Rapid whole cerebrum myelin water imaging using a 3D GRASE sequence

Abstract

Myelin water imaging, a magnetic resonance imaging technique capable of resolving the fraction of water molecules which are located between the layers of myelin, is a valuable tool for investigating both normal and pathological brain structure in vivo. There is a strong need for pulse sequences which improve the quality and applicability of myelin water imaging in a clinical setting. In this study, we validated the use of a fast multi echo T₂ relaxation sequence for myelin water imaging. Using a multiple combined gradient and spin echo (GRASE) technique, we attain whole cerebrum myelin water images in under 15 minutes. Region of interest analysis indicates that this fast GRASE imaging sequence produces results which are in good agreement with pure spin echo measurements (R² = 0.95, p < 0.0001). This drastic improvement in speed and brain coverage compared to current spin echo standards will allow increased inclusion of myelin water imaging in neurological research protocols and opens up the possibility of applications in a clinical setting.

Introduction

Myelin water imaging (MWI) is an increasingly active field of magnetic resonance imaging (MRI) research. The ability to generate maps of the myelin water fraction (MWF) from a human brain in vivo enables clinicians and researchers alike to directly examine the myelination state of white matter in the central nervous system (CNS). As myelin is of great importance in the CNS and there are many known diseases which lead to degradation of the myelin sheath which surrounds white matter neurons (Barkovich, 2000), most notably multiple sclerosis (MS), MWI is a valuable tool in the investigation of the patho-physiological causes and possible treatments of white matter diseases.

The majority of MWF maps are created via multicomponent T₂ analysis. That is, decay curves from multi echo spin echo images exhibiting T₂ contrast are decomposed into discrete T₂ components to create a T₂ distribution (MacKay et al., 1994, Whittall et al., 1997) whereby no a priori assumptions are made about the number of contributing components. A short T₂ component (corresponding to water within the myelin sheath) may then be identified and its magnitude normalized to the total water signal, yielding a value for the MWF between zero and one. Post mortem MRI-pathology correlation studies in both CNS tissue (Kozlowski et al., 2008, Laule et al., 2006, Laule et al., 2008) and animal peripheral nerve (Webb et al., 2003) have demonstrated a good quantitative relationship between the MR-derived MWF and histological staining for myelin.

Spatially resolved data required for multi-component T₂ analysis are usually acquired with multi spin echo (MSE) based image sequences (MacKay et al., 1994, Poon and Henkelman, 1992). However, in order to prevent magnetization transfer effects from exciting multiple slices (Vavasour et al., 2000), these images should be acquired in 3D, potentially resulting in long data acquisition times. If MWI is to become a commonplace tool in clinical imaging or research that can be used in conjunction with other MR-imaging modalities that exhibit full brain coverage with isotropic voxel size, e.g. functional MRI, diffusion tensor imaging, magnetization transfer imaging, etc., whole brain data at reasonable spatial resolution must be acquired within clinically feasible scan durations (< 15 minutes).

A combined gradient and spin echo sequence known as gradient and spin echo (GRASE) (Feinberg and Oshio, 1991) has previously been used to accelerate clinical MR acquisitions (Fellner et al., 1995, Fellner et al., 1997, Melhelm et al., 1998, Patel et al., 1995, Rockwell et al., 1997, Umek et al., 1998). By acquiring multiple gradient echoes per refocusing pulse on either side of the spin echo, an accelerated acquisition trajectory may be realized which exhibits pure T₂ weighting in the k-space centre (spin echo signal) and T₂^⁎ weighting in the k-space periphery (gradient echo signal). Some researchers reported that GRASE was less effective in the detection of small or hypointense lesions (Patel et al., 1995, Umek et al., 1998), and produced images with lower signal to noise ratio (SNR) and less pronounced contrast (Fellner et al., 1995, Fellner et al., 1997). However, others found that GRASE showed better detection of lesions, especially those with low signal (Melhelm et al., 1998) or those which exhibited paramagnetic susceptibility characteristics (Rockwell et al., 1997). Although not necessarily in agreement about all factors, all authors point out that in circumstances where rapid scanning is indicated, GRASE is very useful. Rockwell et al. go as far as to say that GRASE is a potential replacement for turbo spin echo in routine MRI of the brain (Rockwell et al., 1997).

A similar gradient and spin echo sequence has also been used to collect data for multiexponential T₂ analysis (Does and Gore, 2000). This sequence demonstrated dramatic decreases in scan time while maintaining good agreement of multiexponential T₂ parameters compared to a conventional spin echo acquisition.

Mädler (Mädler and MacKay, 2007) introduced whole brain MWI based on 3D-GRASE. Due to both hardware and software limitations at that time, scan times required for whole brain coverage were on the order of 20 to 30 minutes. We were able to improve scan efficiency of a similar acquisition to less than 15 minutes. In this study we demonstrate whole cerebrum imaging in a scan time under 15 minutes and also compare myelin water fraction results from the 3D-GRASE sequence with results from the 3D-MSE technique (Mädler and MacKay, 2006).