On 7/25/18 8:31 AM, John Anderson wrote:
External Email - Use Caution
Dear Petsurfer experts, I used partsurfer in FS6 and had applied partial volume correction (PVC) on PET data. I followed the same steps as in wiki which was very well explained! Thank you!. The analyses went fine... I would like to inquire about some issues, and I highly appreciate any clarification:
- The flag --psf in "mri_gtmpvc": Is the value of "psf" means the
amount of smoothing (fwhm in mm) that will be applied on the corrected maps? If I give "psf" a value of "zero", does this means that the corrected maps are no longer going to be smoothed? The motivation of the question is that I would like to smooth the images at the end of the analysis, so I can apply the amount of smoothing on the PVC images vs analysis involve non-pvc images.
It has nothing to do with applied smoothing but rather the amount of blurring incurred during the PET acquisition and reconstruction process. It models the blurring function as an isotropic Gaussian filter of the given FWHM. Any smoothing you apply will be applied later. Note that applying volume smoothing to PV-corrected data is invalid. Surface smoothing is the only valid way to smooth PVCed data.
- For some of my group analyses, I don't see any difference between
the groups. When I apply PVC using "mri_gtmpvc", I see difference in pathologically relevant regions, and vice versa for some other analyses where I see difference between the groups without PVC then the difference disappear following PVC. I highly appreciate if you please clarify why PVC can find results or eliminate findings. What is the idea of partial volume effect PVE in this case?
This is a big topic. See https://www.ncbi.nlm.nih.gov/pubmed/26915497. That paper shows that atrophy can make the apparent effect of age on FDG be bigger than it really is. This is because PVEs reduce the FDG signal in GM which tends to be the same direction as the actual effect (ie, the PVEs reinforce what is already there making it too big and too significant). For another tracer where one expects the uptake to go down with age or disease (eg, PiB), then PVEs can make effects go away.
- Group analysis: following PVC, I used "mri_concat" to combine the
images "?.mgx.gm.nii.gz", then I smoothed the final combined images using the flag "fwhm" in "mri_surf2surf". I used fwhm=3mm and then fwhm=6mm. For all analyses I can see clusters of difference between the groups at fwhm=3mm. For some analyses the difference disappears when I use fwhm=6mm . Is this till a partial volume effect even after applying pvc or this is relevant to an extra smoothing effect which obscure the PET signal? is there any recommended fwhm value for PET images?
The FWHM you chose to smooth by for analysis purposes is somewhat arbitrary. The ideal FWHM depends on the size of the activation you are expecting, which you don't know until you do the analysis, which you can't do until you set the FWHM. If you have some idea about the size of the effect you expect, then you can set the FWHM to that level. Otherwise, people just fall back to typical values between 5-10.
Thank you for your guidance, John
Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer