Towards Automating Patient- Specific Finite Element Model Development Kiran H. Shivanna 1,4, Brian D. Adams 2,1, Vincent A. Magnotta 3,1,4, Nicole M. Grosland 1,2,4 1 Department of Biomedical Engineering, 2 Department of Orthopaedics and Rehabilitation, 3 Department of Radiology, 4 Center for Computer-Aided Design The University of Iowa, Iowa City, IA

Finite Element Method Invaluable tool in musculoskeletal research Demands associated with modeling the geometrically complex structures of the human body often limit its utility – restricting analyses to baseline models Conventional meshing techniques often prove inadequate

Patient Specific Models In order to bring FE to the bedside for guiding surgical procedures the technique must be unencumbered from the image segmentation and mesh generation process Overcome the limitations associated with individualized, or patient-specific models

FE Model Development Acquire Medical Imaging Data Segment Regions of Interest Generate FE Mesh Apply Boundary/Load Conditions and Material Properties Finite Element Analysis Surface Generation

Tetrahedral Meshes Most commonly used solid meshing technique Several automated techniques for filling a surface based definition of a region of interest –Paving, advancing front, others –Advantages: well developed algorithms, straight forward to implement –Disadvantages: overly stiff elements

Voxel Based Meshing Techniques Direct conversion of CT data to hexahedral elements –Keyak et al –Advantages: easy to implement, voxel-wise material properties, fast –Disadvantages: stair step artifacts in mesh, not appropriate for contact analysis

Hexahedral Meshes Most commonly used meshing technique for surface contact analysis Few methods to generate the meshes –Shelling, whisker weaving, mapped mesh –Advantages: More appropriate for surface contact analysis –Disadvantages: Less well developed algorithms, prone to element shape problems, regional control of mesh density difficult

Objective Automate the generation of high quality hexahedral meshes –Projection method

Bones of Interest Why initiate with the bones of the hand? Long bones and cuboidal bones Number of bones per cadaveric specimen Readily extended to the other long bones of the body

Bones of Interest Extend to irregular bones such as the vertebrae

Image Analysis Cadaveric specimens were imaged with CT scans –Hand: Cadaveric specimen amputated above the elbow –Spine: Visible male dataset

Regions of Interest

Projection Method Carpal Bone Initial Bounding Box Bounding Box with Assigned Mesh Seeding Projected Mesh

Projection Method Example – Proximal Phalanx Bone

Extending Projection Method A single bounding box coupled with the projection technique may not always prove sufficient Method has been extended to add multiple boxes and/or subdivide existing boxes

Projection Method Multiple Boxes

Projection Method Movie

Solid Mesh Smoothing Projection of initial mesh onto the surface oftentimes yields distorted elements Need to smooth resulting mesh – Iterative Laplacian smoothing for solid mesh Method –Apply Laplacian smoothing to surface nodes holding interior nodes fixed –Project nodes back onto the original surface –Smooth interior nodes with surface nodes held fixed –Iterate for specified number of iterations or until convergence threshold is reached

Results of Mesh Smoothing Unsmoothed Smoothed Unsmoothed Smoothed

Multiple Bounding Boxes Spine

Acknowledgements Grant funding –R21 (EB001501) –R01 (EB005973) Nicole Kallemeyn, Nicole DeVries, Esther Gassman