13 Orientation Distribution Function Reconstruction
One of the primary aims of collecting diffusion MRI data is to estimate or “reconstruct” the voxel-wise Orientation Distribution Functions (ODFs) and Fiber Orientation Distributions (FODs). Briefly, these functions describe how water diffuses within each 3D pixel (voxel) of the brain. In white matter tracts, water preferentially diffuses along the length of axons rather than across them. By modeling this diffusion, researchers can infer the underlying microstructural architecture of the brain.
There are many different mathematical models for reconstructing these functions from the raw diffusion signals. The A2CPS project relies on two main methods: one that serves as a precursor for calculating regional summary metrics (like Fractional Anisotropy), and another designed for full-scale tractography (tracing long-distance connections). This kit describes the precursor to the former (the tensor model). For the derived regional summary statistics themselves, see Chapter 14. For the more complex model used as a precursor to probabilistic tractography (which handles crossing fibers), see Chapter 15.
13.1 Starting Project
13.1.1 Locate Data
On TACC, the data are stored underneath the releases. For example, data release v2.1.0 is underneath
/corral-secure/projects/A2CPS/products/consortium-data/pre-surgery-release-2-1-0Paths in this kit are written relative to that release folder, so the neuroimaging data are underneath mris.
These outputs are under mris/derivatives/qsirecon_fsl_dtifit. There are three groups of outputs.
$ ls mris/derivatives/qsirecon_fsl_dtifit
dtifit qsirecon-fsl split_shellsThese outputs were generated in sequence. First, the qsirecon-fsl were generated by the qsirecon workflow reorient_fslstd, which takes the outputs of Chapter 11 and reorients them into the format that is expected by the FSL Diffusion Toolbox. For details on the organization of that folder, see the QSIRecon documentation. Then, the split_shells outputs were generated. They contain the reoriented outputs, but split into different bvals. For example
$ tree -l mris/derivatives/qsirecon_fsl_dtifit/split_shells/sub-10010
mris/derivatives/qsirecon_fsl_dtifit/split_shells/sub-10010
└── ses-V1
└── dwi
├── sub-10010_ses-V1_acq-b0_space-T1w_dwi.bval
├── sub-10010_ses-V1_acq-b0_space-T1w_dwi.bvec
├── sub-10010_ses-V1_acq-b0_space-T1w_dwi.nii.gz
├── sub-10010_ses-V1_acq-b1000_space-T1w_dwi.bval
├── sub-10010_ses-V1_acq-b1000_space-T1w_dwi.bvec
├── sub-10010_ses-V1_acq-b1000_space-T1w_dwi.nii.gz
├── sub-10010_ses-V1_acq-b2000_space-T1w_dwi.bval
├── sub-10010_ses-V1_acq-b2000_space-T1w_dwi.bvec
├── sub-10010_ses-V1_acq-b2000_space-T1w_dwi.nii.gz
├── sub-10010_ses-V1_acq-b3000_space-T1w_dwi.bval
├── sub-10010_ses-V1_acq-b3000_space-T1w_dwi.bvec
├── sub-10010_ses-V1_acq-b3000_space-T1w_dwi.nii.gz
├── sub-10010_ses-V1_acq-b500_space-T1w_dwi.bval
├── sub-10010_ses-V1_acq-b500_space-T1w_dwi.bvec
└── sub-10010_ses-V1_acq-b500_space-T1w_dwi.nii.gz
2 directories, 15 filesFinally, the dtifit outputs were generated by running FSL’s dtifit on different combinations of the bvalues.
$ ls mris/derivatives/qsirecon_fsl_dtifit/dtifit/
b1000 b2000 b3000 multishellMost of those are self-explanatory (e.g., the b1000 subfolder contains the fits with the b1000 shell). The multishell outputs were generated from the full dMRI scan (that is, all shells). Which of these outputs to use is up to the individual researcher. According to the ABCD, the protocol and analysis plan on which A2CPS dMRI is based, many researchers may prefer the multishell outputs. That said, the tensor model is known to fit poorly to diffusion data with bvalues larger than about 1500 Jensen et al. (2005), which is a majority of this sequence. Given that, some researchers may prefer the b1000 outputs, but note that these are based on only 15 directions.
