Connectivity matrix calculation

[Mattia De Francisci]

Dear Frank,

I used DSI studio to obtain connectivity matrices starting from whole brain tractographies stored in .trk formats. I did the same using MRtrix (after converting my tractographies from .trk to .tck with https://github.com/nipy/nibabel/blob/master/nibabel/cmdline/trk2tck.py). I compared the mean matrices I obtained by using the two tools (from the same pool of subjects), and , although the general structure is preserved, DSI studio version of the mean structural connectivity matrix is sparser. I was wondering whether some filtering is applied in the connectivity matrix calculation in DSI (apart from the connectivity threshold, which I manually set to 0).

Best regards,

Mattia

[Fang-Cheng Yeh]

Yes, there is a threshold used to assign zero to the matrix entries if
their values are lower than a ratio multiplied by the maximum value of
the matrix. Usually, this ratio is 0.001 and can be adjusted at GUI or
command line (e.g. --connectivity_threshold).
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/8eec1606-667c-45bd-9a21-1ae5c02cf108n%40googlegroups.com.

[Mattia De Francisci]

Dear Frank,

Thank you for your reply. I'm aware of that threshold and in fact I set it to 0, so that no entry would be artificially put to 0.

But I still get different results from MRtrix and DSI-studio. It seems like the number of connections reported in DSI matrix is systematically lower than the MRtrix version. So i was wondering if this could be explained somehow by the algorithm DSI studio uses to assign the streamlines to the node. Which is the condition for assigning a streamline endpoint to a node of the parcellatiion? Is there any constraint, maybe based on some anatomical assumption?

Thank you a lot,

Mattia

[Fang-Cheng Yeh]

Hi Mattia,

I don't recall anything special in calculating the matrix. It is
just a simple match of the rounded coordinates.

Perhaps we can run a simplified example (e.g. left arcuate
fasciculus) and see what may have caused the difference?

Best regards,
Frank

On Wed, Aug 31, 2022 at 11:07 AM Mattia De Francisci

> To view this discussion on the web visit https://groups.google.com/d/msgid/dsi-studio/bd87dfec-9ef5-4c14-bd03-7734b53d36f1n%40googlegroups.com.

[Mattia De Francisci]

Hi Frank,

I think that MRtrix uses a radial search for assigning the streamlines to the nodes. I tried to change this rule and simply match the coordinates: now the matrices are much more similar. Thanks a lot!