Using Nilearn to Extract Seitzman Networks

Software Support
1

Hi Neuroimagers:

A quick question about what the best approach would be to extract the Seitzman networks with spheres of different sizes using Nilearn.

Summary of what happened:

Previously, I’ve used Nilearn’s NiftiSpheresMasker to extract resting state networks. In my current projects, I am hoping to extract the Seitzman networks which have differently sized spheres. I was hoping I could pass in a list of radii that would match with my coordinates, but it seems like NiftiSpheresMasker only accepts one radius to describe all coordinates passed in.

Some solutions I’ve thought of:

  • Extract the timeseries for all nodes with NiftiLabelsMasker using the Seitzman .nii and the ROIs_anatomicalLabels.txt label file and match them up later with the ROIs_300inVol_MNI_allInfo.txt which has the network affiliation of the nodes.
  • Extract nodes using NiftiSpheresMasker and group all nodes that have the same size radius by network and then combine different time series from the different radius nodes to get a single time series for each network.

I feel like there must be an easier way. Has anyone else experienced this and found an easier work-around?

Thanks for brainstorming!!

Command used (and if a helper script was used, a link to the helper script or the command generated):

N/A

Version:

N/A

Environment (Docker, Singularity / Apptainer, custom installation):

N/A

Data formatted according to a validatable standard? Please provide the output of the validator:

Data is bids valid!

Relevant log outputs (up to 20 lines):

N/A

Screenshots / relevant information:

2

Hi,
I think that your first solution is actually the simplest: if there exists a Seitzman.nii file, then you can get the ROI time courses easily. The association with mets information (names, network structure etc.) should not pose any difficulty.
Happy to help if you encounter any issue.
Best,
Bertrand

3

Attaching the code I used below in case this is helpful for anyone in the future!

### This script extracts the Seitzman networks
###
### Katharina Seitz
### June 2024

# Python version: 3.8.4
import os
import json
import pandas as pd #1.0.5
import nibabel as nib #3.2.1
import numpy as np #1.19.1
from nilearn.input_data import NiftiLabelsMasker #0.8.1

def extract_seitzman_nodes(sub, ses, funcindir, procindir, sesoutdir):   
    flist = os.listdir(funcindir)
    # Location of the fMRIprep & mask image
    file_rest_mask = os.path.join(funcindir, [x for x in flist if ('brain_mask.nii.gz' in x and 'task-rest' in x)][0])
    rest_mask_img = nib.load(file_rest_mask)

    # Load parameters to get TR
    param_rest_file = open(os.path.join(funcindir, sub+'_ses-1_task-rest_space-MNI152NLin6Asym_desc-preproc_bold.json'),)
    param_rest_df = json.load(param_rest_file)
    rest_tr = param_rest_df['RepetitionTime']
    
    # Load processed resting state scan
    rest_cen2 = nib.load(procindir+'/'+sub+'_'+ses+'_task-rest_final_nr6.nii.gz')
    
    # Load the atlas image and labels
    labels_img = nib.load('/projects/b1108/templates/Seitzman300/Seitzman300_MNI_res02_allROIs.nii.gz')
    labels_path = '/projects/b1108/templates/Seitzman300/ROIs_anatomicalLabels.txt'
    labels_df = pd.read_csv(labels_path, sep='\t')  
    labels_list = labels_df.iloc[:, 0]
    
    
    ####Begin network extraction
    masker_rest = NiftiLabelsMasker(labels_img=labels_img,
                                labels=labels_list,
                                mask_img=rest_mask_img,
                                smoothing_fwhm=None,
                                standardize=True,
                                detrend=False,
                                low_pass=None,
                                high_pass=None,
                                verbose=5,
                                t_r=rest_tr
                            )
    rest_time_series = masker_rest.fit_transform(rest_cen2)
    
    ##### Save Output
    #save timeseries data
    np.savetxt(sesoutdir+sub+'_'+ses+'_task-rest_seitzman_timeseries_nr6.csv',
    rest_time_series, delimiter=',')

    return
    
def get_network_timeseries(sub,ses, sesoutdir):
    # Load file with node's network affiliations
    all_info = '/projects/b1108/templates/Seitzman300/ROIs_300inVol_MNI_allInfo.txt'
    network_affiliations = pd.read_csv(all_info, sep=" ")
    timeseries = pd.read_csv(sesoutdir+sub+'_'+ses+'_task-rest_seitzman_timeseries_nr6.csv',
                             header=None)
    
    # Get a list of the networks
    networks = network_affiliations['netName'].unique()
    
    #Iterate through each network and create timeseries for just that network
    for network in networks:
        network_indeces = network_affiliations[network_affiliations['netName'] == network].index
        network_timeseries = timeseries.iloc[:, network_indeces]
        np.savetxt(sesoutdir+'/'+sub+'_'+ses+'_seitzman_' +network + '_timeseries_nr6.csv',
            network_timeseries, delimiter=',')
        # Correlate every column with every other column
        corr_matrix = np.corrcoef(network_timeseries, rowvar=False)
        np.savetxt(sesoutdir+'/'+sub+'_'+ses+'_' + network + '_network_corrmat_nr6.csv',
                corr_matrix, delimiter=',')


def main():
    ses = "ses-1" #bids ses-1
    fmriprepdir = '/projects/b1108/studies/TEAM/data/preprocessed/neuroimaging/fmriprep_23_2/'
    procdir = '/projects/b1108/studies/TEAM/data/processed/neuroimaging/aib_networks_rest/'
    outdir = '/projects/b1108/studies/TEAM/data/processed/neuroimaging/seitzman_networks_rest/'
    subjects = os.scandir(fmriprepdir)
    for sub in subjects:
        if("sub-" in sub.name and not(".html" in sub.name) and sub.name =='sub-01270'):
            #define subject specific file paths
            funcindir = fmriprepdir + sub.name + '/' + ses + '/func/' 
            sesoutdir = outdir + sub.name + '/' + ses + '/'
            procindir = procdir + sub.name + '/' + ses + '/'
            preproc_rest = os.path.join(funcindir, sub.name +'_'+ses+'_task-rest_space-MNI152NLin6Asym_desc-preproc_bold.json')
            if(os.path.exists(preproc_rest)):
                os.makedirs(os.path.join(outdir, sub.name, ses), exist_ok=True)
                try:
                    extract_seitzman_nodes(sub.name, ses, funcindir, procindir, sesoutdir)
                    get_network_timeseries(sub.name,ses, sesoutdir)
                except Exception as e:
                    print(e)

if __name__ == "__main__":            
    main()