Batch Processing for Automatic Fiber Tracking

Introduction


Started from the new version from April 2020, DSI Studio provides batch automatic fiber tracking. The method does not use T1W parcellation. It only uses DWI data and the tractogrphy atlas to realize fiber tracking and target track selection. For technical details, please refer to this paper.

The following is the steps to run automatic fiber tracking:


Step T1 Generate SRC files


Please refer to Read DICOM, NIFTI, Bruker 2dseq, or Varian FDF Files to create SRC files.

Step B4 Select tracks and parameters

In the main window, select [Batch Processing][B4: automatic fiber tracking] to bring up the automatic fiber tracking interface.


Step B4(a): Specify SRC Files


Click on the [Open] button to select SRC files for reconstruction and fiber tracking. You can also select FIB files (change file type when browsing files) to skip the reconstruction steps.


Step B4(b): Select Bundles


Select the bundle to be tracked. 

Step B4(c): Parameters


GQI Length ratio: If SRC files are assigned in Step B4(a), GQI reconstruction will be done using the assigned length ratio (details about this parameter see Diffusion MRI Reconstruction in DSI Studio). If FIB files are assigned, then this parameter is ignored.
Resample DWI: IF SRC files are assigned in Step B4(a), DSI Studio will rotate the DWI volume to align with the MNI orientation and interpolate it to 2mm or 1mm resolution before GQI reconstruction. If FIB files are assigned, then this parameter is ignored.
Tract/Voxel Ratio: The number determines the number of streamlines be mapped. A higher value will lead to much longer computation time, whereas lower value will result in incomplete mapping.
Error Tolerance: Individuals may have track different a lot from the atlas. This distance value decides whether to consider a valid streamline. A higher value will allow for more complete mapping with the increasing risk of including false connections.
Pruning Count: Number of iteration to carry out "topology-informed pruning" (Yeh, 2019). Higher counts allow for removing more false connections, but it may also remove thin pathways if the streamline counts are not sufficient enough.

Output Statistics: Checked this to output track-related statistics including shape metrics and diffusion metrics.
Output Trk files: Specify whether to output tractography
Default mask: Check this to use default mask assigned by DSI Studio. If no check, when no mask is used (more computation time, but won't lose brain regions).
Overwrite Previous Tract Output: Check this to overwrite any existing files (otherwise, DSI Studio will skip the analysis

Run Fiber Tracking


Click on the [Run] button to start processing. You may stop the process at any time and click again to continue (if [Overwrite Previous Track Output] is unchecked).

DSI Studio will generate track files and statistics that can be inspected in [Step T3 Fiber Tracking]



Troubleshooting


Automatic fiber tracking may fail to generate tracks if the data are not correctly reconstructed, or if the data has a quality problem.

If "no result" happens very often (e.g. > 50%), then likely there is a post-processing problem such as (1) b-table flipped, (2) image volume flipped. The following video shows how to check the quality of your DWI data:


Possible solutions to quality issues are the following:

1. run quality control for SRC files: http://dsi-studio.labsolver.org/Manual/Parse-DICOM#TOC-Batch-Quality-Control-for-SRC-files and remove problematic dataset.
2. For b-table problems, uncheck the "check b-table" function at [Step T2 Reconstruction], and manually flip the b-table in y-direction if you use bval and bvec from FSL.

If "no result" happens only in very few subjects' (< 5%) small pathways (e.g. right AF, SLF I, corticopontine tract...etc)(see examples reported in 1,2), then likely your data acquisition is not perfect enough to map those small tracts for all subjects. 

The following is a list of common acquisition issues that may lead to poor fiber tracking results:

1. nonisotropic spatial resolution (e.g., slice thickness much larger than in-plane resolution)
2. b-value not enough (e.g., b-value less than 1,500)
3. insufficient diffusion sampling directions (e.g., less than 100 directions)

If you still have concerns or unexpected issues, please upload the problematic SRC file(s) using the upload link provided on the left navigation bar and notify me at the Discussion forum.

(Optional) Add New Tracts

The tracking list can be added by adding new atlas tracks. To do so, download HCP1065.1mm.fib.gz from brain.labsolver.org, and open it in Step T3 Fiber Tracking. Then use [Tract][Open Tracts] to load tractography atlas "icbm152.tt.gz" located in the \atlas\icbm152 folder (included in the DSI Studio Windows program. For Mac, right-click on dsi_studio.app and "show content" to find it inside the MacOS folder). You can track new pathways, add to the existing list, and save all pathways together as a new tractography atlas using [Tracts][Save Tracts][Save All Tracts As...].



Comments