Thinning

[Manik]

Hi,

I would like to know if there is a command (or a function in Aims librairies) for the "thinning" of a 3D shape :
- I have a 3D GIS image that contains an object extracted from an MRI.
- I have a template of this object which is 1-voxel wide
- I want to get the thin version (1-voxel wide) of my extracted object to allow registration with the template.

I suppose that the software I am currently using for this task extracts the skeleton of the object.
I know there are morphomath functions in Aims libraries, so if there is no Aims command for this, could you please explain how they work, so I can make the command myself ?

Thanks in advance,

[Jean-Francois Mangin]

I am afraid our thining algorithms are only hidden in Vip Commandlines...
Moreover, thinning is not a very robust operation. You have to pay attention to the topology of the initial object. The details of its boundaries may change the result a lot, depending on the thinning method. Finally the shape of the object may influence robustness a lot. In 3D, a flat object will reliably yield a flat skeleton, while a cylindric or spherical object may yield non controlledresults. We have a collegue here who spent some time playing with this for caudate, using Aims Library (Renaud Maroy). You may be interested getting in touch with him, even if this remains a difficult problem.

[Manik]

Ok, but I suppose the problem is easier in my specific case, because I just want to get the skeleton of a stereotactic frame (two rectangles, one for each side of the head), not a complex object.
Thank you for your help.

[Jean-Francois Mangin]

You could explore the possibilities of VipSkeleton, If you do not mind using a commandline. I am not sure of the best parameters, however, because this command is dedicated to the cortical fold pipeline. The good options may be missing.

Try something like

VipSkeleton -i frame -v n -c n

If it does not work, I may help you but not next week, I am not at work.

PS You may also get some help from Serge who played with this stuff during is PhD.

[Manik]

Thanks a lot !

The results I get with

Code: Select all

VipSkeleton -i frame.ima -v n -c n -sk s 

look like the image I expected.
I still have to check that my other programs work correctly with the computed skeleton.