GUI AutoTrack versus CLI AutoTrack

13 views

Skip to first unread message

Karen Aguilar Mateu

unread,

May 14, 2026, 3:13:43 PM (11 days ago) May 14

to DSI Studio

Hello,

I am running batch AutoTrack processing using the Linux CLI version of DSI Studio (Ubuntu 24.04 CPU build, May 2026 version).

I observed a consistent discrepancy between:

· GUI AutoTrack

· CLI AutoTrack (--action=atk)

for several anatomically difficult bundles, especially:

· ProjectionBrainstem_CorticospinalTractL/R

· ProjectionBasalGanglia_FornixL/R

· Association_ArcuateFasciculusR

Important observations:

1. The same .fz file is used in both GUI and CLI.

2. GUI AutoTrack successfully reconstructs these tracts.

3. CLI AutoTrack often fails to generate any .tt.gz output for the same bundles.

4. This persists even after: increasing terminate_if up to 50000, lowering fa_threshold to 0.12, running tracts individually and using the latest Ubuntu 24.04 CPU build

5. Other bundles reconstruct correctly in CLI.

6. Some difficult bundles reconstruct in some subjects but not others.

Example CLI command:

dsi_studio --action=atk \

--source=subject.fz \

--track_id=ProjectionBrainstem_CorticospinalTractL \

--tracking_index=fa \

--fa_threshold=0.12 \

--terminate_if=50000

My question is:

Are there implementation differences between GUI AutoTrack and CLI AutoTrack?

Or is there another recommended parameter for improving CLI AutoTrack convergence in difficult bundles?

Thank you very much.

Frank Yeh

unread,

May 18, 2026, 12:13:39 PM (7 days ago) May 18

to DSI Studio

Hello Karen,

The GUI AutoTrack and CLI --action=atk are not always directly comparable. In the GUI, AutoTrack is more like running tracking with --action=trk and specifying track_id, while the exact tracking parameters may be set internally by the GUI. Because of this, the tracking parameters can be different from those used in your CLI --action=atk command.

The best way to check this is to look at the console output or command history from the GUI run. It should show the actual tracking command and parameters used by the GUI. You can then copy those parameters into the CLI command and see whether the results match.

For example, instead of comparing GUI AutoTrack directly with:

dsi_studio --action=atk \
--source=subject.fz \
--track_id=ProjectionBrainstem_CorticospinalTractL

please check whether the GUI output is closer to something like:

dsi_studio --action=trk \
--source=subject.fz \
--track_id=ProjectionBrainstem_CorticospinalTractL \
[other tracking parameters shown in the console]

For difficult bundles such as CST near the brainstem, fornix, or arcuate fasciculus, small differences in parameters can affect whether the tract is generated. Please compare the full GUI console output with the CLI command, especially the tracking index, threshold, seeding/tracking settings, and stopping criteria.

If you can send the console output from a successful GUI run and the corresponding CLI command that fails, I can help identify which parameter is different.