Difference between GUI AutoTrack result and command-line result using the same parameter_id

[Chiara Matti]

Hi,

I am trying to reproduce in the command line the same tractography and structural connectivity matrix that I obtain in the DSI Studio GUI using the AutoTrack preset. I am using subject 103818a.qsdr.fz of hcp-ya-retest.

In the GUI, the tracking report shows:

A deterministic fiber tracking algorithm (Yeh et al., PLoS ONE 2013) with augmented tracking strategies (Yeh, Neuroimage 2020) was used.

The anisotropy threshold was randomly selected between 0.5 and 0.7 of the Otsu threshold. The angular threshold was randomly selected between 45° and 90°. Step size was set to voxel size. Tracks shorter than 30 mm or longer than 200 mm were discarded. A total of 1,000,000 tracts were tracked. parameter_id=c9A99193Fb803FdbF041b484340420Fcba803FcCDCC4C3Ec

I also compute the connectivity matrix with the following parameters: custom parcellation in MNI, end in region, count, no threshold.

I then run the following command:

dsi_studio --action=trk \
  --source=103818a.qsdr.fz \
  --thread_count=8 \
  --output=103818a_test.tt.gz \
  --parameter_id=c9A99193Fb803FdbF041b484340420Fcba803FcCDCC4C3Ec \
  --connectivity=/Applications/dsi_studio.app/Contents/MacOS/atlas/human/human_schaefer_100_TH_mni.mni.nii.gz \
  --connectivity_threshold=0.0 \
  --connectivity_value=count \
  --connectivity_type=end

However, when I load the resulting .tt.gz file into the GUI using Load Tracts, the tractogram looks noticeably different from the tractogram generated in the GUI using the same preset. Both runs produce 1,000,000 tracts, but when I load the CLI output is seems that there are way less tracts.

Question 1: Why does this happen?

I also tried outputing trk.gz (instead of  tt.gz) and manually tuning the parameters to:

'--tract_count=1000000',

'--min_length=30.0',

'--max_length=200.0',

'--track_voxel_ratio=1',

'--tip_iteration=0', but the results are the same. Here are the differences in connectivity matrices:

- when i use the parameter_id

- when I manually change the parameters

Note: the CLI manual params and the CLI parameter_id are equal.

Question 2: why are the count values of the GUI matrix almost double the CLI ones?

If I divide by 2 the GUI matrix and plot the same matrices as before:

Question 3: there are still some differences. Are these acceptable?

Thank you for your help. Best, Chiara

[Frank Yeh]

I will check this in detail and get back to you soon.

Best regards,

Frank

--
You received this message because you are subscribed to the Google Groups "DSI Studio" group.
To unsubscribe from this group and stop receiving emails from it, send an email to dsi-studio+...@googlegroups.com.
To view this discussion visit https://groups.google.com/d/msgid/dsi-studio/b3e31ade-d735-402b-9474-3bfbb0e47394n%40googlegroups.com.

[Frank Yeh]

Could you update DSI Studio and see if this issue is solved?

Older versions GUI and CLI may differ in whether they compute the symmetric part of the connectivity matrices. The issue was fixed recently.

Best regards

Frank

[Chiara Matti]

Dear Professor, 

Thank you for your reply. It seems that after updating the dsi_studio version, the command line argument --connectivity_value that I was using before is not working anymore. If I omit it, all the possible connectivity matrices get saved. Is there a way to avoid this and select only a specific metric to compute the matrix?

Also, I noticed that the threshold option is not there anymore, does this mean that the matrices are unthresholded now? Could you point me to where I can find the updated documentation regarding this matter?

Thank you again for your help,

Best, 

Chiara

[Frank Yeh]

Thank you for your reply. It seems that after updating the dsi_studio version, the command line argument --connectivity_value that I was using before is not working anymore. If I omit it, all the possible connectivity matrices get saved. Is there a way to avoid this and select only a specific metric to compute the matrix?

Currently DSI Studio will output all of them and there is no command to select only a metric.
We don't have a plan to modify this because we can easily copy or delete files right after the command line.
However, for users needing to get values, specifying each of them is much more difficult and cumbersome.

 

Also, I noticed that the threshold option is not there anymore, does this mean that the matrices are unthresholded now?

Yes, the new version won't apply a threshold to zero the connectivity matrices.
The reason is that users can easily apply thresholding to the matrix, but if DSI Studio applies it first, then users cannot undo them.
 

Could you point me to where I can find the updated documentation regarding this matter?

The command line documentation have mentioned the new syntax at https://dsi-studio.labsolver.org/doc/cli_t3.html
Best regards,

Frank 

To view this discussion visit https://groups.google.com/d/msgid/dsi-studio/82622849-5591-4fc7-8f2f-3bc8efcb8689n%40googlegroups.com.