Different results using mrtrix pre-processed data vs DSI studio pre-processed data

[Maria Ana]

Hello all,

I have recently started using DSI studio to apply differential tractography (using QSDR for reconstruction) to some multi-shell diffusion data of multiple sclerosis patients. I have pre-processed the raw diffusion images using DSI studio, and then computed differential tractography between baseline and follow-up, but no tracts were selected. I also tried using the already pre-processed diffusion data from Mrtrix, which yielded some tracts in differential tractography for the same patient (I later realised that the bvec used in construction of SRC file along with the pre-processed diffusion file was not updaded after pre-processing in mrtrix, and I wonder if this is why there is a difference between methods). Is there any reason as to why these results should be different? Also, shouldn't differential tractography show something for multiple sclerosis patients (considering I do the pre-processing steps using DSI studio)?

Thank you!

PS: mrtrix pre-processing steps are as follows: denoising, removal of Gibb's artifact, concatenation of AP and PA images, topup and eddy, bias field correction

[Frank Yeh]

The results may not be the same, but should be consistent if the
effect size is large enough.
Did you follow the steps in the tutorial videos at
https://practicum.labsolver.org/ ?

[Maria Ana]

Thanks for your response.

Yes, I did. I've repeated the differential tractography step now, just to make sure. And now I get results using the pre-processed images in DSI studio (and not using the ones from mrtrix). Also, I now get results in another patient where no tracts appeared before when using differential tractography with dsi studio-processed images. I'm not sure what's the problem, as I've done now what I did before and the results are different. I might have done something wrong before without realising it.

But still, mrtrix and dsi studio results are different, now with tracts only appearing for dsi studio-processed images. Do you think this could be because of mrtrix pre-processing steps? I read somewhere that the bvec file should be updated after pre-processing to be able to construct the SRC files if I already have the pre-processed images. The bvec file I use is never updated and thus I am constructing a SRC file with the image processed with mrtrix + the raw bvec file. Could this be hindering results in any way? 

[Frank Yeh]

There could be numerous possible causes.
If you have a sample dataset that you are really sure there is a
neuronal change, I can process on my side to show you my results as a
comparison.

[Maria Ana]

This is an exploratory analysis so I don't know if there should be any neuronal change or not, even though I would suspect some change (it's ten months between baseline and follow-up, and these are early multiple sclerosis patients without medication at baseline). If you could give me an e-mail, I could send you the raw and mrtrix-preprocessed files, if you think that could help. Thank you!

[Frank Yeh]

DSI Studio website has a data upload link on the left. It is private
and only send the data to me.
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 visit https://groups.google.com/d/msgid/dsi-studio/2af144c3-3c95-43a2-a86d-07b209c658ebn%40googlegroups.com.

[Maria Ana]

I've just uploaded the files. Thank you so much for your help

[Frank Yeh]

I checked the data. DTI FA shows no obvious differences. But QA has differences that could be mapped using (m1-m2)/m1 with 1,000,000 seeds:

There is a bug in the current version using the the undivided metrics m1-m2. It will be fixed today.

To view this discussion visit https://groups.google.com/d/msgid/dsi-studio/776f9373-6e30-4c46-897c-4d54b640866dn%40googlegroups.com.

[Maria Ana]

Sorry, I don't think I understood: so there are no differences between timepoints using DTI FA (i.e. in differential tractography) using either DSI or mrtrix-preprocessed data?

Also, is it feasible to estimate QA using multi-shell data? I was under the impression that QA should be estimated using grid sampling and not multi-shell. Sorry to keep bothering you, I'm just a little confused.

[Frank Yeh]

Sorry, I don't think I understood: so there are no differences between timepoints using DTI FA (i.e. in differential tractography) using either DSI or mrtrix-preprocessed data?

Yes

Also, is it feasible to estimate QA using multi-shell data? I was under the impression that QA should be estimated using grid sampling and not multi-shell. Sorry to keep bothering you, I'm just a little confused.

GA can be estimated from single shell, multiple shell and grid.

To view this discussion visit https://groups.google.com/d/msgid/dsi-studio/e6e8bead-6bdd-41be-a444-19fb3e30eb34n%40googlegroups.com.

[Maria Ana]

Sorry for taking so long to get back to you. Thank you so much for all your help. Just one last question, do you have any references about QA being estimated with shell sampling protocols? Just so I can be more informed about this.

[Frank Yeh]

Yeh FC et. al. IEEE TMI 2010

[Maria Ana]

Hi again,

Sorry, I have tried to replicate the result you showed me, but I can't seem to do it. Could you tell me what were the parameters you used? I used all default parameters with a threshold metrics1>metrics2 of 0.1 and the threshold type (m1-m2)/m1. The results from images pre-processed in DSI and in mrtrix are also different, when I think they shouldn't be?

