Real-time feedback optimization of z-shim gradient for automatic compensation of susceptibility-induced signal loss in EPI

Abstract

Signal loss in gradient-echo echo planar imaging (GE-EPI) due to susceptibility-induced magnetic field inhomogeneity makes it difficult to assess the blood oxygenation level-dependent (BOLD) effect in fMRI investigations. The z-shim method that applies an additional gradient moment is one of the more popular methods of compensating for GE-EPI signal loss. However, this method requires a calibration sweep scan and post-processing to identify the optimal z-shim gradients, which slows down fMRI experiments. This study attempts to decrease the calibration time by introducing a real-time feedback framework. Creating a feedback loop between the image processing and the GE-EPI pulse sequence converts the calibration of z-shim gradients to an optimization problem, which can be accelerated by local search methods. This study proposes an interleaved scan that allows the simultaneous optimization of two z-shim gradient moments and allocates sufficient processing time for networking and computation. The z-shim compensated images obtained by the proposed real-time method are comparable to those created by the sweep method. The optimization procedure for obtaining negative and positive gradient moments generally requires about twenty GE-EPI repetitions. In conclusion, the proposed z-shim method includes an automated real-time framework to achieve a significant reduction in susceptibility-induced signal loss in GE-EPI with a minimal increase in calibration time. The proposed procedure is fully automatic and compatible with conventional GE-EPI and can thus serve as a pre-adjustment module in EPI-based fMRI researches.

Introduction

Gradient-echo echo planar imaging (GE-EPI) (Mansfield, 1977) is a fast acquisition technique that is commonly employed in functional magnetic resonance image (fMRI) because it provides a high sensitivity to blood oxygen level-dependent (BOLD) signals (Ogawa et al., 1990). However, brain images acquired with GE-EPI often exhibit significant signal loss in areas near air-tissue or bone-tissue interfaces due to susceptibility-induced local magnetic field inhomogeneity. These areas include the ventral prefrontal, temporal lobes, and orbitofrontal cortex (Lipschutz et al., 2001, Ojemann et al., 1997, Stenger et al., 2000). This inhomogeneity accelerates intravoxel dephasing and substantially attenuates the signal, especially in high magnetic fields.

Researchers have proposed several methods of compensating for susceptibility-induced signal loss by z-shimming, tailored RF pulse, high-resolution imaging, reducing TE using parallel imaging with phased array coils, or employing local shimming coils (Bellgowan et al., 2006, Chen and Wyrwicz, 1999, Frahm et al., 1993, Hsu and Glover, 2005, Schmidt et al., 2005, Wilson et al., 2003, Wong and Mazaheri, 2004). Z-shimming, which is based on adjusting the refocusing gradient in the slice selection direction, is one of the more popular methods in fMRI exams (Constable and Spencer, 1999, Cordes et al., 2000, Deichmann et al., 2002, Frahm et al., 1988, Gu et al., 2002, Ordidge et al., 1994, Song, 2001). Adding additional gradient moments allows through-plan spin refocusing, which reduces the signal loss caused by intravoxel incoherent phases. A general z-shimming exam acquires several images with different strengths of refocusing gradient in the slice direction (referred to as the z-shim gradient hereafter). These images are then combined using a composite formula such as direct sum (Frahm et al., 1988), weighted sum, maximum intensity projection (MIP)(Constable, 1995, Marshall et al., 2009), or the square root of the squared sum method (SSQ)(Ordidge et al., 1994). The optimal strength of a z-shim gradient is generally determined by a discrete global search procedure. A series of GE-EPI scans (i.e., a sweep scan) is acquired with stepping the z-shim gradients (Constable, 1995, Deichmann et al., 2002, Frahm et al., 1988). A post-processing method then calculates the optimal z-shim gradients for subsequent fMRI experiments. However, sweep scans and post-processing delay the overall fMRI studies.

A real-time feedback mechanism for advanced MRI systems may be a solution to this problem. Researchers have used real-time feedback methods, which adjust the scan parameters or the MR pulse sequences according to the acquired signal, to solve MRI problems. Application examples include, but are certainly not limited to, prospective motion correction (McConnell et al., 1997, Zaitsev et al., 2006), stabilization of frequency drifts (Wu et al., 2007), and the determination of scan parameters (Xie et al., 2010). This study proposes a real-time feedback method that automatically and rapidly determines the optimal z-shim gradients for GE-EPI experiments. This method employs the same pulse sequence (i.e., GE-EPI) as the following scans and requires neither user interactions nor pulse sequence interruption. Thus, the proposed method can serve as a pre-adjustment module for z-shim GE-EPI sequences.