> Infact we notice that we are using all in a no property way.
> Our objective is to be able to distinguish a normal patient by an other
> ill using the information that come from SPECT. in a first time we had
> convert the DICOM images (that give us the hospital) in .hdr, .img using
> the programm MRIcro. then we went on with SPM99 with the phase of realign,
> etc.
> in an our discussion today, we notice that our procedure is (suppose)
> uncorrect, infact we realign for a single patient but we give (for example)
> 7 images(7 complete images, every one obtained with many scans), this
> because we are not able to select the scans that compose the image (really
> we notice that we obtain the matrix and also the results, but the images
> that come from the normalization are not compatible with the model)
> observed the web manual we notice that on the our images there isn't the
> little white preceding number, that have to indicate the total number of
> scans. what can we do?
> is a problem with MRIcro, or what?
The small white number preceeding the scans is just the number of files
that have the pattern (e.g.) nrfM01770_0*.img . When the number is clicked,
it just changes the Filter from *.img to nrfM01770_0*.img, and is intended to
make file selection easier.
Each file is typically a volume, made up by putting several slices together
in the same file. I assume that when you say an image is made up of many
scans, you mean this.
If you have a single volume for each subject, then you shouldn't realign
them, as the realignment is intended for registering together several volumes
from the same subject. This is typical for fMRI or PET activation studies,
but un-necessary for studies where there is just one scan per subject.
The pre-processing you need probably involves spatially normalising the data,
and then smoothing it, before entering it into the statistical analysis.
Where you have only one scan per subject, it is probably better to use a
single subject analysis. I realise that you aren't using single subject
data, but the design matrix generated will be much more applicable to the
kind of study you are doing. If you enter the data as two conditions, then
you should be able to compare patients and controls.
there is one more thing that you may need to watch out for. I don't know
what the field of view of your scanner is like, but it may not cover the
whole brain. Between spatially normalising the data, and smoothing it, you
should probably also include a masking procedure. Type the following in
Matlab:
spm_mask
Then select all your spatially normalised volumes to define the mask from,
and select them again as the images to apply the mask to. This cuts off
brain regions that are not present in all subjects (because of limited FOV,
and possibly different positioning in the scanner). Having the same brain
regions included for all subjects will reduce certain artifacts that you may
otherwise see in the results.
Best regards,
-John
--
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
|