QSIRecon: What is the best reconstruction workflow for our dMRI dataset?

Hi developers,

Our clinical-grade DTI data have only 20 directions and a b-value of 1000, without reverse phase-encode collection. What is the best reconstruction workflow would you suggest for our dMRI data to do an automated fiber quantification? I searched this question with Google, and it suggested to use pyAFQ pipeline for these low-angular resolution data. Do you agree with this response?

At the same time, I also want to compute white matter free water measures along the segmented tracts. Do you think our data can fulfill this purpose?

Thanks for your advice in advance.

Best,

Angel

Hi @angelwong and welcome to neurostars!

I’d recommend the ss3t_fod_autotrack recon spec. See this paper for benchmarking: https://onlinelibrary.wiley.com/doi/abs/10.1002/hbm.70429

QSIRecon does not have a single-shell workflow for free water estimation. Although by running the autotrack recon spec above, you will still get a GQI fit on your data which quantifies (among other things) isotropic diffusion.

If you need tract profiles, you would need to use a PyAFQ pipeline. But as far as I know, you can only get profiles for certain common scalars (e.g., DTI and DKI measures), not any arbitrary scalar map (e.g. your free water measures). So in your case, no out-of-the-box QSIRecon spec will do everything you want.

Best,

Steven

1 Like

Hi @Steven

Thanks for your prompt response!

Sorry that I was distracted by other project tasks. I just fully checked our dMRI data which are a bit heterogeneous: majority of our dMRI data have 60 directions with a single b-value of 1000 (n=814), the second most of the data have 20 directions with a b-value of 1000 (n=498), and minority of the data have other number of directions (e.g. 32, 33 [single-shell] or 99 [multi-shell]) (total n=92).

If I want to group the data with 60 and 20 directions together for the statistical analysis, is it more appropriate to use the same reconstruction workflow? Or should I use a better optimized workflow for the data with 60 directions, and don’t group the heterogeneous data together? If yes, can you recommend a workflow for the data with 60 directions?

Thanks again for your advice in advance!

Best,

Angel

Hi @angelwong,

You should use the same workflow and the harmonize your statistics afterwords with something like ComBat.

Best,

Steven

1 Like