Streamline-Based Analysis User Guide

This document provides a user guide to conduct Streamline-Based Analysis (SBA) as presented in [Preprint/Paper].

Overview of the SBA pipeline. Multimodal MRI data are used to construct a study-specific tissue-unbiased template tractogram (which is first filtered and subsequently refined for each subject, while retaining streamline-to-streamline correspondence); this defines a tractogram-based coordinate system for streamline-wise analysis. Imaging measures are sampled and aggregated into streamline-wise metrics. Streamline similarity is used to enable both streamline-wise data smoothing and statistical enhancement within a permutation-based inference framework.

1. Pre-processing

The desirable pre-processing steps will vary depending on the specific dataset. Please, refer to external resources, such as this tutorial or this tutorial up to step 8.

2. Template construction

The specific strategy used to construct the template tractogram for the presented aplication closesy follows what described in a previous tutorial. An exception is made for the "fibre tracking" step.

In this case, we proposed the following approach:

• Raw tractogram generation: this step yields a 1M template tractogram

tckgen FOD_template.mif.gz tracks_raw_template.tck -angle 22.5 -power 1.0 -select 1M -seed_dynamic FOD_template.mif.gz -act 5tt_template.mif -backtrack -minlength 10 -maxlength 250 -max_attempts_per_seed 1000 

• Template tractogram filtering: this step allows filtering the template tractogram to reduce streamline redundancy

tcksift tracks_raw_template.tck FOD_template.mif.gz tracks_filtered_template.tck 

• Further filtering: at this stage the user may decide to implement further filtering according to criteria of choice (e.g. required WM intersection in each subject space)
• SIFT2 multipliers computation:

tcksift2 tracks_filtered_template.tck FOD_template.mif.gz tracks_filtered_template_sift2.txt

3. Template tractogram refinement

To minimise erroneous streamline sampling in subject space, SBA offers a back-tracking/re-tracking tool for tractograms warped to subject space. Please, refer to available community forum resources regarding tractograms registration.

tckbacktrack tracks_filtered_template_subjectspace.tck FOD_subject.mif.gz 5tt_subjectspace.mif.gz tracks_filtered_template_subjectspace_refined.tck 

Overview of the back-tracking/re-tracking process for one exemplar template streamline projected to subject space. In the middle, the streamline is overlaid on a tissue type segmentation image (Cerebro Spinal Fluid – CSF, cortical Gray Matter – cGM, subcortical Gray Matter – sGM, White Matter – WM). Tissue profiles are shown for both the raw subject space projected (left) and back-tracked/re-tracked (right) versions of the streamline.

Qualitative comparison between template streamlines projected to subject space (“raw”) and their “refined” back-tracking/re-tracking version in four magnified regions from a slice of a representative subject. Streamlines are overlaid on tissue segmentation to provide high anatomical contrast (Cerebro Spinal Fluid – CSF; cortical Gray Matter – cGM; White Matter - WM). The refinement effect of this procedure can be observed in the improvements of regions with lack of WM coverage, and elimination of spurious cGM and CSF incursions.

4. Streamline-wise metric computation

SBA can flexibly accommodate various metrics, depending on the research question of interest. The only requirement is that one value per streamline is achieved. In the [Preprint/Paper] the along-streamline Fibre Density and bundle Cross section is used (sFDC). Please refer to steps 11-17 in the Fixel-Based Analysis tutorial.

fixel2tsf FD.mif tracks_filtered_template_subjectspace_refined.tck FD_samples.tsf
tsfinfo FD_samples.tsf -ascii_combined FD_samples_combined.txt
# Python script available in SBA/aux_files/
python mean_fixelsamples.py FD_samples_combined.txt FD_samples_mean.txt

fixel2tsf FC.mif tracks_filtered_template_subjectspace_refined.tck FC_samples.tsf
tsfinfo FC_samples.tsf -ascii_combined FC_samples_combined.txt
# Python script available in SBA/aux_files/
python mean_fixelsamples.py FC_samples_combined.txt FC_samples_mean.txt

# compute the sFDC metric
paste -d '*' FC_samples_mean FD_samples_mean.txt | bc > sFDC.txt

5. Streamline-streamline similarity computation

This step produces a streamline-streamline similarty matrix that is necessary for downstream processing steps. This command employs piping to provide a template grid that is 3mm isotropic based on the T1-weighted template image (it can be another one, as only the dimensions of the volume are used rather than voxel values).

mrgrid T1w_template.nii.gz regrid -voxel 3 - | tcksimilarity tracks_filtered_template.tck - similarity_matrix

6. Streamline data smoothing

This step produces smoothed streamline-wise data values as optimised via ROC-based evaluation:

tckfilter sFDC.txt sFDC_smoothed.txt similarity_matrix -tck_weights_in tracks_filtered_template_sift2.txt

7. Streamline-wise statistical inference with Similarity-informed Streamline Enhancement (SSE)

tckssestats in_streamline_directory subjects.txt design.txt similarity_matrix tracks_filtered_template_sift2.txt stats_results -nonstationarity_intrinsic -ttest contrast.txt