Dear John:
Thanks again. Since I am new to SPM99, and do not know Matlab programming,
could you please tell me where should I add these scripts?
Jiansong
-----Original Message-----
From: John Ashburner [mailto:[log in to unmask]]
Sent: Tuesday, March 04, 2003 10:31 AM
To: [log in to unmask]
Subject: Re: normalization statistical images
> Thanks for you response. But, I still do not quite understand your answer
> for the second question. Since these con, spmT images already have been
> created and normalized, how can I use "Result" function again? To my
> understanding, if I am going to use the "Result" function again, I will
> create new non-normalized con and spmT images. But what I need is to
> overlay the normalized con and spmT images on anatomical image. Further
> suggestions are highly appreciated.
OK, so it won't work in your particular case. The Results won't be able
to use the warped results images. If you want to work with spatially
normalised results, then you need to have done the stats on the
spatially normalised fMRI.
Instead, you could try:
BV = spm_vol(spm_get(1,'*.IMAGE','Select background image'));
FV = spm_vol(spm_get(1,'*.IMAGE','Select foreround image'));
spm_check_registration(BV);
spm_orthviews('addcolouredimage',1,FV,[1 0 0]);
The foreground image should be thresholded first. You could use ImCalc
for this, with an expression such as i1.*(i1>3)
It might work.
Best regards,
-John
> > I am using spm99 analyzing fMRI data. For some reason, I did not do
> > normalization immediately after realignment, instead, after I did the
> > "Result", I normalized the beta, con, spmT images to a EPI T2 template
> > created by myself. But, when I compare the image before normalization
and
> > after normalization, some of them looks really weird, even lost the
shape
> > of "brain image". So, my question is:
> >
> > 1. Can I use SPM99 normalize function to normalize these beta, con,
> > and spmT images?
>
> You can, but you should not use sinc interpolation as this will cause a
> lot of the images to be severely eroded. The beta, con and spmT images
> contain NaNs (the computers way of representing "Not a Number", which
> is the result of dividing zero by zero - i.e. completely unknown) outside
> the masked region. If any voxel under the sinc kernel used for
> interpolation
> is NaN, then the output will be NaN.
>
> In SPM2b, I would suggets not using B-spline interpolation for the same
> reason,
> although with B-spline interpolation, the artifacts are even more severe
> (i.e.
> all voxels in the resampled images are likely to be NaN).
>
> If you use trilinear interpolation, you only lose a shell of voxels around
> the
> mask. As the images have been realigned and resliced (probably with sinc
> interpolation), then you don't need to use a high degree interpolation
> method
> at this stage.
>
> I may be wrong about the reason for the wierd results. It is always
> possible
> that there is some other explanation.
>
> > 2. If it is ok, how can I overlay these statistical images on high
> > resolution functional images?
>
> Images can be spatially normalised to whatever resolution you want.
> Therefore
> you can write your spatially normalised images with e.g. a resolution of
> 1mm.
> The various overlay routines in the results allow you to overlay your low
> resolution fmri results on to a higher resolution image, in much the same
> way
> as the Check-reg button allows data of different voxel-sizes to be shown
> together.
--
Dr John Ashburner.
Functional Imaging Lab., 12 Queen Square, London WC1N 3BG, UK.
tel: +44 (0)20 78337491 or +44 (0)20 78373611 x4381
fax: +44 (0)20 78131420 http://www.fil.ion.ucl.ac.uk/~john
|