Question Posted 06/03/14:
Sound procedure to normalize processed PET scans to MNI space. Dear ADNI experts, we are interested in normalizing LONI PET scans in the most processed format (Co-reg, Avg, Std Img andVox Siz, Uniform Resolution) to MNI space. Our proposal initially started by scaling the PET intensity values with respect to the mean activity in the pons region, then normalizing it to MNI using the standard PET MNI template and finally smoothing it with a gaussian kernel of 12x12x12. However, after reading the acquisition protocols, since the most processed PET scan has already been smoothed and scaled in intensity, we are not sure if we should smooth them with a smaller kernel (maybe 8x8x8?) or scale it with respect to pons in the first place. What do you experts consider to be the most robust protocol to normalize processed LONI PET scans? Thanks a lot in advance.
Response posted 06/03/14 by Bob Koeppe:
The conversion into MNI space is independent of the intensity normalization (scaling) of the images. The processed images you refer to for FDG are globally normalized, and hence for final analysis should be renormalized to a region that on average changes the least, such as pons as you suggest, or cerebellar vermis. The images in this set have been smooth to a common resolution of 8x8x8 mm FWHM. I think additional smoothing of 12x12x12 (mm FWHM?) is far to much additional smoothing. In you last question, I am not sure what you mean by "normalize". Is this intensity, spatial, or both?