
• Keywords: Computer Vision, Deformable Tubular Segments, ICPR 2006, Statistical Deformable Model, Tubular Structure