how CMP deals with GM-voxels having low FA?

23 views
Skip to first unread message

Qingyang Li

unread,
May 5, 2012, 4:16:54 PM5/5/12
to cmtk-...@googlegroups.com
Hi CMTK developers,

Because FA values in GM are low, often tractography algorithms do not extend all the way in to the actual GM, rather, they stop somewhere in the WM.  

Some package deals with this is by ballooning out the GM boundary, such that any WM-voxel near the boundary "counts" as being within the GM, and more precisely, within a particular region. 

I would like to know how CMP deals with this problem. 

Thanks

Yang

Qingyang Li

unread,
May 10, 2012, 6:23:36 PM5/10/12
to cmtk-...@googlegroups.com
Could you please guide me which code should I read to get the answer?
Thanks

Yang
------
Qingyang Li

Balance and focus.


Alessandra Griffa

unread,
May 15, 2012, 4:28:57 AM5/15/12
to cmtk-...@googlegroups.com
Hi Yang,

in the CMP we do not use any constraint about the (g)FA, i.e., the FA values are a not a criteria for fibers termination.

Concerning the WM mask, we consider the FreeSurfer one, and we slightly expand it (through morphological operations) in proximity of the deep nuclei. Anyway in principle there is not overlap between the WM mask and the cortical parcellation.

For the tractography, we start the fibers in the WM and we propagate them till a voxel labeled differently than WM is met (or the criteria on the angle is not respected).
It is true that because of the drop of FA in the WM-GM boundary region, the fibers could stop before reaching the GM because of not respected angle criteria. For this reason we tried to extend the fibers ending inside the WM, but close to the GM (3x3x3 voxels neighbourhood), toward the closest GM region. We could recover some fibers, but we did not observe significant variation of the overall connectivity. We therefore decide not to include this 'fibers endpoints extension' in the CMP.

Hope this can help
Alessandra

jovo

unread,
May 15, 2012, 8:51:15 AM5/15/12
to cmtk-...@googlegroups.com
hi alessandra - that is *very* helpful indeed!

i wonder 2 very simple questions: 

1) do you "seed" streamlines in every WM voxel? 
2) do you then propagate in both directions of the principal mode of ODF?

best,
joshua

ps - i am working with yang.

Alessandra Griffa

unread,
May 16, 2012, 9:49:52 AM5/16/12
to cmtk-...@googlegroups.com
Hi Joshua!

1) yes
2) yes. For each seed we throw, we propagate along the local main directions of diffusion (maximum 3), and in both directions

Alessandra

jovo

unread,
May 16, 2012, 3:26:02 PM5/16/12
to cmtk-...@googlegroups.com
awesome thanks.


On Wednesday, May 16, 2012 9:49:52 AM UTC-4, alessandra griffa wrote:
Hi Joshua!

1) yes
2) yes. For each seed we throw, we propagate along the local main directions of diffusion (maximum 3), and in both directions

Alessandra

Reply all
Reply to author
Forward
0 new messages