Thank you for any help!

[Frank Yeh]

The results won't be the same as mrtrix.

For details, you may check out tutorial videos at practicum.labsolver.org

Best

Frank

To view this discussion visit https://groups.google.com/d/msgid/dsi-studio/0daa8bc4-dece-456c-82ae-fbe8738746a2n%40googlegroups.com.

[Frank Yeh]

Sorry for all the trouble. I will do my best to troubleshoot.

Let work out the crash first.

Are you using the Windows version?

Could you send me the .src.gz or .sz files that you used in Step T2 again? (I don't keep a copy of users data)

Hope to get this issue fixed asap.

Best

Frank

Hi again,

Sorry for being insistent. But I've watched the workshop video twice now, and I've used the same parameters as the ones used there. I still don't achieve the same results as you. This is what I get (using a threshold of 0.1 with QA and contrast (m1-m2)/m1, as well as 1 million seeds and I pre-process images using just DSI studio with all the recommended steps):

Also, I can't seem to use the most recent version of DSI studio, as it crashes anytime I try to do the GQI reconstruction, and anytime I try to upload a slice for differential tractography. 

Thank you.

Maria

You received this message because you are subscribed to a topic in the Google Groups "DSI Studio" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/dsi-studio/4ZcvM7uIN3o/unsubscribe.
To unsubscribe from this group and all its topics, send an email to dsi-studio+...@googlegroups.com.
To view this discussion visit https://groups.google.com/d/msgid/dsi-studio/CAG_Qrt%2BJ2t6EZvC%3Dz2iV%2B10qZkgT7yG--yh0BGdgGMyHra%3D56A%40mail.gmail.com.

[Maria Ana]

Hi again,

Sorry for being insistent. But I've watched the workshop video twice now, and I've used the same parameters as the ones used there. I still don't achieve the same results as you. This is what I get (using a threshold of 0.1 with QA and contrast (m1-m2)/m1, as well as 1 million seeds and I pre-process images using just DSI studio with all the recommended steps):

Also, I can't seem to use the most recent version of DSI studio, as it crashes anytime I try to do the GQI reconstruction, and anytime I try to upload a slice for differential tractography. 

Thank you.

Maria

Frank Yeh <fran...@gmail.com> escreveu (segunda, 16/12/2024 à(s) 22:30):

You received this message because you are subscribed to a topic in the Google Groups "DSI Studio" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/dsi-studio/4ZcvM7uIN3o/unsubscribe.
To unsubscribe from this group and all its topics, send an email to dsi-studio+...@googlegroups.com.
To view this discussion visit https://groups.google.com/d/msgid/dsi-studio/CAG_Qrt%2BJ2t6EZvC%3Dz2iV%2B10qZkgT7yG--yh0BGdgGMyHra%3D56A%40mail.gmail.com.

[Maria Ana]

Thank you.

Yes, I'm using the windows "Chen" version ("Hou" is the one that crashes). I'll upload the files. 

Thanks again

[Frank Yeh]

Sorry about the bug.

Could you let me know the exact steps that induced the crashed on the new version?

[Maria Ana]

If I do GQI reconstruction, it crashes when I try to load a slice in step T3 to do differential tractography.

[Frank Yeh]

Could you upload the .fz or fib.gz file you created from GQI reconstruction and also the slice you loaded?

Once I reproduce the crash, I will have a good chance of finding the bug and fix it.

Best regards,

Frank

[Maria Ana]

Yes, I just uploaded the files. The fz file for the baseline scan and the QA slice from follow-up.

[Frank Yeh]

I tested on my side. I can successfully load the slice with the fz file.

Did you use the GPU version? If yes, could you try the CPU version?

If the CPU version works, then it is likely the CUDA toolkit is not installed or you do not have an NVIDIA graphic card. You may need to install CUDA toolkit or use the CPU version.

Best

Frank

[Maria Ana]

Yes, I was using the GPU version. I'll try the CPU version, thanks.

Also, what about getting different results from the ones you showed me here? Why do you think that happened? I used the default parameters and the ones from the workshop.

[Frank Yeh]

Sorry for the delayed reply.

Also, what about getting different results from the ones you showed me here?

The results should be consistent, and may not be identical.

Why do you think that happened?

If the effect size is small, then you may get inconsistent result. To confirm, we always compute FDR to see if the finding is reliable.

I used the default parameters and the ones from the workshop.

The SCA dataset has good effect size and should give very similar results.

If you get a very different results, please let me know the steps. Once I can reproduce the error, I will fix it right away.

Thanks again for reporting the issues, and hope we can help here.

Best

Frank