Dear Mohamed, I assume you co-register the Fallypride microPET images to the same animal MRI. Then normalize the MRI and carry along the PET images. This gives two stages to examine: How good is the co-registration of the PET/MR; and how good is the normalization of the animal's MR. If you have issues with the normalization, start by testing a simple normalization, and progressively try harder ones. First off, how does a linear normalization work, not allowing any non-linear iterations. I typically pick "Averaged sized template" for the Affine regularization. If you want to allow non-linear iterations, start with a fewer number, say 4, and work up. Also you can try setting the "Nonlinear Regularisation" to 10 or on bad cases even 100. I haven't work with the template by Schweinhardt. et. al.;we use a lab template of animals we collected. Our template and animal MRIs coming from the same camera is likely an advantage. Hope this helps. dave [log in to unmask] On Fri, Mar 20, 2009 at 4:30 AM, Bahri <[log in to unmask]> wrote: > Dear David, > > I have the same rat brain normalization problem as Raimo but with microPET > rat images. So I tried to normalize microPET Fallypride images to MRI > Template (Schweinhardt 2003) but unfortunately it does not work. Of course I > applied normalization after a manual co-registration and extraction of the > brain from the microPET image. I will be very glad if you have some advices > about this type of normalization. > > > Best regards, > > Mohamed > > > > David Wack a écrit : > >> My experience is with only structural rat MRI and microPET, but I >> think the basic process applies. I like to start by realigning using Check >> Reg->re-Orient until the images are close >> to my template, so that the brains are reasonably aligned and overlap the >> template >> fairly well. >> >> I follow this by (SPM) Realigning the images to the template. There is an >> advantage >> over human mri in that variability between rat brains is less, which >> allows this >> realignment process to work fairly well. The >> disadvantage is that there is a much higher percentage of voxels coming >> from >> outside the brain. If my hand realignment is reasonably close, I use a >> large mask >> as a weighting image in the realignment. My large mask comes from ROIs >> of >> the brain of the template, which I enlarged using Check Reg. >> >> This last step won't be perfect because the template and MRI are from >> different >> animals, but it should be good enough to get the MRI to fall underneath a >> tighter >> mask of the template brain. >> At this point you should be able to proceed with normalization. You could >> mask >> out non-brain, or specify the template mask as a weighting image for the >> template >> or rat MRIs. >> >> My basic approach is to take smaller steps than I would in processing >> human brains, >> verify constantly, and try to prevent far out solutions using brain masks >> as weighting images. >> >> The problem becomes harder and trickier the closer the nuisance activation >> is to the >> brain. Eliminating signal from around the eyes isn't too bad. >> >> Hope this helps some. >> >> >> dave >> [log in to unmask] >> >> >> >> On Thu, Mar 19, 2009 at 12:41 PM, Joensuu, Raimo < >> [log in to unmask] <mailto:[log in to unmask]>> >> wrote: >> >> Hello, >> >> I am doing fMRI with rats. My fMRI images contain a lot of signal >> from eyes and extracranial muscles and matching them to the >> segmented anatomical images (which contain only the brain) fails >> badly. I wonder if anybody anybody has succeeded in this type of >> normalization? >> >> Cheers, >> Raimo >> >> Raimo Joensuu, PhD >> Senior Research Scientist >> AstraZeneca R&D Molndal, DECS - Imaging >> S-43183 Mölndal, Sweden >> phone: +46 (0)31 706 5993 >> fax: +46 (0)31 776 3758 >> >> >> >