Use of mulitple Shell DTI for ALPS Method

[marKa53]

Dear Dr. Barisano,

Thank you very much for sharing this great code. I am still new to the topic of DTI ALPS. In the 2024 paper "Diffusion Tensor Image Analysis Along the Perivascular Space (DTI-ALPS): Revisiting the Meaning and Significance of the Method" by Taoka et al., it is stated that "The use of a single b-value instead of multiple shells makes it impossible to separate the velocity components of diffusion, and the information is considered to be a mixture of diffusion components of different velocities. This issue needs to be verified in the future."

As part of a study, we acquired a multiple shell DTI sequence with various b values (5 2005 995 2000 1995 995 2000 2005 1000).

1.I successfully processed a subject yesterday, and the ALPS indices are about 1.5 on both the right and left sides. I wanted to ask if the script is designed to also process multiple-shell DTI, or would I need to make some adjustments?
2. What is your assessment of using multiple shell DTI for ALPS? Does it make sense to calculate DTI ALPS and ALPS Index this way?

Best regards,
Marcel

[gbarisano]

Hi Marcel,

Thank you for your message! And sorry for the delay in my response.
Here my answers to your questions:

1. No, I did not design the script to specifically process multi-shell DTI data. The tensor model which is fitted by my script is generally not a good model for multi-shell data: as the b-value increases, the diffusion signal becomes sensitive to more complex microstructural features of tissue, such as cell membranes, myelin, and other barriers that restrict or hinder water diffusion. These structures lead to non-Gaussian diffusion patterns, which the tensor model cannot capture accurately. The DTI model typically works with single-shell data, having b0 value and typically b-value = 1000s/mm^2 (maximum b-value 1200s/mm^2, less frequently up to 1500s/mm^2)
2. Despite the point above, even though by a theoretical point of view it might seem inappropriate to calculate DTI-ALPS on multi-shell data with multiple b-values above 1200s/mm^2, I did some tests on multi-shell data with my script and the ALPS results were reasonable. I also tried to re-run the script using only the b0 images and the b values close to 1000 or below 1000, but the DTI model did not fit very well.
   One thing to keep in mind is that FSL's eddy in my script likely will not work for multi-shell data.
   Based on these tests, I think that as long as the FA maps look good, the ALPS results should be acceptable, but the actual values likely will not be directly comparable with DTI-ALPS values obtained from traditional single-shell DTI, so you would need to make sure that all the diffusion data included in your study have been acquired with the same multi-shell protocol.
   I will try to run some additional tests and assess more systematically DTI-ALPS on multi-shell data.

I hope this helps. Let me know if you have any question or suggestion.

Giuseppe

[marKa53]

Hello Giuseppe,
First of all, thank you for the helpful and detailed answer and please excuse the longer delay in responding.

Unfortunately, the study has already been closed for a year. All our DTIs are multiple shells, so we will now run the script over our DTIs and then see if the FA maps and values look reasonable as you suggested.

I will get back to you soon when we have finished processing and give an update on how the script was able to handle our multiple shell DTI and where we may have difficulties.

Many thanks and best regards
Marcel

PS: In fact I found 2 publications with your script which also used a mulitple shell DTI with higher b values.

[xiiao-b]

Thank you very much for your code.

I would like to ask if, in my data (e.g., HCP), which contains multiple b-values, I need to select a specific b-value to use with this script?

[gbarisano]

Hi @xiiao-b.
I think that using the b-values closest to 0 and to 1000 will give you DTI-ALPS values that will be consistent with most papers using DTI-ALPS. However, this selection may lead to having a low number of volumes, insufficient for proper DTI fitting, resulting in FA maps of bad quality.
In that case, I would try to use the multi-shell data as they are, as this should give you good FA maps, and would mention in the discussion that these DTI-ALPS measurements might not be strictly comparable to DTI-ALPS measurements traditionally obtained from single-shell diffusion data.

I hope this helps. Let me know if you have any question.

Giuseppe