[vtkusers] 3D reconstruction from CT: pipeline and its options.

[Chen, Elvis]

Greetings,

For my learning of vtk/python, I have decided to write
a "mesher" that would reconstruct mesh of bones from
CTs.  Based on Sebastine Barre's email from VTK
mailing archive, I constructed the following pipeline:

	 - Image/Volume16Reader
         - MarchingCubes
		- SetValue( 0, 1125 )
         - CleanPolyData
         - [PolyDataConnectivityFilter]
		- SetExtractionModeToLargestRegion()
         - Decimate[Pro]
		- PreserveTopologyOn()

Since I intend to use the resulting mesh for
image-guided surgery, where surface-based registration
is used, I would like to preserve as much original
data as possible, while getting rid of
generated-surfaces that is "inside" of cortical bones.

I'm using VTK 4.0 (Release Candidate) and Python 2.2

Problems I have found so far:
1) for some reasons, 1 CT out of every 9 or 10 is
displaced along x-axis.  I have checked the original
CT data and made sure that it wasn't present in the
CT.

2) the final mesh is distorted, as if one of x (or y)
axis is scaled down.  I have set SetDataSpacing(
0.390625, 0.390625, 2.5) in vtkVolume16Reader where
the dimensions are obtained from dicom2.

3) even if I set SetExtractionModeToLargestRegion() in
vtkPolyDataConnectivityFilter(), I still end up with
excessive polygons that is disconnected from the outer
(largest) mesh.  Is there anyway to get rid of them?

4) any suggestion on the pipeline, what options to
use, is very much appreciated.

Thank you very much!

______________________________________________________________________ 
Web-hosting solutions for home and business! http://website.yahoo.ca