Elsevier

NeuroImage

Volume 49, Issue 2, 15 January 2010, Pages 1601-1611
NeuroImage

Gradient distortions in MRI: Characterizing and correcting for their effects on SIENA-generated measures of brain volume change

https://doi.org/10.1016/j.neuroimage.2009.08.008Get rights and content

Abstract

Precise and accurate quantification of whole-brain atrophy based on magnetic resonance imaging (MRI) data is an important goal in understanding the natural progression of neurodegenerative disorders such as Alzheimer's disease and multiple sclerosis. We found that inconsistent MRI positioning of subjects is common in typically acquired clinical trial data – particularly along the magnet's long (i.e., Z) axis. We also found that, if not corrected for, the gradient distortion effects associated with such Z-shifts can significantly decrease the accuracy and precision of MRI-derived measures of whole-brain atrophy – negative effects that increase in magnitude with (i) increases in the Z-distance between the brains to be compared and (ii) increases in the Z-distance from magnet isocenter of the center of the pair of brains to be compared. These gradient distortion effects can be reduced by accurate subject positioning, and they can also be corrected post hoc with the use of appropriately-generated gradient-distortion correction fields. We used a novel DUPLO-based phantom to develop a spherical-harmonics-based gradient distortion field that was used to (i) correct for observed Z-shift-associated gradient distortion effects on SIENA-generated measures of brain atrophy and (ii) simulate the gradient distortion effects that might be expected with a greater range of Z-shifts than those that we were able to acquire. Our results suggest that consistent alignment to magnet isocenter and/or correcting for the observed effects of gradient distortion should lead to more accurate and precise estimates of brain-related changes and, as a result, to increased statistical power in studies aimed at understanding the natural progression and the effective treatment of neurodegenerative disorders.

Introduction

Precise and accurate quantification of ongoing whole-brain atrophy based on magnetic resonance imaging (MRI) data is an important goal in understanding the natural progression of neurodegenerative disorders such as Alzheimer's disease (AD) (Fox et al., 1996, Smith et al., 2007), and multiple sclerosis (MS) (De Stefano et al., 2007, Giorgio et al., 2008). It is also an important goal in understanding how the natural progression of such disorders can be affected by different treatments (Altmann et al., 2009, Anderson et al., 2007a, Filippi et al., 2004).

SIENA (Structural Image Evaluation using Normalisation of Atrophy) is a freely available software package that is widely used for the fully automated estimation of longitudinal changes in brain volume (Smith et al., 2001, 2002). SIENA provides an estimate of the percent brain volume change (PBVC) across two points in time for which appropriate MRI data are available for the same individual. SIENA-PBVC values have been shown to be practical outcome measures for monitoring treatment effects in patients with MS (Altmann et al., 2009, Anderson et al., 2007a, Filippi et al., 2004), and SIENA-PBVC values have been shown (i) to be highly accurate (with a median absolute error of about 0.15%) (Smith et al., 2002, 2007); (ii) to be largely independent of slice thickness (Smith et al., 2002, 2007); (iii) to have only about half of the measurement error of semiautomated seed-growing techniques (Sormani et al., 2004); (iv) to be highly correlated with boundary-shift-interval measures of brain atrophy (Fox et al., 1996) – both in patients with AD (Smith et al., 2007) and in those with MS (Anderson et al., 2007b); (v) to be highly predictive of the amount of “realistic” atrophy applied to simulate longitudinal brain-volume changes in patients with AD (Camara et al., 2008); and (vi) to be highly reproducible across analysis centers (Jasperse et al., 2007).

Unfortunately, the precision and accuracy of SIENA – as well as that of other such measures of global atrophy – can be affected by local volume changes related to a combination of the following three factors. First, the precision and accuracy of such measures can be affected by nonlinear gradient distortions (GD) that are often typical of newer generation scanner gradient systems that are designed to have short bores (in order to take up less space) and short gradient rise times (in order to acquire images faster for functional MRI, diffusion tensor imaging, and MRI of the heart) (Wang et al., 2004); importantly, as shown previously by Jovicich et al. (2006), the potential effect on morphometric analyses of the “barrel-shaped” distortions associated with such scanners seems to increase with Z-distance from magnet isocenter (see Fig. 4B for an example of such a barrel-shaped distortion). Second, they can be affected by inconsistent positioning of subjects within the scanner – particularly along the long (i.e., Z) axis of the magnet – and, as we will show in Part 1 below, such inconsistent positioning is very common in typically acquired clinical trial data. Third, they can be affected by typical canthomeatal (CM) alignment within the magnet, which, as illustrated in Fig. 1, results in an individual's cerebrum being centered several centimeters further into the magnet than isocenter. As a result of these three factors, inadvertent Z-shifts of several centimeters into the magnet would result in the bulk of the brain moving even further away from isocenter, where it would experience even greater effects of GD; on the other hand, Z-shifts of similar extent out of the magnet would result in the bulk of the brain moving closer toward isocenter, where it would experience lesser effects of GD (see Fig. 1).

