IndNet

Defining subject-specific brain networks by individualizing templates

by Florian Krause & Nikos Kogias

Introduction

IndNet is a Nipype inplementation of a dual-regression-like seed-based approach to individualize general binary templates to specific subjects. Templates are assumed to be in MNI space. Raw functional (e.g. resting state) and anatomical images are preprocessed and transformed into MNI space. Timecourses are extracted from the functional images for all templates (using the first eigenvariate) and fed into a GLM analysis, together with mean WM and CSF signals. Specific template combinations, representing networks of interest, are tested as contrasts in the GLM and results are thresholded using spatial mixture modeling and subsequently binarized. In addition to the resulting main network masks (which might be overlapping), IndNet also outputs exclusive masks of the networks of interest (i.e. non-overlapping masks). All masks are available in native (functional and anatomical) as well as MNI space/resolution.

Prerequisites

1. Install Nipype
2. Install FSL
3. Install graphviz
4. Install ICA-AROMA
   • Download
   • Install dependencies (see requirements.txt)
   • Make ICA_AROMA.py executable
   • Add ICA_AROMA.py to path, so it can be called system wide
5. Download IndNet

Donders cluster

If you are working on the compute cluster of the Donders Institute, please follow the following steps:

1. Load Python module by running command: module load python
2. Install virtualenv by running command: python -m pip install virtualenv --user
3. Create new environment in home directory by running command: cd && python -m virtualenv IndNet
4. Activate new environment by running command: source ~/IndNet/bin/activate
5. Update pip in environment by running command: pip install -U pip
6. Install Nipype into environment by running command: pip install nipype
7. Install ICA-AROMA
   • Download and extract ICA-Aroma-master to somewhere
   • Within ICA-Aroma-master
     • Install dependencies by running command: pip install -r requirements.txt
     • Make ICA_AROMA.py executable by running command: chmod a+x ICA_AROMA.py
   • Add ICA_AROMA.py to path, so it can be called system wide
     • Open file ~/.bash_profile in a text editor and add: export PATH=/path/to/ICA-Aroma-master/:$PATH
     • Run command source ~/.bash_profile
8. Download IndNet and extract IndNet-master to somewhere

Usage

1. Write script with custom workflow (see Indnet-master/example.py for an example)
2. Run script
3. Results are in results directory within the base_dir set in the script

Donders cluster

If you are working on the compute cluster of the Donders Institute, please follow the following steps:

1. Start a new interactive job by running command: qsub -I -l 'procs=8, mem=16gb, walltime=4:00:00'
2. Load Python module by running command: module load python
3. Load graphviz module by running command: module load graphviz
4. Activate environment by running command: source ~/IndNet/bin/activate
5. Write script with custom workflow (see Indnet-master/example.py for an example) and save it as my_script.py
6. Run script by running command: python my_script.py
7. Results are in results directory within the base_dir set in the script

When done, deactivate environment by running command: source ~/IndNet/bin/deactivate