The coarse resolution bone_region mask was OK for gathering the statistics (2D-histogramming and optimizing the probability models).
However, it is unsufficiently accurate near the implant. For two reasons:
The coarse voxels make for an inaccurate interface to the implant.
The morphological closing (for including blood vessels in the region) also closed off the valleys from the implant threads: important regions for e.g. bone-to-implant contact that now can't be properly segmented!
Further: There are problems near the cylinder edge as well.
In conclusion: We need a better bone_region mask with as high resolution and accuracy as possible near the implant threads, and which covers the entire bone region.
The coarse resolution bone_region mask was OK for gathering the statistics (2D-histogramming and optimizing the probability models).
However, it is unsufficiently accurate near the implant. For two reasons:
Further: There are problems near the cylinder edge as well.
In conclusion: We need a better bone_region mask with as high resolution and accuracy as possible near the implant threads, and which covers the entire bone region.