3D Statistical Shape Models to Embed Spatial Relationship Information

Abstract. This paper presents the creation of 3D statistical shape models of the knee bones and their use to embed information into a segmentation system for MRIs of the knee. We propose utilising the strong spatial relationship between the cartilages and the bones in the knee by embedding this information into the created models. This information can then be used to automate the initialisation of segmentation algorithms for the cartilages. The approach used to automatically generate the 3D statistical shape models of the bones is based on the point distribution model optimisation framework of Davies. Our implementation of this scheme uses a parameterized surface extraction algorithm, which is used as the basis for the optimisation scheme that automatically creates the 3D statistical shape models. The current approach is illustrated by generating 3D statistical shape models of the patella, tibia and femoral bones from a segmented database of the knee. The use of these models to embed s...

• Type: Conference
• Year: 2005
• Where: CVBIA
• Authors: Jurgen Fripp, Pierrick Bourgeat, Andrea J. U. Mewes, Simon K. Warfield, Stuart Crozier, Sébastien Ourselin
• Keywords: Computer Vision, CVBIA 2005, Knee Bones, Segmentation Algorithms, Statistical Shape Models