Extract features for classification

Hi,

Im trying to make classification model for disease classification using rs-fMRI data.

Im know that many studies use fALFF, Reho, Functional Connectivity as features.

I want to know how can I extract such features.

I tried to make functional connectivity using Nilearn package and I did.

But some research papers are saying that preforming ICA might provide better result.

This part is quiet tricky for me.

As far I as know, ICA gives info about which regions have similar BOLD signal patterns. Isn’t that information already included in functional Connectivity?

To de concluded, I jus wanna know whether should I perform ICA and make functional connectivity or just skip perform ICA.

Best,
Junyong Oh.

Hi @dhwnsdyd21,

These are all output by default by XCP_D.

ICA is used to generate individual-specific functional networks, without the need to specify seeds and ROIs beforehand.

If personal network topography is important to your goals, than you can use ICA or similar techniques. Otherwise if you are okay with using a common atlas, XCP will do that for you.

Best,
Steven

1 Like

Will I might be able to use XCP_D without preforming surface reconstruction in preprocessing step?

Yes it works on niftis too.