Home
Update this later. The simplified workflow as a schematic shown there:
The segmentation could be done for example with level sets which would give us implicit model (e.g. Ségonne and Fischl 2015) of the vessel shape possibly eliminating the need to do the volumetric (voxel-based) surface extraction using the Marching Cubes algorithm or some other similar algorithm (see for example Wiemann et al., 2015)
We can also think whether it is the best to register the stacks from different time points using point clouds or the meshed vessels. And whether we need the registration at all for quantitative analysis, but it could be nice for visualization of the BBB leakage for example. The meshed reconstruction could be then refined in Rhinoceros 3D or in Blender using texture and bump maps (can be maybe borrowed from ready-made vessel models)