DTI for non-accidental head trauma in children

[johnk2...@gmail.com]

Hello Dr Yeh,

I really like DSI-studio, it has worked great for my needs. I am attempting DTI or GQI data analysis in children with non-accidental head trauma (with retinal hemorrhages). 

The data were acquired for standard DTI protocol on Siemens 3T. Some DTI are in the acute phase, some are >3 months follow up, some both.

B-value = 1000 s/mm2, in-plane resolution is 1.79688 mm, slice thickness was 3.5 mm, a minimum of 20 diffusion sampling directions, 

The subjects were sedated, but imaging is carefully reviewed to remove data sets with artifacts / bad suscept. 

These scans occurred over years, the plan is to match a control of similar age / date of scan.

My main research question is is to correlate visual electrophysiology (evoked potentials) with DTI, the hypothesis is that diffusion metrics mean something important to neural function.

I did some preliminary tractography of the optic radiations. The regions appear accurate, but it is clear the tracts have too much random orientations. Differences between DTI and GQI result in slightly different outcomes, but still have similar problems.

Now I plan to have auto registering  ROI in DSI-studio, then get REGION metrics from this such as 

fa, md, ad, rd, 

My question: does this approach look correct? Is there a better approach?

2) Can I other metrics such as QA, ISO, RDI with this DTI scheme?

3) should I resample to isotropic for my reconstructions?

Thank you so much. I am in total awe of your skills!

John Kelly

[Frank Yeh]

>
> My question: does this approach look correct? Is there a better approach?

You may also test "differential tractography" which can track the
precise segment of tracts with a decrease of FA or any metrics:
http://dsi-studio.labsolver.org/Manual/differential-tractography

The data acquisition is nonisotropic, and an isotropic voxel
resolution of 2 mm will provide better results.

> 2) Can I other metrics such as QA, ISO, RDI with this DTI scheme?

Yes, but with diminished sensitivity and/or specificity (see Fig 7 at
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6919327/ )

RDI may need multiple b-value to differentiate restricted and
nonrestricted diffusion.

> 3) should I resample to isotropic for my reconstructions?

Yes, but better to acquire isotropic voxel at the beginning, IMHO.

Best regards,
Frank

>
>
> Thank you so much. I am in total awe of your skills!
> John Kelly
>

> --
> 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/6255660c-54eb-420f-9360-06d0a0de30d8o%40googlegroups.com.

[johnk2...@gmail.com]

Thank you for your answer.

Is it possible to reconstruct an isotropic voxel resolution of 2 from the command line? (e.g., --action=rec ....)

I will have about 60 DTI scans 

:)   John Kelly, Seattle

[Frank Yeh]

You can add --cmd="[Step T2][Edit][Rotate to MNI2]" with the
--action=rec command:
http://dsi-studio.labsolver.org/Manual/command-line-for-dsi-studio#TOC-Image-reconstruction

DSI Studio also has GUI interface that provides batch processing. You
don't need to write a command-line script:
http://dsi-studio.labsolver.org/Manual/batch-processing-using-gui

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/81d87167-e933-490c-8f20-f73ea40d4128o%40googlegroups.com.

[Johnnie Kay]

Thank you for your response. Differential tractography would be an exciting application for this data set. For example, one important question is do these children have axonal shear injuries that emerge later. I realized a big problem is that early imaging can have subdural hemorrhage causing a mass effect, then the follow-up scan has significant post-event brain atrophy. These changes cause low R2 values in the differential tractography alignment. I have uploaded a sample image of large changes( fa map)  occurring in the anterior frontal lobe -- the R2 value was 73. Altering the brain mask did not greatly help the R2, but I am not sure if I could edit the follow-up brain mask.

Is it still appropriate to proceed?

[Frank Yeh]

The mass effect will create a nonlinear distortion, and recently I have updated DSI Studio to improve the registration.

You may update DSI Studio and see if it improves. If not, I can check your two SRC files and figure out a way to align them.

Best,

Frank

To view this discussion on the web visit https://groups.google.com/d/msgid/dsi-studio/e163bbba-91f8-404e-99c0-380b2cf2d8f1n%40googlegroups.com.

[Johnnie Kay]

Yes, the new version (3/18/2021 build)  improved the alignment, but R2 still says 72. Looking at it more carefully, I believe the program did a good job of alignment, but there is local volume loss as well in superior frontal cortex, and the R2 may reflect that. Do you still want me to upload both SRC files? I get a bug report in (chinese?) when I tried these jpg images. Thank you Frank!!

[Frank Yeh]

You may upload the SRC files, and I will check.

The mask needs to be enlarged, BTW.

To view this discussion on the web visit https://groups.google.com/d/msgid/dsi-studio/5214c982-0bb1-4d1e-a4fd-5c12f30dc6fbn%40googlegroups.com.

[Frank Yeh]

I got your point about the "volume loss."  It is due to the mask.

Please check out the documentation here: http://dsi-studio.labsolver.org/Manual/Reconstruction#TOC-Step-T2a:-Setup-a-Mask and enlarge the mask to cover the entire brain region.

Frank

[Johnnie Kay]

Just an update if anyone is interested. This is what I found works so far with my "clinical single shell" scans in DSI studio (3/18/2021 build):

-Motion correction (even if done prior by Siemens)

-Rotate to MNI, 2mm, resample isotropic

-On baseline scan, allow mask to be bigger to include subdural hem areas as this could resolve then brain can expand.

-Carefully remove artifacts on fiber tracking that included the areas with subdural hem. In general, there are few. 

John

[Johnnie Kay]

I believe I am getting close, but I need help understanding the Tracts -> statistics... output. 

I have differential tractography set "dec_qa" threshold to 0.3 (parameters are set similar to that in your publication) and I tracted the Left optic radiation. The statistics output I am interested shows a value for inc_qa and dec_qa. What does the inc_qa mean here? It is low, but shouldn't it be null? Here are sample values:

qa   0.069354
inc_qa   0.0105824
dec_qa   0.632226
inc_fa   0.00740802
dec_fa   0.42602
dti_fa   0.323737
md   0.68381
ad   0.91409
rd   0.56867
gfa   0.0980929
iso   0.183399
base_fa   0.323737
study_fa   0.191282

Thank you Dr. Yeh.

[Frank Yeh]

It is increased qa.

Since the values are very low, it is likely due to systematic noise.

In the differential tractography paper (Yeh 2019), I used "inc_qa" as the negative control to calculate the FDR. 

You may check out the sham scans section of the paper.

Frank

To view this discussion on the web visit https://groups.google.com/d/msgid/dsi-studio/817f47ba-6042-4560-8ec4-05f4f18aec15n%40googlegroups.com.

[Johnnie Kay]

Is it meaningful that inc_qa is significantly correlated with  baseline  ISO (r2 = 0.75 )?  In contrast, the correlation between dec_qa  and  baseline  ISO   is weak (R2 = 0.078).  Sorry if I am asking too many questions. This is just fascinating to me!

[Frank Yeh]

Honestly speaking, I have no idea, but this is an interesting finding (I did not notice this before in other disease models) and potentially meaningful.

You may need more case numbers to confirm whether ISO is predictive of inc qa.

Best,

Frank

To view this discussion on the web visit https://groups.google.com/d/msgid/dsi-studio/c3a1f354-6dbe-4147-9aad-3257e42116a4n%40googlegroups.com.