[Insight-users] FEMRegistrationFilter Questions

D R Magee drm at comp.leeds.ac.uk
Tue Mar 7 06:40:00 EST 2006


Hi,

I've been trying to get FEMRegistrationFilter to work using mutual 
information. I'm starting with the DeformableRegistration1 program to 
start with (although I've also coded a version without the normalisation 
filters). My questions are:

1) Although the config file seems to work with uni-modal images using 
the "mean sq" metric I can't get anything to work with the mutual 
intormation (MI) metric. I've created some very simple 
synthetic multi-modal images at (for testing):

http://www.comp.leeds.ac.uk/drm/ref2.PNG
http://www.comp.leeds.ac.uk/drm/moving2.PNG

The config file I'm using is:

% ---------------------------------------------------------
% Parameters for the single- or multi-resolution techniques
% ---------------------------------------------------------
5 % Number of levels in the multi-res pyramid (1 = single-res)
5 % Highest level to use in the pyramid
16 16 % Scaling at lowest level of pyramid
4 4 4 4 4 % Number of pixels per element
1.e4 1.e4 1.e4 1e4 1e4 % Elasticity (E)
1.e4 1.e4 1.e4 1e4 1e4 % Density x capacity (RhoC)
1 1 1 1 1 % Image energy scaling (gamma) - sets gradient step size
2 2 2 2 2 % NumberOfIntegrationPoints
4 4 4 4 4 % WidthOfMetricRegion
30 30 30 30 30 % MaximumIterations
%  -------------------------------
% Parameters for the registration
% -------------------------------
3 0.99 % Similarity metric (0=mean sq, 1 = ncc, 2=pattern int, 3=MI, 5=demons)
1.0 % Alpha
1 % DescentDirection (1 = max, 0 = min)
2 % DoLineSearch (0=never, 1=always, 2=if needed)
1e1 % TimeStep
0.5 % Landmark variance
0 % Employ regridding / enforce diffeomorphism ( >= 1 -> true)
% ----------------------------------
% Information about the image inputs
% ----------------------------------
352 % Nx (image x dimension)
288 % Ny (image y dimension)
0 % Nz (image z dimension - not used if 2D)
..etc ...

Am I doing someting badly wrong? The deformation field calculated is near 
zero, which seems to result from the method stopping after only a couple 
of iterations at each step as the energy change is not monotonic.

2) What I really want is to be able to set E/RhoC on a per-element basis 
(so I can have fairly rigid and fairly non-rigid areas). Is this possible?


Thanks a lot in advance

Derek Magee


More information about the Insight-users mailing list