[Insight-users] branch labeling 3D tree structure

[Olivier Sebastien Jerome Cros / Region Nordjylland]

Hello,

I am in an hospital in the ENT department, more specifically ear surgery. I am currently working on the extraction of the air cells in the mastoid bone (right behind the ear). I am currently investigating the tree structure of the air cells.

I have a very high-resolution Dicom CT scan of the temporal bone, 512*512*110 (pixel spacing:0.296825*0.296825, slice thickness: 0.625).

After binarization of the data and cropping the region of interest (both using ImageJ), it has been processed with the newly implemented 3D thinning algorithm (available at: http://www.insight-journal.org/midas/view_item.php?itemid=1943). After visualization, the result seems correct, but I am not sure.

So far I get satisfactory results. But I am hitting against the wall, while finding a solution about counting the number of splitting, the number of branches, etc.

My best idea so far is to run a kernel of 27 voxels and detects for each voxels belonging to my skeleton if there is a neighboor having only one vertex in common, only one edge in common or a face in common.

My question remains on how should I do it. Is it related to iterators? maybe neighbours iterators?

Could anyone enlight me on this problem,

Thank you in advance,

Olivier Cros.