The contents of each of the dtifit subfolders are in a “bidsish” organization, containing the standard outputs of dtifit, grouped by subject and session.
# within mris/derivatives/qsirecon_fsl_dtifit/dtifit/b1000
$ tree sub-10220
sub-10220
└── ses-V1
└── dwi
├── sub-10220_ses-V1_acq-b1000_space-T1w_dwi_FA.nii.gz
├── sub-10220_ses-V1_acq-b1000_space-T1w_dwi_L1.nii.gz
├── sub-10220_ses-V1_acq-b1000_space-T1w_dwi_L2.nii.gz
├── sub-10220_ses-V1_acq-b1000_space-T1w_dwi_L3.nii.gz
├── sub-10220_ses-V1_acq-b1000_space-T1w_dwi_MD.nii.gz
├── sub-10220_ses-V1_acq-b1000_space-T1w_dwi_MO.nii.gz
├── sub-10220_ses-V1_acq-b1000_space-T1w_dwi_S0.nii.gz
├── sub-10220_ses-V1_acq-b1000_space-T1w_dwi_sse.nii.gz
├── sub-10220_ses-V1_acq-b1000_space-T1w_dwi_tensor.nii.gz
├── sub-10220_ses-V1_acq-b1000_space-T1w_dwi_V1.nii.gz
├── sub-10220_ses-V1_acq-b1000_space-T1w_dwi_V2.nii.gz
└── sub-10220_ses-V1_acq-b1000_space-T1w_dwi_V3.nii.gz
2 directories, 12 files13.1.2 Extract Data
Most of these outputs are nifti images, and so any tool capable of reading them will work. Some neuroimaging graphical user interfaces also have built in tools for viewing the tensors. For example, by default the FSL viewer fsleyes will accept as input a folder with dtifit outputs and by default display the estimated tensors.
$ fsleyes dwiThere are several options for refining the display. For example, fsleyes can also produce variations on color FA maps.
13.2 Considerations While Working on the Project
13.2.1 Data Generation
These outputs were generated by the qsirecon-fsl_dtifit_app. As discussed above, the orientation distribution functions were reconstructed with the standard diffusion tensor imaging model, as estimated by FSL’s dtifit (using the weighted-least squares option).
For a good general introduction to DTI, see the QSIRecon documentation.
Note that the tensor model is known to fit poorly to diffusion data with bvalues larger than about 1500 Jensen et al. (2005). Several advanced models have been proposed, and the A2CPS neuroimaging DIRC is actively testing these. Outputs from another model are expected to be available in the release 4 series.
13.2.2 Variability Across Scanners
Many MRI biomarkers exhibit variability across the scanners, which may confound some analyses. For an up-to-date assessment of the issue and overview of current thinking, please see Appendix D.
13.2.3 Data Quality
As with any MRI derivative, all pipeline derivatives have been included. This means that products were included regardless of their quality, and so some products may have been generated from images that are known to have poor quality—rated “red”, or incomparable. For details on the ratings and how to exclude them, see Appendix A. Additionally, extensive QC has not yet been performed on the derivatives themselves, and so there may be cases where pipelines produced atypical outputs. For an overview of planned checks, see Confluence.
13.2.4 Citations
The standard DTI reference is Basser et al. (1994).
In publications or presentations including data from A2CPS, please include the following statement as attribution:
Data were provided (in part) by the A2CPS Consortium funded by the National Institutes of Health (NIH) Common Fund, which is managed by the Office of the Director (OD)/Office of Strategic Coordination (OSC). Consortium components and their associated funding sources include Clinical Coordinating Center (U24NS112873), Data Integration and Resource Center (U54DA049110), Omics Data Generation Centers (U54DA049116, U54DA049115, U54DA049113), Multi-site Clinical Center 1 (MCC1) (UM1NS112874), and Multi-site Clinical Center 2 (MCC2) (UM1NS118922).
When using neuroimaging derivatives, please also cite Sadil et al. (2024).