DSI Studio

[tang ying]

Excuse me, some problems occurred to me when I was using DSI-studio. 

 My GFA value was much lower than that in the literature（it‘s around 0.9 in the literature, while mine is below 0.2). The processing steps are as follows.

Step 0：moving correction and eddy using FSL

Step 1: after correction, the data was changed to an SRC file. 

Step 2: reconstruction. In this step, l used batch processing and imported the brain mask manually.

    I also reconstructed using the uncorrected data. The GFA value of uncorrected data is slightly higher than that of corrected data, but they are all lower than 0.3.

   I wonder if the number of b-values would result in the situation.

    I would be very grateful if you could provide some help.

[Frank Yeh]

GFA is usually for higher b-value DWI data (e.g. HARDI).

It is likely that you have a low b-value acquisition of DTI.

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/2eb66177-0d6d-473f-8bad-6edcd149d31an%40googlegroups.com.

[Runtian Miao]

Hi Frank and Tang,

We have a similar issue here.

The maximum GFA value in our 179 subjects is "0.16647057235240936" which is much lower than the literature.

We are using DTI data with b value of 1000. Is the GFA a good connectivity map to look in this case? If not, do you have any other suggestions if we want to look at the white matter integrity?

Best,

To view this discussion on the web visit https://groups.google.com/d/msgid/dsi-studio/CAG_Qrt%2Bp8rVQrc4JrHYN7hmkubSWHgzgBZazLL8SDa9EhxJqNg%40mail.gmail.com.

[tang ying]

Thanks for your reply. The maximum b-value of our data is 3000. What is the suitable B-value for this model?

[Frank Yeh]

GFA value depends on the ODF calculations.

If the ODF is calculated from DSI or QBI (currently not available in DSI Studio), the isotropic part is largely reduced due to the radial integration parameter used, and this leads to a much higher GFA. 

In DSI Studio, the current GFA calculation using GQI includes the full isotropic component, and thus the value is much lower.

A lower b-value further reduces the GFA value, but the major effect is whether the isotropic component is included.

We can remove part of the isotropic component here, but the effect is similar to scaling the GFA so that it is closer to one. But this is entirely artificial and heavily depends on the parameters and may not be reproducible.

The ODF currently calculated does not have this additional step to handle the isotropic part, but the reality is a lower GFA.

Hope this explanation helps,

Best,

Frank

[Frank Yeh]

3000 is a suitable range. But the isotropic component has a larger effect here that reduces the GFA value.

To view this discussion on the web visit https://groups.google.com/d/msgid/dsi-studio/8374f5a4-f00e-49e7-b25d-4595af04e091n%40googlegroups.com.

[Runtian Miao]

Thanks for the explanation. I think it helps.

Another issue we have is that the overall connectivity of the matrix is low for ( Schaefer400*7 atlas): the average of the sparsity of the GFA  connectivity matrix is around 97% across 180 subjects. If we take the mean of the connectivity matrix and calculate the sparsity, it is around 64%. We obtained similar sparsity values when using other connectivity maps (etc mean length)

Do you have any suggestions about this? Should we use some coarser atlas like schaefer 100?

Best,

Rexy

To view this discussion on the web visit https://groups.google.com/d/msgid/dsi-studio/CAG_QrtLtQ8J70nh8LZGFo7YOYdsJqu0C7kBXY8zxAAHk558hkw%40mail.gmail.com.

[tang ying]

Thank you very much. I have other questions as follows:

1) whether the number of diff.direction would influence GFA value?

2) what factors would decrease the weight of  isotropic component?

3) can you offer me a data example which have relative high GFA value?

Thank you again.

[tang ying]

Thank you very much. I have other questions as follows:

1) whether the number of diff.direction would influence GFA value?

2) what factors would decrease the weight of  isotropic component?

3) can you offer me a data example which have relative high GFA value?

Thank you again.

On Tuesday, November 21, 2023 at 7:05:35 PM UTC+8 Runtian Miao wrote:

[Frank Yeh]

Thank you very much. I have other questions as follows:

1) whether the number of diff.direction would influence GFA value?

GFA depends on ODF shape, and the shape will be affected by number of dif directions

 

2) what factors would decrease the weight of  isotropic component?

The total amount of water content, including free water and isotropically-restricted water

 

3) can you offer me a data example which have relative high GFA value?

Most HARDI data reconstructed by QBI will have high GFA.

 

To view this discussion on the web visit https://groups.google.com/d/msgid/dsi-studio/61d20aa0-e703-4247-a92f-d85864b32cb4n%40googlegroups.com.

[tang ying]

Hello. I have used the sequence which is recommended on the website( Issues in HCP-Style Multishell Acquisitions | DSI Studio Documentation (labsolver.org)  ),

with both 101 diff. directions and 258 diff. directions with bmax=4000. However, the value of GFA is still very low, around 0.3.

Whether there are some potential factors that l have neglected?

Thanks for your kindness.

[tang ying]

Additionally, l download a data and its fib.gz file from the website  src - OneDrive (sharepoint.com) . I notice the value of GFA is also below 0.5. It is a little strange. 

whether this tool would calculate GFA relative low?

[Frank Yeh]

The low GFA with GQI reconstructed data is common and normal.

You will only get a high GFA on HARDI reconstructed by QBI.

Best regards,

Frank

To view this discussion on the web visit https://groups.google.com/d/msgid/dsi-studio/a93b269d-0e8d-4179-9ad2-88135922f21en%40googlegroups.com.

[tang ying]

I know! Thank you very much. 

By the way, what ready-made tools or scripts are usually used with QBI ?

I have tried "qboot" in FSL. But there was something wrong.

[Frank Yeh]

Sorry, I don't know.

Best regards,

Frank

To view this discussion on the web visit https://groups.google.com/d/msgid/dsi-studio/aae93f2b-5ac7-48c9-a3e4-c390d697ab1dn%40googlegroups.com.