Print Email Facebook Twitter Automated Expansion of Statistical Shape Model Training Set for Femur Title Automated Expansion of Statistical Shape Model Training Set for Femur Author Li, H. Contributor Vilanova, A. (mentor) Faculty Electrical Engineering, Mathematics and Computer Science Department Intelligent Systems Programme Digital Media Technology Date 2015-08-31 Abstract Active Shape Model (ASM) uses Statistical Shape Model (SSM) to fit the manually labeled landmark point in Magnetic Resonance Image (MRI) scan and segment the femur from the MRI scan. The SSM is trained from a training set containing 275 complete femur meshes. In general, with more femur meshes in the training set, there would be more shape variance information in SSM, and the segmentation result would be more accurate. Currently, it is hard to get complete femur meshes. However, 2000 partial femur meshes are available. A method to incorporate these partial femur meshes to the SSM training set is proposed. This method includes building point to point correspondence between partial femur and complete femur, filling missing values to perform Principal Component Analysis(PCA) and designing the mechanism to utilize the content in partial femurs. Before incorporation, a rejection criteria is set to reject the unsatisfactory partial femur meshes that cannot contribute to the SSM. A evaluation method is designed and performed to validate the incorporation method. The evaluation results show the accuracy of outliers of SSM fit can be improved by incorporating partial femurs to the SSM training set. Subject Statistical Shape ModelTraining Set Expansion To reference this document use: http://resolver.tudelft.nl/uuid:410aea7a-0397-40dd-b5ed-9f2a9a5f5974 Embargo date 2017-07-31 Coordinates 51.993789, 4.386978 Part of collection Student theses Document type master thesis Rights (c) 2015 Li, H. Files PDF Li2015.pdf 4.48 MB Close viewer /islandora/object/uuid:410aea7a-0397-40dd-b5ed-9f2a9a5f5974/datastream/OBJ/view