Hello,

I'm trying to analyze my rsfMRIs, and would like to have all outputs in the same space and resolution as my anatomical scans.
Specifically, the steps I'm following are:

1. Load the anatomical scans under Setup: Structural and the rs functional scans under Setup: Functional,
2. Go to Setup.Options and click the "Create confound-corrected time-series" as well as change the Analysis Space to "Volume: Same as structurals,"
3. Click on Preprocessing and select: "preprocessing pipeline for surface-based analyses (in subject space),"
4. Set the "Structurals target resolution" to the voxel resolution of my anatomical scans, and
5. When prompted to enter a number of diffusion steps for smoothing, I select the default value of 40.

I was wondering whether the above process is the best way to preprocess/denoise the rsfMRIs so that all outputs are in structural space and resolution. I'm mainly asking because I wasn't sure whether the surface-based approach was the best in my case since I'm not loading Freesurfer segmentations.

Thank you in advance,
Panos

Hi Panos,

I would suggest to modify the "preprocessing pipeline for surface-based analyses (in subject space)" to:

1) remove the three steps from "functional resampling..." to "functional smoothing..."
and 2) add a "functional smoothing (spatial convolution...)" step and place it third to last

The resulting pipeline should look like the attached image, and that would leave your functional data in the same space as your anatomical scans (naturally, these will be different for each subject so that will preclude any group-level voxel-based analyses), skipping those steps in the original pipeline involved in projecting your functional data to FreeSurfer cortical surface.  

Hope this helps
Alfonso
Originally posted by Panos Fotiadis:

Thanks Alfonso, that's extremely helpful!

Best,
Panos

Hi Alfonso,

In addition to the T1 anatomical space I mentioned earlier in this thread, I also tried to perform the fMRI processing analysis in the subject's native diffusion space. In order to do that, I registered my original T1 scan into the subject's diffusion b0 space and then used that as the structural scan in the CONN pipeline. The pipeline I ran was the modified "preprocessing pipeline for surface-based analyses (in subject space)" you kindly suggested in the previous message of this chain (I additionally changed the target structural voxel-size in the preprocessing setup to the diffusion scan's voxel size rather than the default 1mm.)

I had two follow-up questions:

1. Does that methodology sound correct if I want to perform the processing on diffusion space, or is there a more "appropriate" way to do it? After running it for one subject the results seem to make sense (including the segmentations), and the final denoised volume seems to be correctly registered to diffusion space!

2. I would like to assess functional connectivity (i.e., Pearson correlations between BOLD timeseries) between all given voxels within a cortical mask I have (this mask is in diffusion space and each voxel within that mask is assigned a different integer value - in that sense it's not really a binary mask, although I could convert it to one if needed).

I can think of a tedious way of doing that by matching those specific voxels to their equivalent rows within the vvPC_Subject*_Condition*.matc file and then calculating the Pearson correlation, but I was wondering whether there was a more efficient way (such as for instance by adding this mask as an ROI under Setup and then have a matrix file be exported).

Thanks in advance for all your help!

Best,
Panos

Just wanted to re-circulate this. Thanks in advance!

Best,
Panos

Just wanted to re-circulate this. Thanks in advance for your time and help!

Just wanted to re-circulate this!

Thanks in advance,
Panos

Hi Panos,

Regarding (1) yes, that looks perfectly fine as well, I see no problems at all with the approach and that will allow you to perform your analyses in (subject-specific) native diffusion space. 

Regarding (2) again yes, if you "mask" file contains is named mymask.nii and it contains values between 1 and N, simply create an associated file named mymask.txt (you may either leave that text file empty or enter in it N lines, each line containing a label for that individual voxel ;e.g. its coordinates) and save this file in the same folder as your original mymask.nii file. Then, when you enter that mymask.nii file as a new ROI into CONN, it will automatically be interpreted as an atlas file, so you will be able to use the standard ROI-to-ROI analyses in CONN to compute your voxel-to-voxel correlation matrices for each subject (of course, assuming that this mask is also subject-specific remember to also check the 'Subject-specific ROI' checkbox in Setup.ROIs so you may import all of your different mask files simultaneously)

Hope this helps
Alfonso
Originally posted by Panos Fotiadis: