search

ramp-kits / autism

Data Challenge on Autism Spectrum Disorder detection
https://paris-saclay-cds.github.io/autism_challenge/
67 stars 43 forks source link

WM/CSF regression #49

Open makism opened 5 years ago

makism commented 5 years ago

Dear all,

First of all, thank you for your efforts and publishing the dataset. I am just wondering is whether or not the WM/CSF are regressed out. It's a pretty common practice for resting state fMRI if I'm not mistaken.

Thanks, Makis

ntraut commented 5 years ago

Hello,

Maybe @glemaitre can confirm but according to https://github.com/ramp-kits/autism/blob/master/preprocessing/extract_time_series.py, no confounding variable was regressed out. The only confounding variables that we give are motion values, but it seems that regressing out that kind of variables is not so important when using partial correlation or tangent matrices. We can effectively see that some of the ten best submissions did not use fMRI motion at all.

ntraut commented 5 years ago

In fact it seems that regression was done in https://github.com/ramp-kits/autism/blob/master/preprocessing/scripts_f1000/5_nuisance.sh, but I am still not sure it is really important in the case of partial correlation...