Question: Sparse connectivity matrix

58 views
Skip to first unread message

Kayleigh Campbell

unread,
Jun 24, 2020, 1:00:35 PM6/24/20
to DSI Studio
Hello Frank, 

I had a question regarding connectivity matrices. I have 64 regions (cortical and subcortical) and would like to produce n x n connectivity matrices. 

I first perform whole-brain tractography (100,00 seeds, etc.), for example:
/Applications/dsi_studio.app/Contents/MacOS/dsi_studio --action=trk --source=subj01.fib.gz --parameter_id=CDCCCC3D9A99193F441DAF3E9A99193FCDCCCC3Db2041b9643A08601cba01dca  --output=subj01_whole_brain.trk.gz
 
Then generate the connectivity matrix:
/Applications/dsi_studio.app/Contents/MacOS/dsi_studio --action=ana --source=subj01.fib.gz --tract=subj01_whole_brain.trk.gz --connectivity=subj01_64regions.nii.gz --connectivity_value=count  --connectivity_type=end --connectivity_threshold=0 > subj01_matrix_logs.txt

However, when I view the resulting connectivity matrix, most of the edges = 0. The connectivity matrix appears quite sparse. I have attached a connectogram.txt file as an example. 

I was expecting to see more/higher counts between regions (edges), given the relatively few regions I am using here (and therefore larger volume per region). Do you have any suggestions for why this may be occurring? 

Thank you for your help!

subj01_count.end.connectogram.txt

Fang-Cheng Yeh

unread,
Jun 24, 2020, 2:04:00 PM6/24/20
to dsi-s...@googlegroups.com
--connectivity_type=end requires the track to strictly end in the region.
I would suggest using  --connectivity_type=pass in your case (maybe the region is too thin?)
Best,
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 on the web visit https://groups.google.com/d/msgid/dsi-studio/26cbe1c3-97e0-4952-9c66-0a149cfd65dfn%40googlegroups.com.

Kayleigh Campbell

unread,
Jun 24, 2020, 3:28:42 PM6/24/20
to DSI Studio
Hi Frank, 
Thank you for getting back to me - much appreciated!
Attached is a screenshot of an axial view of one subject's regions overlaid on FA. I don't think the regions are too thin - any thinker would overlap onto WM tracts, etc. 

I tried setting --connectivity_type=pass (example attached). However, with this I worry about structures such as the cingulate gyrus that have many fibres passing through on route to the cortex. What does "connectivity_type=pass" actually represent? Is it the number of fibres that enter, for example region 1 and region 2, that may either end in the region or continue to some other region? 

Thanks for the help!!

subj01_ax_regions.png
subj01_count.pass.connectogram.txt

Fang-Cheng Yeh

unread,
Jun 24, 2020, 4:35:38 PM6/24/20
to dsi-s...@googlegroups.com
Attached is a screenshot of an axial view of one subject's regions overlaid on FA. I don't think the regions are too thin - any thinker would overlap onto WM tracts, etc. 


What I would do is to (1) find where edge=0 seems problematic, (2)identify the two corresponding regions, (3) add regions as "end" region in [Step T3a] , (4) load the whole brain track, and (4) use [Step T3][Tracts][Filter by ROI...] function to figure out there are tracks ended in these two end regions.
 
I tried setting --connectivity_type=pass (example attached). However, with this I worry about structures such as the cingulate gyrus that have many fibres passing through on route to the cortex. What does "connectivity_type=pass" actually represent? Is it the number of fibres that enter, for example region 1 and region 2, that may either end in the region or continue to some other region? 

"pass" count any tracks that have reached to the region, regardless of whether they ended within. 
Your concern is right. The "pass" will have a flaw with cingulum.

Best,
Frank
Thanks for the help!!

On Wednesday, June 24, 2020 at 11:04:00 AM UTC-7 Frank Yeh wrote:
--connectivity_type=end requires the track to strictly end in the region.
I would suggest using  --connectivity_type=pass in your case (maybe the region is too thin?)
Best,
Frank

On Wed, Jun 24, 2020 at 1:00 PM Kayleigh Campbell <k.camp...@gmail.com> wrote:
Hello Frank, 

I had a question regarding connectivity matrices. I have 64 regions (cortical and subcortical) and would like to produce n x n connectivity matrices. 

I first perform whole-brain tractography (100,00 seeds, etc.), for example:
/Applications/dsi_studio.app/Contents/MacOS/dsi_studio --action=trk --source=subj01.fib.gz --parameter_id=CDCCCC3D9A99193F441DAF3E9A99193FCDCCCC3Db2041b9643A08601cba01dca  --output=subj01_whole_brain.trk.gz
 
Then generate the connectivity matrix:
/Applications/dsi_studio.app/Contents/MacOS/dsi_studio --action=ana --source=subj01.fib.gz --tract=subj01_whole_brain.trk.gz --connectivity=subj01_64regions.nii.gz --connectivity_value=count  --connectivity_type=end --connectivity_threshold=0 > subj01_matrix_logs.txt

However, when I view the resulting connectivity matrix, most of the edges = 0. The connectivity matrix appears quite sparse. I have attached a connectogram.txt file as an example. 

I was expecting to see more/higher counts between regions (edges), given the relatively few regions I am using here (and therefore larger volume per region). Do you have any suggestions for why this may be occurring? 

Thank you for your help!

--
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 on the web visit https://groups.google.com/d/msgid/dsi-studio/26cbe1c3-97e0-4952-9c66-0a149cfd65dfn%40googlegroups.com.

--
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.

Kayleigh Campbell

unread,
Jun 24, 2020, 8:27:02 PM6/24/20
to DSI Studio
Thanks for your response!

I just tried your suggestion. It appears to be functioning correctly. I obtained the same result as what is shown in the connectivity matrix for fibre count ending in a given region pair. However, the fibre count ending in regions that 'should' have a stronger connection is very low. For example, very few fibres that 'should' form somatosensory pathways from thalamus-to-parietal regions are present (< 15). Would you recommend I increase whole-brain seeding from 100,000 --> 500,00 or greater?

If I proceed with using --connectivity_type=pass, there will be a lot of 'noise' and produce, for example, artificially high network density. 
Do you have a recommendation for filtering/thresholding parameters to apply to help with this?

Thanks again for your help!!

Fang-Cheng Yeh

unread,
Jun 24, 2020, 8:39:39 PM6/24/20
to dsi-s...@googlegroups.com
A lot of tracks have synapses in the thalamus, but fiber tracking will just pass through. For thalamic connections, "pass" will be more accurately describing the connectivity.
What you can do is to run "end" and "pass" respectively and use the thalamic connectivity from "pass" to replace those in "end"
This should give more reasonable results.
Frank

Reply all
Reply to author
Forward
0 new messages