Elsevier

NeuroImage

Volume 62, Issue 3, September 2012, Pages 2140-2150
NeuroImage

Parallel-transmission-enabled magnetization-prepared rapid gradient-echo T1-weighted imaging of the human brain at 7 T

https://doi.org/10.1016/j.neuroimage.2012.05.068Get rights and content

Abstract

One of the promises of Ultra High Field (UHF) MRI scanners is to bring finer spatial resolution in the human brain images due to an increased signal to noise ratio. However, at such field strengths, the spatial non-uniformity of the Radio Frequency (RF) transmit profiles challenges the applicability of most MRI sequences, where the signal and contrast levels strongly depend on the flip angle (FA) homogeneity. In particular, the MP-RAGE sequence, one of the most commonly employed 3D sequences to obtain T1-weighted anatomical images of the brain, is highly sensitive to these spatial variations. These cause deterioration in image quality and complicate subsequent image post-processing such as automated tissue segmentation at UHF.

In this work, we evaluate the potential of parallel-transmission (pTx) to obtain high-quality MP-RAGE images of the human brain at 7 T. To this end, non-selective transmit-SENSE pulses were individually tailored for each of 8 subjects under study, and applied to an 8-channel transmit-array. Such RF pulses were designed both for the low-FA excitation train and the 180° inversion preparation involved in the sequence, both utilizing the recently introduced kT-point trajectory. The resulting images were compared with those obtained from the conventional method and from subject-specific RF-shimmed excitations. In addition, four of the volunteers were scanned at 3 T for benchmarking purposes (clinical setup without pTx). Subsequently, automated tissue classification was performed to provide a more quantitative measure of the final image quality.

Results indicated that pTx could already significantly improve image quality at 7 T by adopting a suitable RF-Shim. Exploiting the full potential of the pTx-setup, the proposed kT-point method provided excellent inversion fidelity, comparable to what is commonly only achievable at 3 T with energy intensive adiabatic pulses. Furthermore, the cumulative energy deposition was simultaneously reduced by over 40% compared to the conventional adiabatic inversions. Regarding the low-FA kT-point based excitations, the FA uniformity achieved at 7 T surpassed what is typically obtained at 3 T. Subsequently, automated white and gray matter segmentation not only confirmed the expected improvements in image quality, but also suggests that care should be taken to properly account for the strong local susceptibility effects near cranial cavities. Overall, these findings indicate that the kT-point-based pTx solution is an excellent candidate for UHF 3D imaging, where patient safety is a major concern due to the increase of specific absorption rates.

Highlights

► We explore the potential of parallel-transmission for human brain MRI at high field. ► Different transmission methods at 7 T are compared to the standard MP-RAGE at 3 T. ► Parallel transmission improves image quality while reducing energy deposition. ► Contrast losses due to RF inhomogeneities vanish with kT-point tailored pulses. ► KT-points applied in MP-RAGE at 7 T produce T1-contrasts equivalent to standard 3 T.

Introduction

The magnetization-prepared rapid gradient echo sequence (Mugler and Brookeman, 1990), referred to as “MP-RAGE”, is among the most commonly employed 3D sequences to obtain T1-weighted anatomical images of the brain. To this end, typically an inversion pulse is used followed by a spoiled fast low angle shot (FLASH) train acquiring one partition plane in k-space per repetition (TR). Careful adjustment of the delay between the inversion and the acquisition block, as well as of the usual imaging parameters, allows excellent contrast between gray matter (GM), white matter (WM) and Cerebrospinal fluid (CSF) at field strengths up to 3 T (Deichmann et al., 2000, Han et al., 2006, Mugler and Brookeman, 1990, Mugler et al., 1992).

Strong GM/WM/CSF contrast is a prerequisite for both manual and automated tissue segmentation. Analysis based on such segmented data allows the study and potentially diagnosis of various pathological conditions. Measurements of change in GM, WM and CSF volumes for example provide profound insights into the aging brain (Gur et al., 1991) and pathologic progression in neurodegenerative diseases such as Alzheimer (Silbert et al., 2003) and Huntington's disease (Thieben et al., 2002). