zeigar / niral_dti_pipeline

Python scripts to execute a DTI pipeline.

NIRAL_DTI_PIPELINE

Table of contents

---

1. Pipeline flow diagram
2. Configuring and running the pipeline
3. Pipeline component specification
• DWI to DTI
• AutoSeg
• FDT masks
• FDT bedpost
• FDT probtrack
• Connectome

1. Pipeline flow diagram

--- 

The flow diagram shown above represents the single-process pipeline version. The major difference between the multi-process and single-process version is the multi-process version computes all the seed regions in parallel (i.e. does not include a loop that sequentially processes each seed region).

As illustrated, the pipeline defines six components that perform the sequence of operations listed below:

1. DWI to DTI: This takes the skull stripped DWI image and creates a DTI, RD, AD, FA, and DWI-b0 image volume.
2. AutoSeg: Creates a white matter, gray matter, and CSF segmentations using the Imperial atlas and the T1 and T2 images.
3. FDT masks: Creates the masks required to run FDT bedpost and probtrack. Specifically,
• no_diff_brain mask
• waypoint mask
• exclusion mask
• seed masks (one for each brain region defined in Imperial parcellation)
• termination masks (one for each brain region defined in Imperial parcellation)
4. FDT bedpost: Execute the FDT Bayesian Estimation of Diffusion Parameters Obtained using Sampling Techniques (bedpostx) algorithm that includes modeling for crossing fibers.
5. FDT probtrack: Execute the FDT probabilistic tracking with crossing fibers (probtrackx) algorithm for each seed region.
6. Connectome: Create NxN connectivity matrix using the probtrack results.

2. Configuring and running the pipeline

--- There are two ways to run the pipeline: using the single-process version or using the multi-process version. To run the single process version from the command line type:

python dti_pipeline.py –config config_pipeline.txt

and to run the the multi-process version form the command line type:

python dti_pipeline_mp.py –config config_pipeline.txt

In general, the configuration file defines where the subject folder is located, a list that includes the subjects to be processed, the locations of the Autoseg computation and parameter files, and the pipeline components to be executed. For instance, on a Linux/UNIX/Mac operating system if the study folder is located at

/data/study_subjects

and in this folder the following sub-folders (typically named using subject specific unique ids)

subject_id1
subject_id2
subject_id3
subject_id4

The subjects.txt file can be created using the command ls > subjects.txt. Then the dti pipeline configuration file would have the following key-value pairs.

SubjectFolder:/data/study_subjects
SubjectList:/data/study_subjects/subjects.txt

For the autoseg template and pipeline component flag files, if the git rep was cloned in folder /data/git/NIRAL_DTI_PIPELINE then the dti pipeline configuration file would have the following key-value pairs.

COMPFILE:/data/git/NIRAL_DTI_PIPELINE/autoseg_templates/AutoSeg_Computation.txt
PARMFILE:/data/git/NIRAL_DTI_PIPELINE/autoseg_templates/AutoSeg_Parameters.txt
Flags:/data/git/NIRAL_DTI_PIPELINE/example_config/config_pipeline_flags.txt

In general, the flag file indicates the pipeline components to be executed. A yes value indicates the component will be executed, and a no value indicates the component will not be executed.

Example configuration files and autoseg template files can be found in the github repository.

3. Pipeline component specification

---

DWI to DTI

AutoSeg

FDT masks

FDT bedpost

FDT probtrack

Connectome