Hi John,
My primary purpose of knowing how the scaling works in SPM2 is to calculate
the sum (or mean) squared differences between the template and the finally
warped source. Could you suggest me how to obtain this number for each
normalization task?
Thanks a lot,
James Yu
-----Original Message-----
From: John Ashburner [mailto:[log in to unmask]]
Sent: Wednesday, February 08, 2006 3:31 AM
To: [log in to unmask]; [log in to unmask]
Cc: [log in to unmask]
Subject: Re: [SPM] Scaling in normalization
> Thanks for pointing out the lines 311, 160, and 163 (SPM5).
> I could understand the effects of lines 160 and 163 (global scaling).
>
> I have two more questions on line 311 (the expression s = T((s1+1):end):
> 1) what are these "4" scaling factors? Why not just one?
They model linear intensity gradients over the image. If the parameters are
s1, s2, s3 and s4, then the scaling of the template is by
g(xg,yg,zg)*(s1+xf*s2+yf*s3+zf*s4), where xg, yg and zg are the coordinates
in the template and xf, yf and zf are the corresponding coordinates in the
image that is matched to the template.
> 2) how to use these factors? Multiply or divide them to the source to
make
> it match to the target, or ...?
This is likely to be a bit tricky, and would require some additional coding
to
be done.
Best regards,
-John
> -----Original Message-----
> From: John Ashburner [mailto:[log in to unmask]]
> Sent: Monday, February 06, 2006 3:06 AM
> To: James Yu; [log in to unmask]
> Subject: Re: [SPM] Scaling in normalization
>
> > I read that when SPM2 does spatial normalization, it also searches for
> > optimal pixel value scaling factor. Is there anyway to know what this
> > final value is once the normalization is done?
>
> There are actually 4 scaling factors that are supposed to model linear
> intensity variability over the volume. You can get these from
> spm_normalise.m at around line 311 by including e.g:
>
> s = T((s1+1):end)
>
> and somehow saving the value of s. Note also that you'll need to include
> the
> scaling effects from around lines 160 and 163, which scales the
intensities
> to be within a more reasonable range.
>
> Best regards,
> -John
|