 |
 |
Dear Buyean, I totally agree with what David Wack wrote - misalignment between transmission and emission can be a major problem. I also agree co-registration will be better suited than realignment. You don't say which tracer you work with, but it must be one with slow kinetics seeing your acquisition timing :-) . If you use a "cortical" tracer with lots of signal in the outer parts of the brain, your proposed technique might already work "as is". Particularly if you're using a "subcortical" tracer, however, where signal is more or less confined to the basal ganglia (or the limbic system,...), or if you use a "cortical" tracer with heterogenous kinetics, your late frames are likely to look so different from early frames that even mutual information co-registration may run into problems. Two strategies may help to correct post hoc (assuming you do not have independent movement detection data e.g. from Polaris or other systems): 1. Improving your signal through wavelet denoising - see e.g. Hammers A, Asselin M-C, Turkheimer FE, Hinz R, Osman S, Hotton G, Brooks DJ, Duncan JS, Koepp MJ. Balancing bias, reliability, noise properties and the need for parametric maps in quantitative ligand PET: [11C]diprenorphine test-retest data. Neuroimage 2007, 38(1):82-94. (You correctly wondered about what happens to actual values after co-registration and, therefore, necessarily reslicing the dynamic images. Note a reviewer made us take out the detailed results of the movement correction procedure, but importantly, the parametric values obtained after applying the model did not change in those subjects without substantial movements - this is now briefly mentioned in Methods, p. 85 col.I). 2. Getting a reliable "outer ring" through using non-attenuation corrected images - see e.g. Montgomery, A.J., Thielemans, K., Mehta, M.A., Turkheimer, F., Mustafovic, S., Grasby, P.M., 2006. Correction of head movement on PET studies: comparison of methods. J. Nucl. Med. 47, 1936-1944. - or a combination of the two. Sorry about only quoting work from our place - you'll find references to others' important contributions in the papers cited. Good luck and I hope this helps, Alexander ________________________________ From: SPM (Statistical Parametric Mapping) [mailto:[log in to unmask]] On Behalf Of David Wack Sent: 01 April 2008 21:13 To: [log in to unmask] Subject: Re: [SPM] Can one use coregistration to correct head motion in PET? Realignment routines typically expect the images to be more less the same except they may be slightly out of alignment. What you likely have with the dynamic PET scans is that the individual frames don't look that much alike (even if they were perfectly aligned). This is due to how the tracer is being taken up. For this reason using the Coregistration routine makes some sense. Of course if you do have substantial motion you'll have to deal with the fact that your transmission scan wasn't aligned with at least some of your frames. dave Buyean Lee wrote: Dear SPM users, I would like to know if one can use Coregistration (normalized mutual information) to register PET frames to each other (also called, head motion correction). My concern is that if Coregistration can introduce artifact to the resliced image (or even after only estimation). I know that realignment uses only translation and rotation, which would not change the values in the image except the changes due to interpolation. Thank you, Buyean For someone who is interested in specifics, I will describe the PET procedure and problem in detail below. This is the PET acquisition protocol that we use. 1. 1st transmission scan for attenuation correction 2. Inject the isotope 3. 1st emission scan for 80 minutes 4. 20-minute break (the subject is removed from the scanner bed). 5. Reposition the subject to the scanner bed 6. 2nd transmission scan for attenuation correction for the 2nd emission scan 7. 2nd emission scan Basically, we acquire two scans, even though we are acquiring only one dynamic scan. So far, I have been using 'Realignment' to correct head motion (after making the two scans into one scan); the middle frame (17th) was used as the first image for realignment. By the way, the 17th frame (image) belongs to the first emission scan. Occasionally, I noticed that registration by 'realignment procedure' is not perfect. Specifically, SPM5 fails to realign the frames (images) from the 2nd emission to the first image (17th frame). When I used coregistration (default, normalized mutual information), I got better registration. ________________________________ Planning your summer road trip? Check out AOL Travel Guides <http://travel.aol.com/travel-guide/united-states?ncid=aoltrv00030000000 015> .