Anatomical landmark and region mapping based on a template surface deformation for foot bone morphology Jaeil Kim 1, Sang Gyo Seo 2, Dong Yeon Lee 2, Jinah Park 1 1 Department of Computer Science, KAIST, South Korea 2 Orthopedic Surgery, Seoul National University Hospital, South Korea
< Anatomical landmark and region mapping based on a template surface deformation for foot bone morphology > < Jaeil Kim > My disclosure is in the Final AOFAS Mobile App. I have no potential conflicts with this presentation.
Introduction Anatomical point and region landmarks provide a basis to quantify the morphological changes of bones and the joint orientation/motion based on anatomical knowledge [1-2] However, determining the landmarks on the bone surfaces is a difficult and time-consuming task, due to the large variations of size and shape, and the lack of the salient features of the landmarks How to define the sub-regions of bone surfaces consistently? [ Volume Rendering (left) and bone surface meshes (right) ] 3
Contributions We propose a template-based landmark mapping method for the consistent and automated landmark assignment on individuals bone surfaces. Template model in our approach Encoding the generic shape features of the structure as a surface mesh Including the anatomical landmarks as subsets of the points of the surface mesh Proposed template-based landmark mapping Building a pairwise correspondence between the template and the target volume Achieved by a non-rigid surface-to-volume registration technique To place the anatomical landmarks and compare them via transitive relation between subjects 4
Outline Segmentation 0.0 Point Displacement 1.5 (mm) 2. Non-rigid Template-to-Volume Registration with Shape Correspondence 1. Template Model with Landmarks 3. Individualized Model and Landmarks 5
Template Construction Template surface model Generated from an image atlas of target structures Image atlas = mean shape image of the input binary masks Marching cubes algorithm + regular surface sampling Anatomical region and point landmarks Assigned manually onto the template surface model We used a point selection tool in the Paraview software (www.paraview.org, version 4.1.0) Save the landmarks as subsets of the point indices of the template model [Template model and region landmark] 6
Template-to-Volume Registration Two steps of template-to-volume registration 1 st step: a rigid alignment of the template model to target volume using iterative closest point algorithm 2 nd step: a non-rigid template deformation to boundaries A progressive deformation based on a Laplacian surface representation and a flexible weighting scheme of rigidity [3] Preserving the geometric features (relative area of triangles and local curvature) as strong as possible during the template surface deformation To place the points of the template to anatomically corresponded positions on the target volume [Objective function of the Laplacian surface deformation] 7
Experiments on Human Data - Calcaneus Study Materials CT scans of 10 non-diseased subjects Modeling the calcaneus using the template model, generated from the mean image of all binary masks Template surface sampling: 1.016~2.392mm (distance between points) Landmark assignment to template model Anterior, middle and posterior articular surface for the talus Experiments Evaluation of the template-to-volume registration Measuring the shape similarity between the template and the binary masks using volume overlap and distance metrics (mean and Hausdorff distances) Evaluation of the landmark mapping Comparing the template-based landmarks with manually assigned landmarks on individuals surface models 8
Experiments on Human Data - Calcaneus Study Accuracy of the template-to-volume registration Comparison of the template-based landmarks with manually assigned landmarks 9
Experiments on Human Data - Calcaneus Study CT Volume Rendering Surface Model Manual Landmark Template-based Landmark [Minimum distance btw manual and template-based landmarks] [Maximum distance btw manual and template-based landmarks] 10
Summary Template-based landmark mapping framework Using a template-to-volume registration finding the pairwise correspondence between the template and target volume Provide an accurate surface representation and the individualized positions of the anatomical landmarks Future directions Statistical analysis of the landmark correspondence across multiple subjects Application to large datasets including subjects with disorders of foot and ankle Correlation analysis between joint motion and shape changes 11
References & Acknowledgements References 1. Neogi, T., et al.: Magnetic resonance imaging-based three-dimensional bone shape of the knee predicts onset of knee osteoarthritis: data from the osteoarthritis initiative. Arthritis Rheum. 2013. 65:2048 2058. 2. Peeters, K., et al.: Alterated talar and navicular bone morphology is associated with pes planus deformity: A CT-scan study. J Orthop Res. 2012. 31:282 287. 3. Kim, J.,Park, J.: Organ Shape Modeling Based on the Laplacian Deformation Framework for Surface-Based Morphometry Studies. J Comp Sci Eng. 2012. 6:219-226 Acknowledgements This research was supported by Basic Science Research Program through the National Research Foundation of Korea (NRF) funded by the Ministry of Education, Science and Technology (No. 2011-0009761) 12