Running multiple ROIs via probtrackx2

1

I am trying to run a set of ROI’s on the same subject via FSL probtrackx2, and want each ROI to be run independently. Initially I had my datagrabber (below), match mniROIs=‘addlInfo/subjContainer/%s/DTI_ROIs/Human_Left_Hippocampus.nii.gz’ so that the mniROI only returned a single file. I changed it to return to a wildcard, so it returns a list of all files in the DTI_ROIs subdirectory (and the output looks correct in the _report.rst below…)

In my workflow, I initially just connected the single ROI to the “seed” input for pbx2.

runpbx2.connect( datasource,‘mniROIs’, pbx2, ‘seed’)

After changing to a wildcard, I added an iterfield (i.e. seed)

pbx2 = pe.Node(interface=fsl.ProbTrackX2(), name=‘pbx2’, iterfield=[‘seed’])

But the workflow keeps creating a “seeds.txt” file which is a text file containing all of the ROIs. I feel like I am missing something obvious, but I’m not exactly sure what.

Long story short, I want to run a set of X ROI’s from a datagrabber node independently through probtrackx2.

Any pointers or examples on how to do that? I must just be screwing up the syntax somewhere.

“”"

Setup for DataGrabber inputs needed for probtrackx2

“”"

datasource = pe.Node(interface=nio.DataGrabber(infields=[‘subject_id’],
outfields=[‘nodif_brain_mask’,‘xfm’,‘invxfm’,‘thsamples’,‘phsamples’,‘fsamples’,‘mniROIs’]),
name=‘datasource’)

create a node to obtain the functional images

datasource.inputs.base_directory = “/data/HCP_BedpostData/”
datasource.inputs.template =’’
datasource.inputs.sort_filelist = True
datasource.inputs.field_template = dict(
thsamples=’%s/T1w/Diffusion.bedpostX/merged_%s.nii’,
fsamples=’%s/T1w/Diffusion.bedpostX/merged_%s.nii*’,
phsamples=’%s/T1w/Diffusion.bedpostX/merged_%s.nii*’,
nodif_brain_mask=’%s/T1w/Diffusion.bedpostX/%s.nii*’,
xfm=’%s/MNINonLinear/xfms/%s.nii*’,
invxfm=’%s/MNINonLinear/xfms/%s.nii*’,
mniROIs=’%s/DTI_ROIs/*’
)

datasource.inputs.template_args = dict(
thsamples = [[‘subject_id’,‘th1samples’]],
phsamples = [[‘subject_id’,‘ph1samples’]],
fsamples = [[‘subject_id’,‘f1samples’]],
nodif_brain_mask = [[‘subject_id’,‘nodif_brain_mask’]],
xfm = [[‘subject_id’,‘acpc_dc2standard’]],
invxfm = [[‘subject_id’, ‘standard2acpc_dc’]],
mniROIs = [[‘subject_id’]]
)

  • fsamples : /data/HCP_BedpostData/100206/T1w/Diffusion.bedpostX/merged_f1samples.nii.gz
  • invxfm :
  • mniROIs : [’/data/HCP_BedpostData/100206/DTI_ROIs/Human_BasalForebrain_Bilat_wimt.nii.gz’, ‘/data/HCP_BedpostData/100206/DTI_ROIs/Human_Hippocampus_Left_wimt.nii.gz’, ']
  • nodif_brain_mask : /data/HCP_BedpostData/100206/T1w/Diffusion.bedpostX/nodif_brain_mask.nii.gz
  • phsamples : /data/HCP_BedpostData/100206/T1w/Diffusion.bedpostX/merged_ph1samples.nii.gz
  • thsamples : /data/HCP_BedpostData/100206/T1w/Diffusion.bedpostX/merged_th1samples.nii.gz
  • xfm :
2

You might want to look at the shell script dti_3_tract.sh or the Matlab function doDtiTractSub(). Both assume a region of interest map that has been resliced to the native space of the DWI data (make sure to use nearest neighbor interpolation: a partial volume voxel with portions of region 17 and 19 should not be interpolated as label 18). The Matlab function will use the GPU if available, which is a good idea for these applications. While complex (and tuned for stroke applications), the full pipeline will generate connectomes that can be analyzed with NiiStat (either GLM or machine learning).

3

Thanks Chris-- I’ll check out that script/workflow.

I realized when I specified the iterfield=[‘seed’] I also needed to wrap the pbx2 command as a MapNode instead of a Node.

Now it’s running each ROI separately… I just need to figure out the naming convention for each of the mapNode outputs which are just named numerically for now.

Thanks again