search

translationalneuromodeling / tapas

TAPAS - Translational Algorithms for Psychiatry-Advancing Science
https://translationalneuromodeling.github.io/tapas/
GNU General Public License v3.0
216 stars 89 forks source link

PhysIO toolbox: PCA numbers and other parameters for aCompCor #195

Open gracehsu opened 2 years ago

gracehsu commented 2 years ago

Dear developers,

Thanks for developing this toolbox in the SPM batch. Currently I'm denosing my rsfMRI data using PhysIO toolbox.

I wonder if there's any suggestion for the number of PCA components for aCompCor using WM & CSF masks? (currently I'm using "5")

And I'm not sure what "Number of ROI pixels to be cropped" mean? (currently I'm using "0"). Also, do you have recommendation for ROI threshold?

In addition, I noticed that "fMRI Time Series File (s)" only allowed me to select one file (/volume?) As I assumed I should have selected the whole volumes of my rsfMRI images (e.g., 1:180)

Thank you in advance. :)

benoitberanger commented 2 years ago

Hi @gracehsu,

I wonder if there's any suggestion for the number of PCA components for aCompCor using WM & CSF masks? (currently I'm using "5")

Personally, I would not recommend a specific value, since this depends on your dataset. However, from my personal 5 PCs looks like a reasonable number of components. You can give it a try and visually inspect the components with the verbose option.

And I'm not sure what "Number of ROI pixels to be cropped" mean? (currently I'm using "0"). Also, do you have recommendation for ROI threshold?

Here the objective is to extract signal form compartment of non-interest. You must avoid including signal from the compartment of interest (GM). I recommend using high threshold above 90%, and at least 1 voxel for the 3D erosion "Number of ROI pixels to be cropped" I proposed in PR a function to visually inspect the effect of threshold + 3D erosion here. You can give it a try.

In addition, I noticed that "fMRI Time Series File (s)" only allowed me to select one file (/volume?) As I assumed I should have selected the whole volumes of my rsfMRI images (e.g., 1:180)

The file selector asks for a single 4D volume in "one line": path
/path/to/volume4D.nii
This selection does not follow SPM default file selection such as path
/path/to/volume4D.nii,1
/path/to/volume4D.nii,2
/path/to/volume4D.nii,3
gracehsu commented 2 years ago

Thank you so much for the reply, and the script is very helpful! thank you!

mrikasper commented 2 years ago

Thank you, @benoitberanger , for resolving the issue.

BTW: I haven't forgotten about the pull request, hopefully makes it into the September release :)

All the best, Lars