Author: Simone Poncioni, MSB Group

Date: 29.07.2022

Import a voxel-based model and convert it to a geometrical simplified representation through the use of splines for each slice in the transverse plane.

• Import the MetaImage file of the masked model that we want to convert;
• Using SimpleITK, extract the binary contour for each slice (in cortical bone, this contains inner+outer cortex)
• Extract only the outer contour using the OpenCV library and interpolate them over a numpy meshgrid
• First sorting: counterclockwise sorting of points (i, j) wrt CoM of the object
• Second sorting: sorting according to Mahalanobis distance (because it's more independent of the starting points than Euclidean distance)
• Representation of a B-spline of 3rd order over the transverse surface on the points (i, j)

• Faster implementation of sorting algorithm
• Faster implementation of cortical sanity check
• Implement test robustness over different models
• Understand how undersampling of raw data affects spline definition

• 2D algorithm: Quasi-Structured Quads (experimental)

• 3D algorithm: Delaunay

• 2D recombination algorithm: Simple

• Recombine all triangular meshes: FALSE

• Subdivision algorithm: None