In the present body of work, we will examine four main issues. First, we will describe the extent of variability that was found in the Z-positioning of 100 patients with MS who underwent repeated MRI-scanning sessions as part of a recent multicenter clinical trial. Second, we will describe a novel, DUPLO-based phantom that we have used in order to characterize, and correct for, the GD-field associated with our 1.5-T Siemens Sonata MRI scanner. Third, we will describe the effect of actual Z-shifts on SIENA-generated PBVC values – both before and after correcting for GD – in a sample of nine normal control (NC) subjects who underwent a series of same-day MRI acquisitions. And, fourth, we will describe the results of a series of simulations based on the data from these same nine NC subjects: simulations aimed at examining the GD effects on SIENA-generated PBVC values that might be expected with a greater range of Z-shifts than those that we were able to actually acquire.

Section snippets

Methods

In order to estimate the extent of variability in Z-positioning that might be expected to be found if accurate Z-positioning is not strictly controlled for, we undertook a post hoc analysis of 815 typically acquired T1-weighted MRI scans from a sample of 100 patients with MS who were scanned as part of a recent multicenter clinical trial. Each patient had an initial baseline scan and up to eight subsequent follow-up scans over the course of 48 weeks. These T1-weighted images were acquired with

DUPLO-based phantom

Our approach to characterizing, and correcting for, the GD associated with our 1.5-T Siemens Sonata MRI scanner involved a novel phantom made out of 125 2 × 4 DUPLO plastic building blocks (19 × 32 × 64 mm each), which were arranged in a regular pattern of 11 layers containing 9 to 12 blocks each (about 209 × 192 × 192 mm total). It was assembled inside a NALGENE 8-L plastic container filled with a water solution of 0.15 mM/L MnCL2 and 2.8 g/L NaCL (Fig. 4A). Importantly, such a phantom can be easily and

Methods

The T1-weighted CM-Scan MRIs of the 9 NC subjects in whom we studied the effects of actual and simulated Z-shifts on SIENA-generated measures of brain atrophy are shown in Fig. 7. The anatomical location of magnet isocenter for all subjects is also shown relative to the ICBM-152 image in Fig. 2. Although there was some variability in the 9 NC scans' X- Y- and Z-positioning, it is much less than that which was found in the 815 clinical trial scans described in Part 1. The extent of variability

Methods

In order to examine the effects on SIENA-PBVC values of a greater range of Z-shifts than those that we were able to actually acquire, we carried out a series of simulations based on the same MRI data from the nine NC subjects described above. For each subject, the effect of Z-shift was simulated in 5-mm steps from -50 mm to +50 mm as follows: (i) our phantom-based GDC field was applied to their intensity-corrected and neck-cropped CM-Scan T1-weighted data, (ii) these GD-corrected data were Z

General discussion

In the present body of work, we have shown that inconsistent MRI positioning of subjects is quite common in the clinical trial setting – particularly along the magnet's long (i.e., Z) axis. If not corrected for, the GD effects associated with such inadvertent Z-shifts can significantly decrease the accuracy and precision of MRI-derived measures of brain atrophy. Importantly, these negative effects seem to increase in magnitude with (i) increases in the Z-distance between the brains to be

Acknowledgments

This study was supported by grants from (i) the Canadian Institutes of Health Research; (ii) the Multiple Sclerosis Society of Canada; and (iii) CLUMEQ (Consortium Laval, Université du Québec, McGill and Eastern Quebec), which is funded in part by the Natural Sciences and Engineering Research Council of Canada (Major Resources Support Program), Le Fonds québécois de la recherche sur la nature et les technologies, and McGill University. The authors would like to express their gratitude to (i)

References (20)

There are more references available in the full text version of this article.

Cited by (0)

View full text