Question

Question Posted 06/11/14:
Continuation to: "Sound procedure to normalize LONI PET scans". Dear ADNI experts, thanks so much for answering my previous question. However, I'm still wondering whether I should do any smoothing after transforming the processed PET scans (Co-reg, Avg, Std Img andVox Siz, Uniform Resolution) to MNI space.

My proposal for a robust procedure to run group analysis from processed LONI PET scans is the following: First, scale the intensity of the scans dividing by its mean pons value. Second, spatially normalize them to MNI space. Finally, should we smooth them afterwards? if so, which kernel do you recommend? Thanks a lot in advance.
Response posted 06/11/14 by Bob Koeppe:
I don't believe there is any right or wrong answer to your question. The answer may depend on the scientific question you are asking. For voxel-wise analysis, maybe more smoothing would be beneficial. For larger analysis from predefined atlas regions, I would expect additional smoothing will not be a good idea. I would start with the image sets that are already smoothed to 8 mm, and go from there.
I would probably start without any additional smoothing, and go from there.
Response posted 06/11/14 by Bill Jagust:
I suggest you download the data on which this has all been done already. My recommendation is to smooth all PET data so as to achieve a common resolution. This is a fairly complex process, but it has already been done. If you download the most highly processed ADNI data you will have images that a re all in the same orientation and resolution