Hi all,
I am a researcher exploring the possibility of applying functional connectivity analysis to some FDG PET data. This data is static (or sum over the 4D dynamic image), one volume per subject, and each subject has a 30-month follow-up session. Despite the various tools designed for connectivity analysis (Conn, network-based statistics, Graphvar, Nilearn, etc.), all are built to aim for fMRI/rsfMRI studies. The problem when trying to adapt these toolboxes to my PET data is that they require analysis of the time-series signal (BOLD, in the case of fMRI/rsfMRI).
The Conn toolbox has a step to denoise the images according to covariables generated for each study from quality assessments of the 4D alignment/motion, HRF model (I think?), and PCA 5 components of the WM and CSF. GraphVar expects to work on time signals from a .mat file. I haven’t explored the in-depth network-based statistic (NBS) toolbox or Nilearn yet, but they are also explicitly built for analyzing fMRI.
So my doubts are the following:
- If I expect to find a condition-specific network for both of the sessions acquired independently and jointed, should I first consider the subsets of my subjects related to that condition on each sessions as a time series and create the first level analysis, considering a subject in this case as the sample that represents a condition, to form the network that represents the condition and then apply a second-level analysis for a cross-sectional study?, and when consider the both sessions jointed (longitudinal analysis) should I form the time-series signal of FDG as the two scans from both sessions for each subject and generate the network on a individual level for then apply the second analysis to see the differences on both conditions?
I plan my methodology this way because we are interested on both cross-sectional study (on both sessions) and a longitudinal study.
- Also, Is it fine to try to run this analysis on these fMRI-oriented toolboxes? or do these toolboxes have a more specific model design that won’t let me analyze the PET images without modelling them as fMRI? Should I start from scratch and design all the pipeline without a framework?.
Thank you, any help is welcome!!,