Geometry Reconstruction from MicroCT Images
Previously acquired micro-computed tomographic (microCT) images of the scaffold (Liu et al., 2018) (Fig. 2A) were used to generate computational models (Fig. 2B). Image data were imported into Simpleware ScanIP (Synopsys, Inc.) for segmentation and directly transformed into a volumetric mesh containing both solid and fluid domains. Images were filtered to remove noise and smoothed by applying a mean filter (filter radius = 1 pixel). Grayscale stacked images were then segmented to separate solid from the air (threshold value taken from our previous model (Liu et al., 2018)). The scaffold image data were presented as a voxel model after the segmentation. Next, a voxel model of a cylinder was created to trim the scaffold into a smaller domain (1.2 mm thickness, 2 mm diameter) for shorter computation. Boolean operation intersection between the cylinder and scaffold voxel model was used to create the final solid domain (Foley, 1996). Boolean operation subtraction was used to generate the fluid domain from the image (scaffold) voids (Fig. 2C). Finally, the two domains were meshed together by replacing existing voxels directly to volumetric tetrahedral elements. The final mesh contains both solid and fluid domain with watertight assembly and interfaces with shared nodes (Fig. 3A).