Novel tractography analysis

Differential tractography


Differential tractography tracks the exact segment with a change in anisotropy to reveal pathways with axonal injury.

Conventional tractography maps all existing pathways regardless of whether they have changed their diffusion property, and thus it is insensitive to subtle diffusion differences until the entire axonal loss happens at the very late stage. In comparison. difference tractography [8] adopts a novel “tracking-the-difference” paradigm. This is realized by first conducting a voxel-level comparison between longitudinal scans to filter out regions with no substantial difference. The axons with substantial differences can then be tracked and enhanced for further localization and quantification 


Correlational tractography (connectometry)



Introduction


Correlational tractography is a tractography modality that shows trajectories of pathways correlated with a study variable. The method to derive correlation tractography and test its reliability is called "connectometry." Connectometry includes a fiber tracking routine and permutation test to provide false discovery rate of the findings.

Definition of connectivity

The connectivity is defined by a new anisotropy measure called quantitative anisotropy (QA). QA has a different biophysical meaning from FA (see Diffusion MRI Metrics for the discussion). The analysis is not limited to QA, and DSI Studio allows users to use any voxel-wise metrics as the connectivity measure for connectometry analysis.

How connectometry works

Connectometry adopts a "tracking the correlation" paradigm, which is fundamentally different from the conventional analysis paradigm of "finding the difference in tracks or regions". It uses a nonparametric permutation test to find out the association of white matter pathways with any study factor [1][pdf]. It utilizes a simple linear regression that includes all relevant variables (e.g. sex, site difference) in the model. The results will show "correlational tractography", which only maps the exact segment of pathways exhibiting positive or negative correlation with a variable of interest.

The analysis first identifies voxels that have strong correlations and tracks along axonal fiber directions to identify the consecutive fiber segment that shows a continuous positive and negative correlation. The tracking is only limit to regions with high correlation, thus minimizing the noise contributed from other non-correlated regions.

Connectometry can also be used in a longitudinal study to study the difference between paired data (e.g. pre- and post-treatment). Group connectometry uses a linear regression model and can include other variables to consider in the regression. Group connectometry uses a linear regression model and can include other variables to consider in the regression.


Tracking instead of clustering


Connectometry is conceptually identical to task-based fMRI analysis with the FWE-corrected cluster: The voxel-wise BOLD signals in fMRI are replaced by fiber-wise QA in connectometry. The task sequence in fMRI is replaced by study variables (age, sex, s-weight...etc.) of the subjects. The GLM in fMRI is then reduced to simple linear regression in connectometry because there is no time series. Each voxel in fMRI will have a T-score from GLM, whereas each fiber in each voxel will also have a T-score from the linear regression in connectometry. The clustering of voxels in fMRI (aggregate if T larger than a threshold) is replaced by fiber tracking in connectometry (track if T larger than a threshold). The cluster size in fMRI for significance is replaced by track length of the fiber tracking results.

The statistical testing for the "track length" used a different strategy to avoid the recent "cluster failure" issue in fMRI. In connectometry, the analysis uses a permutation test that permutes the subject label to derive a null distribution of the track length. This allows for testing the track length results using a false discovery rate that considers multiple comparisons.

The final results can be called "correlational tractography", which is a whole-brain, prior-free finding revealing the exact segment of pathways correlated with a study variable (with other covariates considered).

Is connectometry a better approach than existing methods?

Most studies used either track-based or region-based analysis to compare diffusion data along the tracks or within a given region. These approaches require users to assign pre-defined tracks or regions to sample diffusion indices for further analysis. The assigned regions or tracks, however, will inevitably include regions that have low SNR (near gray matter) or show no correlation at all (irrelevant branches), consequently bringing high variance and making results insignificant.

Connectometry adopts an entirely different mindset to avoid including irrelevant regions. The paradigm behind it can be summarized as "tracking the correlation". This is fundamentally different from the conventional paradigm of "finding the difference in tracks or regions". The analysis first identifies voxels that have strong correlations and tracks along axonal fiber directions to identify the consecutive fiber segment that shows a continuous positive and negative correlation. The tracking is only limit to regions with high correlation, thus minimizing the noise contributed from other non-correlated regions.



Pros:

1. Connectometry identifies only the affected segment of the fiber tracks
2. Allows for multiple regression, group comparison, paired group comparison.

Cons:
1. Cannot work on distorted brain structure (e.g. brain tumor)
2. Relatively poor at identifying short-ranged focal difference.


Exercises:

1. Download the sample MS patient data at https://pitt.box.com/v/differential-tractography and find the injury track using Differential tractography.
2. Download CMU60 connectometry database at https://pitt.box.com/v/course-group-cnt run Group connectometry analysis to find connections correlated to BMI.

Reference:
[1] Yeh, Fang-Cheng, David Badre, and Timothy Verstynen. "Connectometry: A statistical approach harnessing the analytical potential of the local connectome." NeuroImage 125 (2016): 162-171.
[2] 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.




Comments