Assessment of brain anatomy with gradient-echo contrasts: A comparison between magnitude, phase & R2* imaging with quantitative susceptibility mapping (QSM)


Journal article


A. Deistung, A. Schäfer, F. Schweser, K. Sommer, R. Turner, J. Reichenbach
2012

Semantic Scholar
Cite

Cite

APA
Deistung, A., Schäfer, A., Schweser, F., Sommer, K., Turner, R., & Reichenbach, J. (2012). Assessment of brain anatomy with gradient-echo contrasts: A comparison between magnitude, phase & R2* imaging with quantitative susceptibility mapping (QSM).

Chicago/Turabian
Deistung, A., A. Schäfer, F. Schweser, K. Sommer, R. Turner, and J. Reichenbach. “Assessment of Brain Anatomy with Gradient-Echo Contrasts: A Comparison between Magnitude, Phase & R2* Imaging with Quantitative Susceptibility Mapping (QSM)” (2012).

MLA
Deistung, A., et al. Assessment of Brain Anatomy with Gradient-Echo Contrasts: A Comparison between Magnitude, Phase & R2* Imaging with Quantitative Susceptibility Mapping (QSM). 2012.


Abstract

INTRODUCTION – Accurate assessment of extent and location of anatomical brain structures is essential for many imaging and surgical applications. Superb anatomical contrast has already been demonstrated with T2-weighted magnitude [1] and phase imaging [2, 3] as well as quantitative R2 mapping [4] at ultra-high magnetic fields. However, magnitude images are not quantitative, phase images suffer from strong non-local signal contributions and R2 images may be affected at locations of steep phase gradients. In contrast, quantitative magnetic susceptibility mapping (QSM) [5], a novel imaging technique that determines tissue magnetic susceptibility from the phase of complex valued gradient echo (GRE) data, provides quantitative and local anatomical contrast. This study compares quantitative susceptibility maps with conventional GRE imaging approaches (magnitude, phase, R2) with respect to anatomic tissue contrast at 7 T and provides quantitative susceptibility and R2 values. MATERIAL AND METHODS – Data Acquisition: For susceptibility mapping 3D single-echo GRE data sets (TE/TR/ FA/TA=10.5 ms/17ms/8°/16:57min:sec, BW = 140 Hz/px, voxel size=0.4x0.4x0.4mm3) of nine healthy volunteers were acquired on a 7T MR scanner (Magnetom 7T, Siemens Healthcare, Erlangen, Germany) at 3 different orientations of the head with respect to B0. Additionally, 3D multi-echo GRE imaging (TE1/TE2/TE3/TE4=5 ms/12.8 ms/20.6 ms/28.4 ms, TR/FA/BW = 34 ms/8°/160 Hz/px, voxel size of 0.8 × 0.8 × 0.8 mm3) was performed in normal position of the head. Data Processing: Single channel magnitude images were combined using the sum-of-squares method [6], whereas single channel phase images were combined according to [7]. For each subject combined magnitude and phase data sets were non-linearly registered to the dataset in normal position. Phase aliasing was resolved by 3D phase unwrapping [8] and background phase contributions were eliminated with the SHARP technique [5]. Background-corrected phase images were used to compute multi-angle (MA) susceptibility maps as described in [5]. Furthermore, backgroundcorrected phase images were divided by 2π·TE to yield frequency maps showing the Larmor frequency variation in Hz. The magnitude signal decay of the 3D multi-echo GRE examination was converted to quantitative R2 maps according to [9]. For direct comparison the R2 images were linearly registered to the susceptibility maps. Data Analysis: Volumes of interest (VOIs) were manually identified for deep gray matter structures (internal globus pallidus (GPi), external globus pallidus (GPe), putamen (Put), caudate nucleus (CN), subthalamic nucleus (STN), red nucleus (RN), substantia nigra (SN)), homogenous white matter structures (frontal and occipital regions) and corticospinal fluid (CSF). Contrast-to-noise ratio (CNR) was calculated for all image contrasts by calculating the differences in signal intensity between the specific VOI with respect to a VOI in homogeneous white matter in the frontal brain, divided by the noise (standard deviation of the homogeneous white matter VOI). Since noise in MR images scales with MR imaging parameters, e.g., voxel volume and number of averages, the CNR values obtained from susceptibility and R2 data were corrected to yield a similar noise level compared to a single GRE data set of the MA study. RESULTS & DISCUSSION – A comparison of image contrast within thalamic tissue between magnitude, frequency, susceptibility, and R2 images is presented in Figure 1. Magnitude contrast is nearly homogeneous within thalamic tissue and only allows clear identification of the pulvinar (not shown) and venous vessels. Although frequency images provide a heterogeneous signal pattern of thalamic tissue, interpretation is hampered by non-local signal contributions (see black arrow). Susceptibility and R2 images, however, provide a very heterogeneous, locally varying pattern of thalamic tissue that clearly enables identification of sub-regions and nuclei of the thalamus. However, the R2 images do not provide the same level of contrast and detail as the susceptibility images. The CNR values of different deep grey matter structures for the different images are summarized in Table 1, suggesting that deep gray matter structures are delineated best on susceptibility images. Furthermore, susceptibility differences of selected brain structures with respect to CSF as well as R2 values are listed in Table 1. The susceptibility differences are in good agreement with values reported in the literature for 7T [10] and 3T [5, 11]. The R2 values (Tab. 1) of the basal ganglia regions agree excellently with literature findings [4]. Susceptibility images provided an excellent and local contrast between white matter and cortical grey matter (Fig. 2) that was superior to the contrast in magnitude, frequency, and R2 images (Fig. 2e). Susceptibility maps also revealed sub-structures of the midbrain and basal ganglia that were not observed on corresponding magnitude, frequency, and R2 images (not shown). Although QSM based on the MA approach, as presented here, is not applicable in clinical routine, this contribution demonstrates the enormous potential of the susceptibility contrast for structural MRI. QSM may reveal a new contrast for automatic identification of (sub-)structures of the brain. REFERENCES – [1] Li TQ et al. NeuroImage. 2006;32(3):1032-40. [2] Deistung A et al. Magn Reson Med. 2008;60(5):1155-68. [3] Duyn JH et al. Proc Natl Acad Sci U S A. 2007;104(28):11796-801. [4] Yao B et al. NeuroImage. 2009;44(4):1259-66. [5] Schweser F et al. NeuroImage. 2011;54(4):2789-807. [6] Roemer PB et al. Magn Reson Med. 1990. 16(2):192-225. [7] Hammond KE et al. NeuroImage. 2008; 39(4):1682-92. [8] Abdul-Rahman HS et al. Appl Opt. 2007;46(26):6623-3. [9] Miller AJ and Joseph PM. Magn Reson Imaging. 1993;11(7):1051-6. [10] Wharton S and Bowtell R. NeuroImage. 2010;53(2), 515-525. [11] Liu T et al. Magn Reson Med. 2011;66(3), 777-83. Figure 1: GRE imaging of the thalamus. Transverse sections of magnitude, frequency, susceptibility, and R2 images are shown in (a), (b), (c), and (d), respectively. The susceptibility image clearly exhibits specific contrast within the thalamus enabling discrimination of subthalamic structures, such as a – internal medullary lamina, b – anterior nuclei group, c – ventral lateral nucleus group, d – lateral posterior nucleus, and e – mediodorsal nucleus. In the frequency image (b) such interpretation is impeded by non-local signal contributions, e.g., close-to veins (black arrow).