Differential tractography for cross sectional data


The differential tractography for a cross-sectional study can be done in the native space or in the MNI space. Both approaches should give consistent results. The native-space analysis provides more individualized analysis, whereas the MNI-space analysis provides better group-level integration. The choice depends on the need of your study. 

Step 1: Reconstruct data

First, the MRI data have to be converted to *.SRC.GZ (see Read DICOM, NIFTI, Bruker 2dseq, or Varian FDF Files or Batch Processing using GUI) and make sure that the quality is good (see quality control procedures here). 

For native-space analysis, select all *.SRC.GZ files at [Step T2:Reconstruction] and specify T2b(1)=[GQI] at Diffusion MRI Reconstruction in DSI Studio. The reconstruction will generate *.FIB.GZ files.

For MNI-space analysis, use [Step C1: Reconstruct SRC files for connectometry] to select the folder that contains the SRC files (or a root folder with subfolders containing the SRC files). DSI Studio will reconstruct the SRC files and generate a FIB file for each of them in the same folder. 

Step 2: Compare Data

We will compare one subject's data with a group average image. 

The group average image can be the QA, nQA, GFA, ISO, FA NIFTI files from the HCP data at http://brain.labsolver.org/diffusion-mri-templates/hcp-842-hcp-1021

If you would like to create your own group-average images from a control group, open [Tools][O8: Create template/skeleton] and include the FIB files of the control subjects (reconstructed by [Step C1: Reconstruct SRC files for connectometry]to create a group-average FIB file. Updated DSI Studio (after September 2021) will also average other metrics (e.g., dti_fa, md, rd1, rd2 ...etc.). The group-average NIFTI images can be exported by opening the group-average FIB file in [Step T3 Fiber Tracking] and export metrics using [Export].  

In the following, we will use QA as an example. The same procedure can be applied to other metrics.

First, open the FIB.GZ file of a subject at [Step T3: Fiber tracking]
Load the group-average NIFTI file (e.g., HCP1065.QA.nii.gzusing [Slices][Insert MNI Images]. DSI Studio will apply nonlinear warping. The spatial normalization may take a while, and then a new metric will be added to the slice menu in the top middle toolbar using the file name of the group-average NIFTI file (e.g., HCP1065.QA).

Then compare voxel-wise difference using[Analysis][Add Tracking Metrics] and input HCP1065.QA-qa for mapping decrease of QA, or qa-HCP1065.QA for mapping increase of QA. Here "HCP1065.QA" is the file name of the group-average NIFTI file.

A new differential tracking metrics will be added to the [Step T3c: Options][Tracking Parameters][Differential Tracking][Metrics]

Step 3: Tracking the difference

1. Quality check using whole-brain tracking:

    In the tracking parameters (right upper window), set [Step T3c: Options][Tracking Parameters][Min length] to "30 mm",  [Terminate if] to "100,000 seeds", [Differential Tracking Index] to "none."

    Click the [Step T3d: Tracts][Fiber Tracking] button to see if you can get a good quality of whole-brain fiber tracking. 

    If there are too many noisy fibers, consider increasing the [Step T3c: Options][Tracking Parameters][Threshold] or [Min length].

    If missing too many branches, considers lowering the [Step T3c: Options][Tracking Parameters][Threshold] or [Min length].

    You may also need to adjust other parameters (check out Diffusion MRI Fiber Tracking in DSI Studio) until you get a good quality of the whole-brain track

2. Tracking difference:

    Clean up all tracks from the tracklist on the right lower window. (e.g. [Tracts][Delete Tracts][Delete all])

    In the tracking parameters, set [Step T3c: Options][Tracking Parameters][Differential Tracking][Metrics] from none to the newly added metrics "HCP1065.QA-qa".

    Adjust [Differential Tracking][Threshold], 0.1 means 10% increase or decrease, 0.2 means 20%. I would suggest starting with a 20% difference (i.e. set [Differential Tracking][Threshold]=0.2) because individual differences in anisotropy are around 10%~20%.

    If you have a target track to study (e.g., optic radiation), you may add ROIs to limit findings (see Fiber Tracking). For example, you can map the affected pathways that pass through the internal capsule by assigning the internal capsule as the ROI. The number and length of tracks can be compared between patients if other tracking parameters are fixed (be sure to fix the seed count).  

    Click on fiber tracking to map the differential stratigraphy. The next step is to test the reliability of the findings.


*Different scans may have different coverage at the cerebellum, and it is recommended to remove findings at the bottom slices.
*To know the anatomical structure of the findings, you can use findings as ROIs by [Tracts][Tracts to ROI] to run fiber tracking and then use [Tracts][Miscellaneous][Recognize Track] to determine the anatomical structures.


Step 4: False discovery rate and statistical testing (Optional)

One key question for differential tractography is the significance of the findings. For example, if we observe a lot of tracks showing up in differential tractography, how many of them are false positive? A way to quantifying this reliability is by calculating the false discovery rate from a group of patients and a group of control subjects. The following steps illustrate how this can be carried out in DSI Studio. 

Example 1: calculate the false discovery rate

1. Acquire pre-treatment and post-treatment scans for 5 patients and 5 controls.

2. Repeat differential fiber tracking for each patient and control (either individual v.s. individual or individual v.s. normative atlas). When tracking the difference, use the same number of seeds (e.g. 10,000) and same length for the minimum length threshold (e.g. 40 mm, longer length increases the specificity, whereas lower length threshold increases sensitivity) for all subjects.

3. Calculate false discovery rate: if you can get a total of 1000 tracks from 5 patients and 10 tracks from 5 controls, then the false discovery rate of the findings in patients is 10/100 = 10%. You can then show the tracks in patients and report the false discovery rate.

note: You can stratify the analysis by using different "dec" or "inc" thresholds. For example, using a "dec" of 0.05 means you are mapping affected tracks with more than 0.05 difference in SDF. Repeat the analysis using "dec" of 0.05, 0.10, 0.15 ... to show the overview of the affected brain networks. 

Example 2: test whether patients have longer affected pathways than controls.

1. 2. The same as the previous example.

3. For each patient and control, use [Tract][Statistics] to get the average length of the tracks. Then you can run a t-test to test whether the patients have "longer" affected pathways than controls.

[1] Yeh, Fang-Cheng, Pei-Fang Tang, and Wen-Yih Isaac Tseng. "Diffusion MRI connectometry automatically reveals affected fiber pathways in individuals with chronic stroke." NeuroImage: Clinical 2 (2013): 912-921.