John,
Alle Meije Wink has been kind enough to consult on this problem, as well,
and he arrived at the same conclusion. Using the Mutual Information program
seems to have done it. That result proved quite good.
Thank you both for your time and help with this.
Best,
David Kareken
-----Original Message-----
From: John Ashburner [mailto:[log in to unmask]]
Sent: Thursday, December 14, 2000 5:49 AM
To: [log in to unmask]; [log in to unmask]
Subject: Re: PET-MRI coregistration problems
| I am having PET to MRI coregistration problems, and I wonder if someone
can
| help.
| The target is in my case a T1 MRI (SPGR), as I need the PET images (water)
| to be in the same space as the MRI. The PET images lack the dorsal part
of
| the brain due to field of view limitations. I specified the average water
| PET image as the object, and then tell it to apply the determined
| registration parameters to individual water images.
| The problem seems to be that the r*.img's (individual or mean) are not
even
| budged from their original orientation, let alone correctly registered on
| the MRI. Am I doing something wrong? Is SPM? I did find something in
the
| mailbase about the origin of the images, and none of the origins are 0, if
| that's relevant.
Which coregistration method are you using (mutual information or
the default method)? If you are using the default, then take a look
at the segmented images for clues. This one begins by affine registering
the images to the appropriate templates. If the images are not roughly
in register with the templates (within about 5cm), then this registration
may fail. You can identify a failure at this stage by the appearance of
some pretty wierd segmentation results. Also, use the <Check Reg> button
to see how well aligned the images are with the templates, and maybe give
some better starting estimates by approximately re-orienting the images
via the <Display> button.
If you are using the mutual information version, then take a look at the
scatterplots (histograms) to see if there are any wierd artifacts.
| Another curious thing is that the supposedly co-registered MRI's are twice
| their original size-- that also seems funny.
The MRI images should not change. With the MR as the target and PET as the
object, there should be no change at all to the MR images.
Bets regards,
-John
|