Evaluating consistency of deterministic streamline tractography in non-linearly warped DTI data

TL;DR

This study assesses whether deterministic tractography on non-linearly warped DTI data produces results comparable to native DTI, supporting its use for statistical analysis in normalized space despite potential reorientation imperfections.

Contribution

It provides a systematic evaluation of tractography consistency between native and warped DTI data, highlighting the viability of using warped data for analysis.

Findings

Tractography results in warped DTI are comparable to native DTI.

Small reorientation imperfections do not significantly affect tractography.

Warped DTI can be used reliably for statistical analysis.

Abstract

Tractography is typically performed for each subject using the diffusion tensor imaging (DTI) data in its native subject space rather than in some space common to the entire study cohort. Despite performing tractography on a population average in a normalized space, the latter is considered less favorably at the \emph{individual} subject level because it requires spatial transformations of DTI data that involve non-linear warping and reorientation of the tensors. Although the commonly used reorientation strategies such as finite strain and preservation of principle direction are expected to result in adequate accuracy for voxel based analyses of DTI measures such as fractional anisotropy (FA), mean diffusivity (MD), the reorientations are not always exact except in the case of rigid transformations. Small imperfections in reorientation at individual voxel level accumulate and could potentially affect the tractography results adversely. This study aims to evaluate and compare deterministic white matter fiber tracking in non-linearly warped DTI against that in native DTI. The data present promising evidence that tractography in non-linear warped DTI data could indeed be a viable and valid option for various statistical analysis of DTI data in a spatially normalized space.