| Dear SPManagers: I am realigning multiple SPECT scans within
| subjects, and am not clear what this option offers me. I have looked
| through help files relating to realign, and am not enlightened.
Weighting of the reference image allows the realignment parameters
to be estimated from only a specific region of the reference image. For
example, there may be artifacts in the images that you do not wish
to influence the estimation of the realignment parameters. By giving
these regions zero weight in the realignment, they have no influence
on the estimated parameters. The weighting procedure uses an image
which can contain zeros and ones, or more properly, it can be thought of
as containing the reciprocals of the standard deviation at each voxel
(unlike weighting in the spatial normalisation where the weight is
the reciprocal of the variance - I must fix this).
The function minimised in the realignment is something like:
\sum_i (wt_i * ( g_i - s * f_i ))^2
whereas for the spatial normalisation it is more like:
\sum_i wt_i * ( g_i - s * f_i )^2
|
| I do understand that the defualt for PEt and SPECT is to make a mean
| image from a "first-pass" realignment, and tehn realigns all images to
| that. Is this other option data-type specific?
It does this whenever you use the PET or SPECT modality. It would also
do this for fMRI, but I figured it would slow things down too much if
it did two passes. Also, PET and SPECT images are noisier, so realigning
to a mean image improves the results more. Ideally, the procedure would
be repeated a few times, but again, this would be too slow.
| 1. is there a number which corresponds to hard-coding "Create mean image
| only" for spm_realign?
I'm afraid the sptl_CrtWht variable only accepts the two values.
|
| 2. what is the range of possible values for regularisation? what numbers
| correspond to "medium" and "heavy"?
Any positive value you like. Medium and heavy are given by 0.01 and 0.1
respectively. A value of zero does not regularise the nonlinear part
of the spatial normalisation.
Best regards,
-John
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|