Susceptibility phase imaging with comparison to R2* mapping of iron-rich deep grey matter

Abstract

Magnetic resonance imaging with susceptibility phase is seeing increasing use, especially at high magnetic fields. Tissue susceptibility can produce unique phase contrast for qualitative or quantitative imaging of iron-rich deep grey matter. However, phase imaging has several established sources of error including inherent susceptibility field effects and artifacts from background phase removal. These artifacts have led to inconsistent findings in past works relating iron to phase in healthy deep grey matter. This study seeks to determine the relative artifactual contributions from inherent susceptibility fields and from high pass phase filtering, currently the most common and accessible background phase removal method. In simulation, phase is compared to a known susceptibility distribution, while R2* maps are used as the in vivo gold standard surrogate for iron in healthy volunteers. The results indicate phase imaging depends highly on filtering, structure size, shape and local environment. Using in vivo phase and R2* profiles, it is shown that different filtering values, commonly seen in the literature, can lead to substantially different phase measures. Correlations between phase and R2* mapping are shown to be highly variable between structures. For example, using a standard filter of 0.125 the slopes and correlation coefficients were 4.28 × 10⁻⁴ ppm*s and R = 0.88 for the putamen, 0.81 × 10⁻⁴ ppm*s and R = 0.08 for the globus pallidus, 5.48 × 10⁻⁴ ppm*s and R = 0.72 for the red nucleus, and − 14.64 × 10⁻⁴ ppm*s and R = 0.54 for the substantia nigra. To achieve the most effective correlation to R2* we recommend using a filter width of 0.094 for the globus pallidus and putamen and 0.125 for the substantia nigra and red nucleus. The baseline phase measure should be obtained directly adjacent to the substantia nigra, and red nucleus to yield the most accurate phase values as demonstrated in simulation and in vivo. Different regression slopes are seen between subROIs within structures suggesting that regional iron accumulation within a structure is best studied with subROIs between different subject groups, not differences in phase values relative to the overall phase in one structure. Phase imaging with the standard high pass filter method has the potential to differentiate subtle iron changes in pathological processes compared to normal tissues with more reliability if specific filter strengths and measurement areas are appropriately applied on a structure dependent basis.

Introduction

Phase susceptibility imaging and susceptibility-weighted imaging (SWI) have demonstrated sensitivity to brain iron (Haacke et al., 2004, Ogg et al., 1999), which has been shown to accumulate in neurodegenerative diseases such as Alzheimer's disease (Bartzokis and Tishler, 2000), Parkinson's disease (Baudrexel et al., 2010), and multiple sclerosis (MS) (Pinero and Connor, 2000). These imaging methods have been used for quantifying iron changes in deep grey matter (Haacke et al., 2009a, Ogg et al., 1999) and qualitatively for enhancing image contrast, particularly between MS lesions and normal tissue (Eissa et al., 2009, Haacke et al., 2009b, Hammond et al., 2008b). While transverse relaxation rate (R2 or R2*) mapping is sensitive to iron in normal individuals (Bermel et al., 2005), phase imaging should be both more sensitive to iron because it depends on subtle phase shifts rather than significant dephasing, and more specific since phase is not significantly affected by water content, which could be a confound in cases of neurodegeneration (Mitsumori et al., 2009).

Putative quantitative iron measures are seeing increasing use with phase imaging (Hammond et al., 2008b, Hopp et al., 2010, Xu et al., 2008), however, studies have not shown consistent reliability of phase imaging for iron measurement because phase is also confounded by certain physical factors including: the angle of brain structure to the Bo field (Schafer et al., 2009), neuronal fiber orientation (Lee et al., 2010b), myelin content (Duyn et al., 2007), calcium and phospholipid content (He and Yablonskiy, 2009), neighboring susceptibility sources (Wharton and Bowtell, 2010), and the type of background phase removal method (Haacke et al., 2004, Neelavalli et al., 2009, Wharton and Bowtell, 2010, Yao et al., 2009). By focusing on the iron-rich basal ganglia, where there are substantial deposits of non-heme iron, factors such as phospholipids, myelin and fiber orientation will contribute a smaller role bringing background phase removal and susceptibility field effects to the forefront.

Background phase removal is necessary to remove the global magnetic field variations created by the geometry of the head and air tissue interfaces, such as the nasal cavity, in order to provide access to the underlying field variations related to the local tissue environment. While new phase background removal methods are continually evolving, standard phase imaging with simple background phase removal through phase filtering has been used extensively throughout the short history of phase imaging (Haacke et al., 2007, Hopp et al., 2010, Wang et al., 2000) and in recent neurological studies (Gupta et al., 2010, Rossi et al., 2010, Szumowski et al., 2010, Zhang et al., 2010, Zhu et al., 2009). As well, the phase filtering approach has the advantage of being relatively easy to implement and is widely available on many clinical MRI systems. Although this method produces visually interpretable images, it can alter the true phase values in certain brain structures. The effect of filtering on phase images has implications in quantitative phase measurements because effects of filtering depend upon object shape and size. This requires an in-depth understanding of shape effects.

The effect of phase suppression from varying filter strength has been previously presented qualitatively (Hammond et al., 2008a), and quantitatively (Haacke et al., 2007, Pfefferbaum et al., 2009) which has led to one common standard filtering approach of utilizing ~ 12% of the image in a low pass filter, in order to suppress background global fields but attempt to retain local phase differences. The quantitative studies either did not examine deep grey matter or did not compare subsections of the structures between filter strengths. Since deep grey matter structures vary in shape and size, different parts of structures could be affected differently by filtering and this could have implications when examining iron accumulation patterns.

As well as phase filtering, susceptibility field effects also substantially affect phase images (de Rochefort et al., 2010, Deistung et al., 2008, Schafer et al., 2009). These dipolar field effects result from the susceptibility difference, Δχ, between the inner and external environment of a structure and phase effects are produced within and around structures. Considering a very simple spherical susceptibility distribution, the analytical solution for field effect changes is well known depending on Δχ inside of the spherical structure, and outside on Δχ and on the directional component 3cos²(θ) − 1, where θ is the angle to the main magnetic field. More geometrically complex susceptibility distributions require a numerical computation by multiplication of a dipole field in k-space (Schafer et al., 2009), which has demonstrated the directional, and nonlocal, field effects of more anatomically representative distributions.

In the human brain, R2* values have shown very high correlation to post mortem iron concentrations r = 0.9 (Langkammer et al., 2010). However, previous studies have correlated phase or R2* to predicted iron content of the basal ganglia with minimal success (Haacke et al., 2010, Wharton and Bowtell, 2010, Yao et al., 2009). These studies compared phase between different structures in the same individual, while examining the same structure across individuals would enable phase-iron correlation without the confounding effects of structure dependent filtering and field shift due to structure shape.

In this work, phase imaging is compared to quantitative R2* mapping across multiple volunteers to demonstrate the role of susceptibility fields, phase filtering and ROI placement in each iron-rich, deep grey matter structure. Phase variations are examined in simulation and in vivo experiments using a wide range of filters with clear separation of susceptibility field effects from filter reconstruction effects. Structure-dependent recommendations for filter size and ROI placement are provided. By quantifying the possible confounds of phase imaging in deep grey matter, a means for better interpretation of quantitative phase imaging is provided.