Skip to content

Latest commit

 

History

History
executable file
·
36 lines (24 loc) · 2.01 KB

README.md

File metadata and controls

executable file
·
36 lines (24 loc) · 2.01 KB
Raw

Linear-Planar double PGSE sequence processing

This repository contains the preprocessing and fitting scripts for double PGSE data ("Bibek sequence")

The current preprocessing pipeline uses the following steps

  1. Estimate slicewise gaussian sigma (or voxelwise if you have noisemap) with autodmri github preprint
  2. Correct for non-gaussian Bias (second moment method)
  3. Denoising (Mrtrix's implementation of MPPCA)
  4. Gibbs ringing correction (Mrtrix's implementation of the subvoxel shifts method)
  5. Split dataset by b-value and b-tensor shape
  6. Compute susceptibility induced distorsions from AP PA B0 (FSL's topup)
  7. Linear registration (motion correction) and averaging of lowest b-value (FSL's mcflirt)
  8. Get brain mask from moco low b image for eddy current correction (FSL's bet)
  9. Estimate gradient non-linearity distorsion (THIS version of gradunwarp, requires "secret" coeff.grad file)
  10. Estimate diffusion gradient eddy current induced distorsions (FSL's eddy)
  11. single interpolation application of (6) (9) (10)
  12. Apply spherical averaging to shell
  13. Compute b0 image from fitting tensor on 2 lowest shells

TODOs and Issues

  • Upgrade debiasing second moment method for recalibration from NLSAM
  • Because the vector norms in the gradient scheme file are relative, we need to detect the bmax probably from the DICOM header to be able to remove all the hardcoded b-values
  • The one step interpolation of eddy and gradient non-linearity relie on hardcoded software paths
  • the coeff.grad file relies on an hardcoded path
  • Fix and move back to Cornelius Eichner's version of one step interpolation of eddy and gradient non-linearity

Steps to make this work:

  1. download and install autodmri github