Recommended Built-In Reconstruction & Tractography Workflows in QSIRecon for DKI Data

[liubowen]

Hello everyone,

I am working with a multi-shell DKI dataset that has been successfully preprocessed using QSIPrep.

Currently, I have used the built-in dipy_dki workflow in QSIrecon to reconstruct the diffusion and kurtosis tensor metrics. To proceed with fiber tractography and further analysis, I would like to ask a few questions:

1. Besides dipy_dki, are there any other built-in workflows in QSIrecon that are particularly recommended for the analysis of DKI data?
2. I have noticed the dsi_studio_autotrack; dsi_studio_gqi workflow, which seems excellent for automated tract segmentation. Is this workflow directly applicable to my DKI data?
3. From an analysis strategy perspective, what are the main differences in principle, results, and application scenarios between using QSIrecon’s dsi_studio_autotrack; dsi_studio_gqi versus using the standalone DKE software to first estimate tensors and then perform tracking with its DKE-FT module?

Any help or insights would be greatly appreciated. Thank you!

[Steven]

Hi @liubowen,

It would help to know the b-vals and number of directions per shell. But in general:

@liubowen

Besides dipy_dki, are there any other built-in workflows in QSIrecon that are particularly recommended for the analysis of DKI data?

Several workflows in QSIRecon are also suitable for multi shell data, including NODDI, GQI, and MAPMRI. I encourage you to read the QSIRecon documentation: Built-In Reconstruction Workflows — qsirecon 1.1.1.dev16+gfa8673957 documentation

@liubowen

I have noticed the dsi_studio_autotrack; dsi_studio_gqi workflow, which seems excellent for automated tract segmentation. Is this workflow directly applicable to my DKI data?

Yes.

@liubowen

From an analysis strategy perspective, what are the main differences in principle, results, and application scenarios between using QSIrecon’s dsi_studio_autotrack; dsi_studio_gqi versus using the standalone DKE software to first estimate tensors and then perform tracking with its DKE-FT module?

You’d have to be more specific about what tracking parameters you’d be using for DKE. But in general the DSI Studio AutoTrack method returns only the bundles (no whole brain tractography), is a result of deterministic tracking (not probabilistic), and includes macro- and micro-structural summaries of the bundles. You can also automatically produce summary measures from other QSIRecon workflows (eg NODDI) if you stay within the Qsirecon ecosystem.

Best,
Steven

[liubowen]

This is great! Thank you, Steven.

Your response was really helpful and gave me a much clearer understanding of QSIrecon’s workflows and the differences between the tracking methods.

I’m diving into the documentation now, and I really appreciate your guidance!

[liubowen]

Sorry to bother you with another question. I was looking into the best way to map scalar data to bundles in QSIRecon.

I know I could write a custom pipeline, but then I noticed the built-in hbcd_scalar_maps workflow. The documentation says it’s a “general-purpose way” to do this. I just wanted to double-check with you – does this mean I can apply it to my own multi-shell dataset, even though it’s not from the HBCD study? It seems like a very powerful option if it’s generally applicable.

Thanks for your insight!

[Steven]

Hi @liubowen,

@liubowen

does this mean I can apply it to my own multi-shell dataset, even though it’s not from the HBCD study?

Yes. Also keep in mind that you can make your own recon spec too that has even more. For example, the HBCD spec does not have NODDI, but you can combine the contents of those recon specs into one.

Best,
Steven