Skip to content

Work With Different Data Formats

csttool accepts two input types: DICOM directories and 4D NIfTI files with separate gradient sidecars. This page covers conversion, naming conventions, and the FOV check that catches most "extraction failed" errors before they happen.

DICOM → NIfTI

csttool import wraps dcm2niix and produces a BIDS-style output.

csttool import --dicom /raw/sub-001/dicom --out ./work --subject-id sub-001

If the DICOM directory contains multiple diffusion series, specify one:

csttool import --dicom /raw/sub-001/dicom --out ./work --subject-id sub-001 --series 4

Use --scan-only to see what would be converted without writing anything:

csttool import --dicom /raw/sub-001/dicom --out ./work --scan-only

Gradient sidecar naming

csttool auto-discovers .bval/.bvec files that share the stem of the DWI NIfTI. Both BIDS-style and dcm2niix-style names are recognised:

DWI file Accepted gradient files
sub-001_dwi.nii.gz sub-001_dwi.bval + sub-001_dwi.bvec
sub-001_dwi.nii.gz sub-001_dwi.bvals + sub-001_dwi.bvecs

If your sidecars live elsewhere, pass them explicitly to check-dataset (the other commands derive everything from import's output):

csttool check-dataset --dwi raw_dwi.nii.gz --bval custom.bval --bvec custom.bvec

FOV and resolution

Atlas-based CST extraction registers a template (FMRIB58_FA) into your subject's space, then projects motor-cortex and brainstem ROIs. If your acquisition does not include both endpoints in the field of view, extraction will succeed but produce empty or truncated bundles.

Whole-brain FOV is required

The DWI volume must cover from the vertex down to the caudal medulla (foramen magnum). Acquisitions clipped at the supratentorial level will silently produce bilateral CSTs that stop at the level of the cropping plane.

Use csttool check-dataset --dwi raw_dwi.nii.gz for a coverage diagnostic before launching the pipeline. See the check-dataset reference for the full report.

Voxel size

Default tracking parameters assume voxels in the 1.5–2.5 mm isotropic range. If your data is highly anisotropic, reslice during preprocessing:

csttool preprocess --nifti raw.nii.gz --out ./preproc --target-voxel-size 2 2 2