feat_native procedure

gitdir = "/data00/layerfMRI/Github_repo/layerfMRI/"
bd = gitdir + /analyses/depth_native/02_feat_native/

EV preparation

The summary of all log files is in log.summary.csv (prepared by Lorenzo). This table is parsed with do_EV_native_preparation.py to derive one EV in fsl format for each sub/session/task/run/movietype.

python do_EV_native_preparation.py

The EVs are saved in the EV_predictors directory. Since the process had already been run initially for the PPI analysis, I also wrote a little script that carries out a check to verify that the EVs written now and then are identical.

$PWD/compare_EV.sh

Prepare FEAT .fsf onerun

To carry out the FEAT for all subs and run, I build a template using the sub_02_task_1_run_1. This will be then dynamically modified in sed for every sub/run. Finally feat will be run in parallel on all the .fsf files.

First I manually create an empty directory 000_subj_level_feat.

mkdir 000_subj_level_feat

Then we need to work in the Feat gui in x2go to create a sample fsf that we will save in 000_template_fsf_onesubrun as sub_02_task_1_run_1.fsf.

Data

It is important to correctly specify the following parameters in the gui:

output dir : bd + /000_subj_level_feat/ + /sub_02_task_1_run_1.feat/
fmri data : /data00/layerfMRI/regdata/sub_02/ses_01/func/task_1_run_1_4D

NB: the fmri data is task_1_run_1_4D. Be very careful not to use the MNI counterpart.

Note that the fmri data comes from the original directory. There is no need to duplicate it here.

Pre-stats

Since we are using raw data, there is also a minimal preprocessing to carry out. Basically motion correction (MCFLIRT) and highpass filtering. Make sure you set:

Slice timing correction : None
BET brain extraction : unchecked
Spatial smooting FWHM (mm) : 0.0

Registration

Uncheck everything: we don’t want to run any registration (we suffered enough with ANTs for this)

Stats

There are two EVs: one for M(otion), the other for S(crambled) (and their temporal derivatives). They are linked to the following EV_predictors

EV Motion : bd + EV_predictors/sub_02_EV_task_1_run_1_M.txt
EV Scrambled : bd + EV_predictors/sub_02_EV_task_1_run_1_S.txt

The Contrast matrix contains 4 contrasts:

Title	EV1	EV2
MOTION	1	0
SCRAMBLED	0	1
M > S	1	-1
S > M	-1	1

Post-stats

We want to run the analyses - and resolve the MCP problem - only within the cortex.

To do this, we use the sub/taskrun-specific cortical depth maps we previously created, which are already stored in:

/data_native/sub_02/depth/sub_02_depth_task_1_run_1.nii.gz

So we point the pre-thresholding masking to this file.

Prepare FEAT template FSF

We cp the single-subject fsf to a template_design.FSF

cp 000_template_fsf_onesubrun/sub_02_task_1_run_1.fsf \
   000_template_design.FSF

and replace the actual values in the first column with the placeholder in the second column:

`sub_02_task_1_run_1.fsf`	`000_template_design.FSF`
`set fmri(outputdir)`	`"OUTPUTFEATDIR"`
`set fmri(npts)`	`NUMBATIMEPOINTS`
`set fmri(totalVoxels)`	`TOTALNUMBAVOXELS`
`set feat_files(1)`	`"NII4D"`
`set fmri(custom1)`	`"EVMOTION"`
`set fmri(custom2)`	`"EVSCRAMBLED"`
`set fmri(threshmask)`	`"DEPTHMAP"`

NB: careful to the “double quotes” for files placeholders.

Running feat

The feat for each subj/taskrun is carried out by the 000_run_1st_level.sh executable bash script. The script does the following:

grab a list of subjects in /data00/layerfMRI/list_subjects
grab the 000_template_design.FSF and sed-replace the placeholders in it with the appropriate files/values for each sub/taskrun, then writes the fsf in the 000_subj_level_feat folder
create a bash array of all the .fsf files in 000_subj_level_feat and a function to run one feat
run all the feat .fsf in parallel - 10 at a time because of RAM limitations.

ETA is ~ 3hrs

## Copy Zstat to data_native Once all the first-level analyses have been carried out, it’s time to copy the final zstats for each taskrun into the /data_native folder. This is achieved by:

do_cp_Zstat_2_data_native.sh

The final structure (of the Zstat) for one subject in /data_native will be like the following:

sub_02
└── Zstat
    ├── task_1_run_1
    │   ├── zstat1.nii.gz
    │   ├── zstat2.nii.gz
    │   ├── zstat3.nii.gz
    │   └── zstat4.nii.gz
    ├── task_1_run_2
    │   ├── zstat1.nii.gz
    │   ├── zstat2.nii.gz
    │   ├── zstat3.nii.gz
    │   └── zstat4.nii.gz
    ...
    └── task_4_run_2
        ├── zstat1.nii.gz
        ├── zstat2.nii.gz
        ├── zstat3.nii.gz
        └── zstat4.nii.gz