Skip to main content

Diffusion MRI Preprocessing

Preprocessing diffusion MRI data for tractography

Pipeline

Part A applies the corrections every diffusion dataset needs: susceptibility distortion, noise, Gibbs ringing, head motion, and eddy currents. It runs either as a single QSIPrep container or as eight separate steps. Part B fits the tensor, registers the anatomical image, and estimates the fiber orientation distributions that tractography reads.

Part A — Core preprocessing · two routes to the same result

Route 1

QSIPrep

A BIDS app that runs the Part A corrections in one containerized command and writes a quality-control report per participant. It performs:

  • DICOM to BIDS (dcm2niix)
  • Skull stripping (SynthStrip)
  • Denoising
  • Gibbs correction (optional)
  • TOPUP
  • Eddy
  • Brain mask
  • QC report
QSIPrep route →

Route 2

Manual steps

The same corrections run individually with FSL, MRtrix3, and ANTs. Every parameter is exposed, and acquisitions QSIPrep does not support can be handled directly. These are the tools QSIPrep calls internally.

↓

Part B — Tractography readiness · common to both routes

BedpostX, shell extraction, intracranial volume, and BIDS conversion are documented separately. They are not part of the required path.