I have resting state fMRI data from three groups (control, psychiatric condition1, psychiatric condition2).
I do have individual fisher transformed z-scores from my seed region to every voxel of the brain (.tsv) from Nilearn.
I would like to statistically compare the groups on the brain level.
Which extension or python module can I use to (i) test the group differences & (ii) visualize the results? Will it come down to ANOVAs and/or a package where I simply feed the .tsv files?
Apparently, researchers analyze data with AFNI, SPM, or another program, however, I would like to stick to Python and not use an extension with Docker for example.
Already MUCH appreciated for anyone willing to chip in a thought or two!
You can use the nilearn GLM module to perform seed-base functional connectivity analysis, given that all relevant preprocessing has been done.
The first thing I would recommend is to read this nilearn tutorial to understand the general workflow of GLM-based fMRI statistical analysis better.
For the specific steps described here, you can look up the example on second level analysis to understand how to do relevant analysis on group differences. To understand how to specify GLM models, FSL documentation is a good place to look up relevant designs.
Nilearn also provides an extensive plotting function.
The Nilearn Tutorials unfortunately do not include examples of resting-state data i.e. data which cannot be used to compute contrasts (as there was no task to perform, other than resting).
The data is also not paired i.e. NOT age or sex matched. The only factor I have is ‘Group’ e.g. control or psychiatric condition1. I do understand that I need the time-series of my seed (which I have) with the respective R or z-values to every other voxel on the subject-level.
Would you be willing to get more specific on how to approach this problem with nilearn? Will I need to average z-maps per group and then compare mean differences between those three maps? This to me doesn’t seem to be sound.
I might not be seeing the way to do the analysis of resting state data with nilearn’s first and second level GLMs.
I am aware that there are no out of the box example of resting state functional connectivity analysis in NiLearn. There is a way because I wrote a wrapper for my own use. It still works but I don’t maintain it anymore. Most importantly, this project will not help you understand how the analysis is done. I would still suggest to go through some basic GLM tutorials on fMRI data. The principle of the analysis you are asking for is the same as task data.
Nilearn has an example data set that is close to resting state data, commonly referred as the developmental data set. This is a great place to start and write your own code for seed based functional connectivity.
Hi,
Second Level analyses are not tied to a certain type of neuorimaging contrast: seed-based connectivity, as long as it can be represented as a spatial map, can be used in second-levle analyses as well as other contrasts.
At least this is perfectly feasible with Nilearn. What I don’t understand is why you store z-values in tsv files: these values would be better represented as Nifti1Images, which you can then vizualize compare etc.
Best,
Bertrand