A Signal Peak Separation Index for axisymmetric B-tensor encoding

TL;DR

This paper introduces a theoretical framework and a Signal Peak Separation Index (SPSI) to evaluate how axisymmetric B-tensors in diffusion MRI can be optimized for better fascicle orientation sensitivity and tractography, especially in crossing-fascicle regions.

Contribution

It proposes a new SPSI metric to assess B-tensor sensitivity to fascicle orientations and provides insights into how different B-tensor shapes affect signal robustness and orientation estimation.

Findings

Linear encoding maximizes B-tensor anisotropy and orientation sensitivity.

Oblate B-tensors can yield higher signals and noise robustness at the same SPSI.

The SPSI relates B-tensor properties to tissue microstructure, guiding sequence design.

Abstract

Diffusion-weighted MRI (DW-MRI) has recently seen a rising interest in planar, spherical and general B-tensor encodings. Some of these sequences have aided traditional linear encoding in the estimation of white matter microstructural features, generally by making DW-MRI less sensitive to the orientation of axon fascicles in a voxel. However, less is known about their potential to make the signal more sensitive to fascicle orientation, especially in crossing-fascicle voxels. Although planar encoding has been commended for the resemblance of its signal with the voxel's orientation distribution function (ODF), linear encoding remains the near undisputed method of choice for orientation estimation. This paper presents a theoretical framework to gauge the sensitivity of axisymmetric B-tensors to fascicle orientations. A signal peak separation index (SPSI) is proposed, motivated by theoretical considerations on a simple multi-tensor model of fascicle crossing. Theory and simulations confirm the intuition that linear encoding, because it maximizes B-tensor anisotropy, possesses an intrinsic advantage over all other axisymmetric B-tensors. At identical SPSI however, oblate B-tensors yield higher signal and may be more robust to acquisition noise than their prolate counterparts. The proposed index relates the properties of the B-tensor to those of the tissue microstructure in a straightforward way and can thus guide the design of diffusion sequences for improved orientation estimation and tractography.