Atlases¶
QSIParc supports all atlases that QSIRecon has warped into subject diffusion space. No atlas-specific configuration is required — QSIParc discovers atlas parcellations and their label files automatically.
How atlases work in QSIParc¶
QSIRecon registers brain atlases from MNI space into each subject's diffusion space during processing. QSIParc consumes these pre-warped files:
*_seg-{name}_dseg.nii.gz— Parcellation volume in subject T1w spaceatlases/atlas-{name}/{name}_dseg.tsv— Label look-up table with region names
QSIParc never performs atlas registration — it is entirely downstream of QSIRecon.
QSIRecon built-in atlases¶
QSIRecon ships with several atlases from the 4S parcellation series:
| Atlas name | Parcels | Description |
|---|---|---|
4S156Parcels |
156 | Multi-scale cortical + subcortical + cerebellar |
4S256Parcels |
256 | |
4S456Parcels |
456 | |
4S1056Parcels |
1056 | Fine-grained; large connectivity matrices |
Schaefer2018N100Tian2020S2 |
100 cortical + subcortical | Schaefer cortical + Tian subcortical |
Schaefer2018N200Tian2020S2 |
200 cortical + subcortical | |
Schaefer2018N400Tian2020S2 |
400 cortical + subcortical |
Availability depends on the QSIRecon workflow configuration used for your dataset. Check which atlases were produced:
Filtering by atlas¶
# Run only for one atlas
qsiparc /data/qsirecon /out --atlas Schaefer2018N100Tian2020S2
# Run for all discovered atlases (default)
qsiparc /data/qsirecon /out
Region naming¶
Region names are loaded from the atlas TSV in <qsirecon_dir>/atlases/atlas-{name}/. The region_name column in the output TSV comes directly from this file.
Hemisphere inference¶
If the atlas LUT does not include a hemisphere column, QSIParc infers it from the region name using common neuroimaging naming conventions:
| Pattern | Assignment |
|---|---|
Prefix: lh_, lh-, left_, l_ |
L |
Suffix: _lh, _left, _l |
L |
Infix: _lh_, _LH_ |
L |
Schaefer style: 7Networks_LH_Vis_1 |
L |
Same patterns with rh/right/r |
R |
| No pattern matched | bilateral |
Fallback LUT¶
If no TSV is found in the atlases/ directory, QSIParc extracts unique integer labels from the NIfTI and assigns generic names:
with hemisphere = bilateral for all regions. This is a last resort — meaningful downstream analysis requires proper region names.
Atlas output directory¶
Each atlas gets its own subdirectory within the subject/session output:
sub-001/ses-01/dwi/
├── atlas-Schaefer2018N100Tian2020S2/
│ ├── ...diffmap.tsv
│ └── ...connmatrix.csv (×4)
├── atlas-4S156Parcels/
│ ├── ...diffmap.tsv
│ └── ...connmatrix.csv (×4)
└── ...
Using a custom atlas¶
To use a custom atlas:
- Register it to each subject's diffusion space (e.g. via ANTs or FSL) — this must be done outside QSIParc.
- Name the parcellation file following the
*_seg-{custom_name}_dseg.nii.gzconvention and place it in the subject'sdwi/directory within the QSIRecon derivatives tree. - Place the label TSV at
<qsirecon_dir>/atlases/atlas-{custom_name}/{custom_name}_dseg.tsv. - Run QSIParc normally — it will discover the file automatically.