Semi-Automated Neonatal Tractography Pipeline

This is an extension of nipype used to conduct analysis of diffusion MRI data in a semi-automated fashion using FSL and DSI Studio. It allows you to specify the analysis steps in a json file to quickly set up and potentially iteratively improve the inputs to the workflow.

• Docker
• Nipype
• FSL
• DSI Studio

1. Download the image.
2. docker load -i /path/to/dingo_dsi-feb2019.tar
3. On Linux for convenient use add the following to ~/.bashrc
   • function DINGO { docker run -it --rm --user=$UID:$UID -v /tmp/.X11-unix:/tmp/.X11-unix:ro -e DISPLAY=$DISPLAY -v $(pwd):/data dingo:dsi-feb2019 "$@"; }

1. Install Nipype, FSL, DSI Studio.
2. Download and extract the repository.
3. Add the DINGO directory to your python path (by adding /path/to/DINGO to your PYTHONPATH environment variable or a new line in a .pth file in your site-packages directory)

An analysis config specifies the workflow to be created. Several examples can be found in res.

• name : String, new nipype workflow directory that will be created in the current working directory.
• data_dir : String, parent directory to each patient's folder, will be prepended to FileIn, FileOut.
• included_ids : List, strings specifying included data "patientID_scanID_uniqueSUFFIX"
• steps : List, analysis steps to perform, ["Step1", ["NameofStep2","Step2"], ["NameofStep3","Step2"] ]

• method : Dictionary, if key does not exist for each step, will use default parameters for it.
  • <Name> : Dictionary
    • inputs : Dictionary, { parameter : value }, ( used parameters found in nipype InputSpec )
    • connect : Dictionary, { parameter : [ "SourceStepName", "SourceStepParameter" ] }
• email : Dictionary, email will be sent by smtp at the conclusion of the workflow
  • server : String, "smtp.server:port"
  • login : String, username
  • pw : String, password
  • fromaddr : String
  • toaddr : String

DINGO run /data/path/to/config.json

python
from DINGO.base import DINGO
aworkflow = DINGO('/path/to/config.json')
aworkflow.run()

OR

python /path/to/DINGO/DINGO/base.py /path/to/config.json