Investigating the accuracy and precision of TE-dependent versus multi-echo QSM using Laplacian-based methods at 3 T.

Emma Biondetti,Anita Karsa,Karin Shmueli,David L Thomas

doi:10.1002/mrm.28331

Abstract

Multi-echo gradient-recalled echo acquisitions for QSM enable optimizing the SNR for several tissue types through multi-echo (TE) combination or investigating temporal variations in the susceptibility (potentially reflecting tissue microstructure) by calculating one QSM image at each TE (TE-dependent QSM). In contrast with multi-echo QSM, applying Laplacian-based methods (LBMs) for phase unwrapping and background field removal to single TEs could introduce nonlinear temporal variations (independent of tissue microstructure) into the measured susceptibility. Here, we aimed to compare the effect of LBMs on the QSM susceptibilities in TE-dependent versus multi-echo QSM. TE-dependent recalled echo data simulated in a numerical head phantom and gradient-recalled echo images acquired at 3 T in 10 healthy volunteers. Several QSM pipelines were tested, including four distinct LBMs: sophisticated harmonic artifact reduction for phase data (SHARP), variable-radius sophisticated harmonic artifact reduction for phase data (V-SHARP), Laplacian boundary value background field removal (LBV), and one-step total generalized variation (TGV). Results from distinct pipelines were compared using visual inspection, summary statistics of susceptibility in deep gray matter/white matter/venous regions of interest, and, in the healthy volunteers, regional susceptibility bias analysis and nonparametric tests. Multi-echo versus TE-dependent QSM had higher regional accuracy, especially in high-susceptibility regions and at shorter TEs. Everywhere except in the veins, a processing pipeline incorporating TGV provided the most temporally stable TE-dependent QSM results with an accuracy similar to multi-echo QSM. For TE-dependent QSM, carefully choosing LBMs can minimize the introduction of LBM-related nonlinear temporal susceptibility variations.

Highlights

IntroductionProcessing pipelines for QSM usually require unwrapping to resolve spatiotemporal 2 aliasing, removing background field variations (ΔBBg) resulting from χ sources outside the brain, and solving a local field (ΔBLoc)-to-χ ill-posed inverse problem.[2]
Quantitative susceptibility mapping aims to determine the underlying spatial distribution of tissue magnetic susceptibility (χ)[1,2] from the phase ( ) of a gradient-recalled echo (GRE) MRI sequence:(r, through multi-echo (TE)) = ΔBTot (r, TE) TE + 0 (r) (1)where TE is the echo time; r is the voxel position in the image; is the proton gyromagnetic ratio; ΔBTot is the χ-induced total field perturbation along the scanner’s z-axis; and 0 is a phase offset at TE = 0.Processing pipelines for QSM usually require unwrapping to resolve spatiotemporal 2 aliasing, removing background field variations (ΔBBg) resulting from χ sources outside the brain, and solving a local field (ΔBLoc)-to-χ ill-posed inverse problem.[2]Both single-echo and multi-echo GRE acquisitions can be used for QSM
For studies on the TE-dependent variation of χ, the Laplacian-based methods (LBMs) in the QSM processing pipeline should be carefully chosen to minimize biasing the results with LBM-related temporal variations

Summary

Introduction

Processing pipelines for QSM usually require unwrapping to resolve spatiotemporal 2 aliasing, removing background field variations (ΔBBg) resulting from χ sources outside the brain, and solving a local field (ΔBLoc)-to-χ ill-posed inverse problem.[2]. Both single-echo and multi-echo GRE acquisitions can be used for QSM. By acquiring multiecho GRE images and processing each TE separately, a method referred to as TE-dependent QSM,[5] some studies[5,6,7] have investigated the TE dependence of the QSM susceptibility resulting from nonlinear temporal variations of the phase. Acquiring multi-echo GRE images and combining all TEs improves the SNR of the resulting field map[3,7] or QSM image[1,7] compared to acquiring single-echo GRE images, but the regional accuracy and precision of χ calculated using TE-dependent versus multi-echo QSM has undergone limited investigation

Objectives

Methods

Discussion

Conclusion