Raw lambda values from DTI model from QSIPrep/QSIRecon?

Software Support
,
1

Summary of what happened:

Does anyone know if you can get the raw lambda values directly (lambda1, lambda2, lambda3 which are extracted to calculate FA MD RD AD from the DTI model) from QSIprep/QSIrecon outputs anywhere?

Or would they have to be, like, reverse engineered from the FA MD RD AD values?

Apologies if I missed something in the standard outputs. Thank you so much!

Command used (and if a helper script was used, a link to the helper script or the command generated):

NA

Version:

QSIprep (ver1.0.0) and QSIrecon (ver1.1.2) (then mrtrix3/pyafq), but not married to those.

Environment (Docker, Singularity / Apptainer, custom installation):

Singularity/apptainer

Data formatted according to a validatable standard? Please provide the output of the validator:

NA - no data

PASTE VALIDATOR OUTPUT HERE

Relevant log outputs (up to 20 lines):

NA - no log

Screenshots / relevant information:

2

Hi @emilymharriott ,

I don’t think you could find this information on recent qsiprep versions, where only the preprocessing of diffusion images is done.
In qsirecon, those lambda values are saved in various built-in workflows (dsi_studio_autotrack, dsi_studio_gqi, ss3t_fod_autotrack, hbcd_scalar_maps, multishell_scalarfest). More information here.

3

Hi @jsein,

Yes, if you run the DSI Studio scalar export, you get (among others)

ad: Axial diffusivity (first eigenvalue) from a tensor fit
fa: Fractional anisotropy from a tensor fit
md: Mean diffusivity from a tensor fit
rd: Radial diffusivity from a tensor fit
rd1: Lambda 2 (second eigenvalue) from a tensor fit
rd2: Lambda 3 (third eigenvalue) from a tensor fit
txx: Tensor fit txx
txy: Tensor fit txy
txz: Tensor fit txz
tyy: Tensor fit tyy
tyz: Tensor fit tyz
tzz: Tensor fit tzz

Best,

Steven

1 Like
4

Hi @Steven , thank you for your comment!

Could you describe the procedure with DSI Studio scalar export? Do you need to input the data produced by qsiprep into DSIStudio?

5

Hi @jsein,

Sorry for not being clearer - this is the DSI Studio scalar export in QSIRecon. E.g.,

-   action: reconstruction
    input: qsirecon
    name: dsistudio_gqi
    parameters:
        method: gqi
    qsirecon_suffix: DSIStudio
    software: DSI Studio
-   action: export
    input: dsistudio_gqi
    name: scalar_export
    qsirecon_suffix: DSIStudio
    software: DSI Studio

which is in several pipelines (as you mentioned) and can be added to custom recon specs.

Best,

Steven

2 Likes
6

Hey everyone!

Thank you so much for your help with this. I’m truly so grateful. Following up – I got dsi_studio_gqi to run, and am looking at the derivatives folder (~/qsirecon_output/derivatives/) – but am not seeing any, well, numbers? or spreadsheets? just json files (labels, really) and nii files? do I get a spreadsheet of eigenvalues out of this or do I have to extract something from the .nii files?

Thank you!!!
Emily

7

Hi @emilymharriott,

To get bundle wise metrics you have to use the bundle_mapper recon action.

# Map scalars to bundles
-   action: bundle_map
    input: autotrack_fod_msmt
    name: bundle_means
    scalars_from:
    - gqi_scalars
    software: qsirecon

Make sure the input uses what you did for tractography, keeping in mind that as of now, it has to use an autotrack output (doesn’t matter which FODs are used though).

Best,

Steven

1 Like
8

THIS WORKED! Woohoo! Thank you so much. You’re the best.

In my tsv file, I end up with a (raw) mean, a masked_mean, and a weighted_mean… do you have any recommendations as to the most representative/accurate/standard mean? I would think masked but am obviously more than open to suggestions. Thank you!

9

Hi @emilymharriott ,

WOO! Personally, I prefer to use medians. Not a big deal for tensor metrics, but for some models (e.g MAP-MRI) that have small (in size) but large (in magnitude) outliers, they can skew means even within masks. I think median or masked_median are fine and should be very very correlated with one another. One thing I have noticed with autotrack using GQI FODS (as opposed to inputting MRTrix FODS) is that the bundles tend to extend out past white matter, and even sometimes out of brain mask. So, I think the masked median is sensible. But again, I doubt the choice of which metric will meaningfully change your results (if it does, that is something to worry about).

Best,

Steven

1 Like
10

Phenomenal. I can’t thank you enough!