Automatic Fiber Tracking

Use --action=atk to initiate automatic fiber tracking.

Automatic fiber tracking maps individual bundles using deterministic fiber tracking and tract recognition. The implementation is detailed in Yeh, Fang-Cheng. “Shape analysis of the human association pathways.” Neuroimage 223 (2020): 117329.

Examples

1. Run fiber tracking on all FIB files to map standard pathways:

dsi_studio --action=atk --source=*.fz

2. Track specific bundles (e.g., corticospinal tracts and optic radiation):

dsi_studio --action=atk --source=*.fz --track_id=Corticos,Optic

3. Output tracts in the template space:

dsi_studio --action=atk --source=*.fz --template_track

Core Functions

Parameter Default Description
source (required) Specify .fz or legacy .fib.gz files for automatic bundle tracking.
track_id Atlas-defined standard groups Bundle names separated by commas. Partial names are accepted and may match multiple bundles. For a reproducible named-bundle workflow, use the current atlas entry; choose the parent entry for the whole tract family and a child only for a specific subdivision.
template Current/default template Select the tractography template. Use the template list provided by the current DSI Studio version rather than relying on fixed numeric names.
tolerance FIB/template dependent Bundle-recognition tolerance in mm. If omitted, DSI Studio derives a tolerance series from the selected FIB/template.
track_voxel_ratio FIB dependent Track-to-voxel ratio. If omitted, DSI Studio uses the value derived from the FIB.
check_ending 1 Apply ending constraints used by automatic tract recognition.
yield_rate 0.00001 Restart/stop an unproductive tracking attempt when tract yield is extremely low.
overwrite 0 Overwrite existing results when set to 1.
trk_format tt.gz Output tractography format.

Optional Functions

The majority of parameters used in --action=trk are also supported.

AutoTrack guidance

Standard named AutoTrack entries already include atlas-defined anatomical constraints. Do not add extra ROI, ROA, End, Seed, NotEnd, Limiting, or Terminative regions by default; add an additional region only when a specific anatomical question requires it.

Topology-informed pruning (TIP) is intended for a coherent bundle rather than generic whole-brain cleanup. When a named bundle reaches roughly 5,000–10,000 or more tracts before pruning, 3–4 TIP iterations are a practical starting point. If the bundle is too sparse, inspect acquisition, reconstruction, tracking yield, and tolerance before applying aggressive pruning.

Streamline count reflects tractography sampling. It is not an axon count or a direct measure of biological connection strength.