I see at least 8-9 methods for doing the correction, and papers that compare their merits, but none of these papers discusses the specifics of how they did it, only mathematical formulas and algorithms. My lab already uses FSL, Freesurfer and AFNI if anyone knows an approach that would work with minimal new coding. Thanks.
I think there is some partial volume correction method implemented in PMOD
Different PET vendors offer in their equipments different partial correction software. The basic idea is to meassure the point spread function (PSF) at different positions within the PET field of view. This set of PSF is then used at reconstruction (OSEM 3D) to perform partial volume correction. In SPECT there is a similar technique. By knowing the physical characteristics of your collimators and the distance between the source and the collimator at each projection one have a collection of PSF that can be used to recover image quality.
Thank you for your replies. PMOD is quite expensive ($2900 for a base license, add-ons for image analysis are extra). Alejandro, the approach you describe sounds like something that needs to be implemented during reconstruction of the images. Since I am analyzing already-acquired data (from ADNI), that's not an option for me; I would need to use a post-reconstruction approach, possibly taking advantage of available MRI data from the same subjects.
There are algorithms for PVC that are suitable for different situations and are easily coded. However, if you have Matlab and want minimal coding, there are toolboxes out there tailored to brain PET images. Try SFSRR, at http://www.noninvasive.med.tohoku.ac.jp/ShidaharaLab/SFSRR.html. They have some quite useful tools and documentation.
But I would caution you to consider your use of PVC - is it appropriate and essential? Of course, many ligands suffer with different groups, when there might be atrophy around small ROIs and so on, but the changes you make to the raw data are not always quantitatively the most accurate approach. Like many pre-processing steps, it should be considered in context and not blanket used.
James, thanks for pointing me toward that website. And I agree that it is not always important to do PVC. However, I am analyzing data on subjects with Alzheimer disease using an ROI approach to examine medial temporal lobe structures which are often very atrophic and adjacent to CSF.
Great, PVC essential then! SFSRR is quite widely used and internationally developed. I think it will be useful, at the very least a good jumping off point.
There is this toolbox that I use which has all the existing PVC methods and a combination of those.
Attached is the link to download it and the respective published article.
https://github.com/UCL/PETPVC
http://iopscience.iop.org/article/10.1088/0031-9155/61/22/7975/pdf
- Badges
- Science topic
- Similar topics
- Engineering
- Materials Engineering
- Powders