Simplified understanding of SDF in the local connectome

[Ninju Menon]

Hi Frank, 

Our lab has just started using DSI Studio with the goal of doing group connectivity analysis. We've read through the base article https://www.sciencedirect.com/science/article/pii/S1053811915009647?dgcid=api_sd_search-api-endpoint as well looking at the documentation on the website, and reading  https://www.sciencedirect.com/science/article/pii/S1053811911006392#bb0235. 

I just wanted to clarity some concepts as we are more used to the DTI-TBSS framework where we get one FA, MD value per voxel, so please feel free to correct me if I am stating something incorrectly. 

SDF: The SDF is basically a density measure (mm3) of water diffusing along different fiber directions in a voxel. Does this mean that if I looked at one voxel after QSDR reconstruction, I would see multiple density values in mm3 units per voxel, for each fiber direction? Does the same apply to QA, that we would get a QA value for every fiber orientation in the voxel? 

Connectometry: Translating this to the local connectome, is every element in the row vector corresponding to the local connectome of one individual is the SDF value along a specific fiber direction? My biggest confusion is with the line that says "sample SDFs at local fiber direction from common atlas". Ultimately, what is each element in the row vector of the local connectome matrix in terms of one subject? Are there specific units? Understanding this would help us greatly in understanding how to interpret the results of the connectometry analysis. 

I apologize if these questions are too basic, or if there is a simplified explanation somewhere that I have missed. Please feel free to direct me toward any simplified material that I can look at to better understand these metrics. 

Thank you so much for your time and patience, and of course kudos on developing a great tool! 

[Fang-Cheng Yeh]

> SDF: The SDF is basically a density measure (mm3) of water diffusing along different fiber directions in a voxel. Does this mean that if I looked at one voxel after QSDR reconstruction, I would see multiple density values in mm3 units per voxel, for each fiber direction?

Not exactly. Currently the SDF is normalized in a way that makes a
free water diffusion voxel has SDF=1 at all directions regardless of
the voxel size.

> Does the same apply to QA, that we would get a QA value for every fiber orientation in the voxel?

QA is measured after removing the isotropic component from the SDF.
It only has the anisotropic component.

>
> Connectometry: Translating this to the local connectome, is every element in the row vector corresponding to the local connectome of one individual is the SDF value along a specific fiber direction?

Yes.

> My biggest confusion is with the line that says "sample SDFs at local fiber direction from common atlas". Ultimately, what is each element in the row vector of the local connectome matrix in terms of one subject? Are there specific units? Understanding this would help us greatly in understanding how to interpret the results of the connectometry analysis.

Each element is the anisotropic SDF value of a subject sampled at one
fiber orientation. Here the fiber orientation is defined by the atlas,
not the SDF itself.

The value is a spin density measurement. In MRI, spin density has an
arbitrary unit, and it can only be scaled with respect to a reference.
Here the SDF is scaled in a way that free water SDF has value equals
one at all directions.

>
> I apologize if these questions are too basic, or if there is a simplified explanation somewhere that I have missed. Please feel free to direct me toward any simplified material that I can look at to better understand these metrics.have to e

No problem at all. This is not a basic question, and I often have to
explain several times in the DSI Studio workshop.
If there is still anything unclear, I will be more than happy to help
you clarify.

>
> Thank you so much for your time and patience, and of course kudos on developing a great tool!

My pleasure.
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/4de5ac74-940b-4f56-b183-ae70441b3f3d%40googlegroups.com.

[Ninju Menon]

Hi Frank, 

Thank you so much for your response. It is much clearer now as to what the connectometry process is. 

To clarify once again, each element in each row of the local connectome matrix is the QA value (SDF - isotropic component) at a given fiber orientation? 

We were able to run a sample connectometry analysis with age as the variable of interest, with the left precuneus from the FreeSurfer DKT atlas as a seed region. We tried with 10K and 30K seeds, and found that the the effect direction seemed to flipped with the 10k seed results (there was no track identified as having negative correlation with age). But at 30K, we did see some results and the effect seemed to be in the correct direction (more negative correlations with age than positive). Do you have any recommendations regarding this? 

Thank you, 

Niranjana 

[Fang-Cheng Yeh]

>
> To clarify once again, each element in each row of the local connectome matrix is the QA value (SDF - isotropic component) at a given fiber orientation?
>

Yes.

> We were able to run a sample connectometry analysis with age as the variable of interest, with the left precuneus from the FreeSurfer DKT atlas as a seed region. We tried with 10K and 30K seeds, and found that the the effect direction seemed to flipped with the 10k seed results (there was no track identified as having negative correlation with age). But at 30K, we did see some results and the effect seemed to be in the correct direction (more negative correlations with age than positive). Do you have any recommendations regarding this?

I would suggest using more seeds like 50k and 100k and see if the
result converges.
Frank

>
> Thank you,
> Niranjana

>
> --
> 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/7671e249-9a4b-491d-973a-6b514fe5bf27%40googlegroups.com.

[Ninju Menon]

Hello Frank, 

Thank you so much for your response. I do have a few follow up questions after our analysis: 

1. I was just looking at the statistics along the tracks (from http://dsi-studio.labsolver.org/Manual/diffusion-mri-connectometry#TOC-Visualize-T-statistics-along-the-tracks) and saw that mean SDF values were output for the subjects that I had marked as N/A (9999). According to the documentation, shouldn't these subjects be excluded from the analysis and therefore not have any SDF value?

2. Is there any other statistic we can extract along the tracks, such as the QA value? Specifically, I am trying to extract the track counts for each subject to know how they vary between groups. Is something like this possible? 

Thank you again for your help, 

Niranjana 

[Fang-Cheng Yeh]

> 1. I was just looking at the statistics along the tracks (from http://dsi-studio.labsolver.org/Manual/diffusion-mri-connectometry#TOC-Visualize-T-statistics-along-the-tracks) and saw that mean SDF values were output for the subjects that I had marked as N/A (9999). According to the documentation, shouldn't these subjects be excluded from the analysis and therefore not have any SDF value?

The output of mean SDF does not consider any demographics (which you
put in 9999) and thus will be exported for all subjects.

>
> 2. Is there any other statistic we can extract along the tracks, such as the QA value? Specifically, I am trying to extract the track counts for each subject to know how they vary between groups. Is something like this possible?
>

You can extract dti_fa, md, rd, ad, iso, rdi...etc, but you have to
create one connectometry databse for each of them. (see "index of
interest" at http://dsi-studio.labsolver.org/Manual/create-a-connectometry-database#TOC-Step-C2:-Create-a-connectometry-database).
Each connectometry database is for one type of diffusion metric.

Best regards,
Frank

> Thank you again for your help,
> Niranjana
>

> --
> 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/019a2379-bf13-4498-b150-f1f51c37fd70%40googlegroups.com.

[Ninju Menon]

Thanks for your prompt reply. 

I'm using the June 5th 2019 build, and I do not see QA or SDF as an option in the index of interest drop down menu. My .SRC files were reconstructed using QSDR. Should I be using GQI instead? 

Thank you, 

Niranjana 

[Ninju Menon]

Frank, 

My apologies, I do SDF as the index of interest, just not QA. Is not possible to load QA as the index of interest? 

[Fang-Cheng Yeh]

QA is calculated from the anisotropic component of SDF, and here the
SDF used in connectometry analysis is the anisotropic component.
They are the same but defined in different space.

> --
> 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/69b08cdb-741c-4dc9-9c81-5aeeb559ec90%40googlegroups.com.