> We are evaluating and comparing the stereotactic normalization of SMP99b
> to another tool. We want to use a FDG PET image for a template. This
> template image consists of
> matrix size: 128*128*60
> voxel size: 2.25mm*2.25mm*2.25mm
> Origin: 64 71 28.
>
> We have two questions:
> 1. Which parameters would you recommend for an optimal stereotactic
> normalization with FDG image and FDG template?
> Nonlinear Basis Functions: 7*8*7?
Ideally, you would use as many basis functions as possible, but with many
more than 7*8*7, you will may begin to have memory problems.
> Nonlinear Iterations: 12?
Again, the more iterations the better, although generally after about 12
iterations, the algorithm does not do much to the parameter estimates.
> Nonliear Regularization: Medium?
I don't know. The selection of the optimum amount of regularization
depends on our a priori knowledge of the amount of brain variability.
Pragmatically, medium regularization appears to give reasonable results.
> Which interpolation,bilinear or sinc?
This makes no difference for the parameter estimation. The choice of
interpolation method only influences the writing of the normalized
images. Sinc is better, but it is very slow.
One thing that may be worth experimenting with is the brain masking.
This can give better results for T1 weighted MR images where the signal
from the scalp and other none-brain tissue is quite strong. However,
for images where there is very little none-brain signal, then you will
get better results without the brain masking.
>
> 2.Which output parameters should be used to get the wanted matrix size?
> We want the preserve the matrix size to that of the template. We changed
> the output defaults of the bounding box, but still got the following
> dimensions:114*113*54. How should the ouput default of the bounding box
> be defined to get the dimensions 128*128*60?
The computations are:
[2.25*[(1-64) (128-64)]
2.25*[(1-71) (128-71)]
2.25*[(1-28) ( 60-28)]]
which give:
-141.7500 144.0000
-157.5000 128.2500
-60.7500 72.0000
Regards,
-John
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|