Possible steps to add to DTI pipeline

[CaseyWeiner]

Possible changes/enhancements to m2g_dwi_pipeline :

Add Conversion Step

• Many doctors receive their images from the scanner in dicom or .dcm format.
• I suggest incorporating Rordens dcm2niix here (installs with apt-get)
  • Possibly even better is using dcm2bids (pip installable)
• Would make m2g more "soup to nuts"
• Also more effective time-management for clinicians

Add Denoising Steps

• MP-PCA (docs), then Gibbs Ringing Effect (docs)
  • Perform these steps before eddy_correct

Add Option to Perform either eddy_openmp or eddy_cuda

• Both parallelized, better than eddy_correct
• See here

Add Bias Field Correction

• From nipype here, also available in mrtrix3
• Do this after eddy correction? Or before? I originally thought after
• ANTs N4BiasField generally considered better than FSL's fast

Streamline Robustness

• From Dipy here
• Remove redundant streamlines, evaluate quality of fiber tracking

Connectome Export Format

• Allow user to choose format of connectome in /outdir/connectomes/atlas_name, including:
  • Edgelist (default)
  • Normalized Edgelist
  • Adjacency Matrix
  • Normalized Adjacency Matrix
• With these choices, will need to add norm_method parameter to cover the normalization approach the user wants

Customization

• Allow user to choose steps to be skipped in a run, without checking that step has been run
• Ex: allowing user to skip eddy_correct without checking for a previous eddy_correct run
  • Some clinicians prefer to skip certain steps to save time
  • This is more personal preference

[CaseyWeiner]

Any of this sound like a good idea @Lawreros ? I can work on these in my fork if any are of interest

[CaseyWeiner]

Also I know the lost ROIs are already written to text files, but we could also write one large text file that lists the directories of all the images that are missing ROIs. That way one can immediately see how many subjects lost ROIs instead of looking in each folder.