Tracking

This tutorial walks through the steps to perform fiber tracking using DIPY. Multiple tracking methods are available in DIPY.

You can try these methods using your own data; we will be using the data in DIPY. You can check how to fetch the DIPY data.

Local Fiber Tracking

Local fiber tracking is an approach used to model white matter fibers by creating streamlines from local directional information. The idea is as follows: if the local directionality of a tract/pathway segment is known, one can integrate along those directions to build a complete representation of that structure. Local fiber tracking is widely used in the field of diffusion MRI because it is simple and robust.

In order to perform local fiber tracking, three things are needed:

1. A method for getting directions from a diffusion data set.

2. A method for identifying when the tracking must stop.

3. A set of seeds from which to begin tracking.

We will be using the stanford_hardi dataset for the local tracking command line interface demonstration purposes.

We will start by creating the directory to which to save the stopping file (e.g.: stop_file):

mkdir stop_file

The workflow for each local tracking method requires the paths to the peaks and metrics files, stopping files, and seeding files. The default value for the local tracking method to be used (specified through the tracking_method parameter) is eudx (EuDX tracking [1]). A number of optional arguments, such as the number of seeds per dimension within a voxel (seed_density), or the output directory (out_dir), can also be provided.

To get the stopping file, we will create a binary file of the GFA of the CSA reconstruction method and set a stopping criterion by calling the dipy_mask command:

dipy_mask recons-csa/gfa.nii.gz 0.25 --out_dir "stop_file"

We will use the white matter (WM) partial volume effects (PVE) map corresponding to the diffusion data being tracked, and available through the stanford_pve_maps dataset, as the seeding file (e.g.: pve_wm.nii.gz).

There are four tracking methods as described below.

EuDX Tracking

We will first create the directory in which to save the output tractogram file (e.g.: eudx_tracking_output):

mkdir eudx_tracking_output

Then, to perform the EuDX tracking we will run the dipy_track command as:

dipy_track recons-csa/peaks.pam5 stop_file/mask.nii.gz data/stanford_hardi/pve_wm.nii.gz --out_dir "eudx_tracking_output"

Deterministic Tracking

Deterministic maximum direction getter is the deterministic version of the probabilistic direction getter. It can be used with the same local models and has the same parameters. Deterministic maximum fiber tracking follows the trajectory of the most probable pathway within the tracking constraint (e.g. max angle). In other words, it follows the direction with the highest probability from a distribution, as opposed to the probabilistic direction getter which draws the direction from the distribution. Therefore, the maximum deterministic direction getter is equivalent to the probabilistic direction getter returning always the maximum value of the distribution.

Deterministic maximum fiber tracking is an alternative to EuDX deterministic tractography, and unlike EuDX, it does not follow the peaks of the local models, but uses the entire orientation distributions.

We will first create the directory in which to save the output tractogram file (e.g.: det_tracking_output):

mkdir det_tracking_output

Then, to perform the deterministic tracking we will run the dipy_track command as:

dipy_track recons-csa/peaks.pam5 stop_file/mask.nii.gz data/stanford_hardi/pve_wm.nii.gz --seed_density 2 --tracking_method "det" --out_dir "det_tracking_output"

Probabilistic Tracking

Probabilistic fiber tracking is a way of reconstructing white matter connections using diffusion MR imaging. Like deterministic fiber tracking, the probabilistic approach follows the trajectory of a possible pathway step by step starting at a seed; however, unlike deterministic tracking, the tracking direction at each point along the path is chosen at random from a distribution. The distribution at each point is different and depends on the observed diffusion data at that point. The distribution of tracking directions at each point can be represented as a probability mass function (PMF) if the possible tracking directions are restricted to discrete numbers of well distributed points on a sphere.

We will first create the directory in which to save the output tractogram file (e.g.: prob_tracking_output):

mkdir prob_tracking_output

Then, to perform the probabilistic tracking we will run the dipy_track command as:

dipy_track recons-csa/peaks.pam5 stop_file/mask.nii.gz data/stanford_hardi/pve_wm.nii.gz --seed_density 2 --tracking_method "prob" --out_dir "prob_tracking_output"

Closest Peaks Tracking

Closest peaks fiber tracking is a type of deterministic tracking. Deterministic streamlines follow a predictable path through the data by selecting the same diffusion orientation when evaluating the same point through the propagation process. There may be several estimated directions at each point (such as a voxel center): closest peaks tracking selects one of these estimated directions on the basis of closeness of match to the previous direction of the streamline.

We will first create the directory in which to save the output tractogram file (e.g.: closest_peaks_output):

mkdir closest_peaks_output

Then, to perform the closest peaks tracking we will run the dipy_track command as:

dipy_track recons-csa/peaks.pam5 stop_file/mask.nii.gz data/stanford_hardi/pve_wm.nii.gz --seed_density 2 --tracking_method "cp" --out_dir "closest_peaks_output"

Particle Filtering Tracking (PFT)

Particle Filtering Tracking (PFT) [2] uses tissue partial volume estimation (PVE) to reconstruct trajectories connecting the gray matter, and not prematurely stopping in the white matter or in the corticospinal fluid. It relies on a stopping criterion that identifies the tissue where the streamline stopped. If the streamline reaches the gray matter, the trajectory is kept. If the streamline incorrectly stopped in the white matter or in the cerebrospinal fluid, PFT uses anatomical information to find an alternative streamline segment to extend the trajectory. When this segment is found, the tractography continues until the streamline reaches the gray matter.

We will use the stanford_hardi dataset in DIPY to showcase this tracking method. As with any other workflow in DIPY, you can also use your own data!

We will first create a directory in which to save the output tractogram file (e.g.: pft_output):

mkdir pft_output

To run the Particle Filtering Tracking method, we need to specify the paths to the diffusion input file, white matter partial volume estimate, grey matter partial volume estimate, and cerebrospinal fluid partial volume estimate for tracking, and seeding file followed by optional arguments. In this case, we will be specifying the threshold for the Probability Mass Function that controls the randomness or probabilistic nature of the tracking (pmf_threshold), and the output directory (out_dir).

White matter, grey matter, and cerebrospinal fluid volume files are available through the stanford_pve_maps dataset.

The Particle Filtering Tracking is performed by calling the dipy_track_pft command, e.g.:

dipy_track_pft recons-csa/peaks.pam5 data/stanford_hardi/pve_wm.nii.gz data/stanford_hardi/pve_gm.nii.gz data/stanford_hardi/pve_csf.nii.gz data/stanford_hardi/pve_wm.nii.gz --pmf_threshold 0.5 --out_dir "pft_output"

This command will save the tractogram file to the specified output directory.

Overview of Fiber Tracking Methods

Fiber Tracking Method	Output
EuDX Tracking
Deterministic Tracking
Closest Peaks Tracking

References

[1]

Eleftherios Garyfallidis. Towards an accurate brain tractography. PhD thesis, Cambridge University, Cambridge, United Kingdom, 2012. URL:, doi:.

[2]

Gabriel Girard, Kevin Whittingstall, Rachid Deriche, and Maxime Descoteaux. Towards quantitative connectivity analysis: reducing tractography biases. NeuroImage, 98:266–278, September 2014. URL: https://doi.org/10.1016/j.neuroimage.2014.04.074, doi:10.1016/j.neuroimage.2014.04.074.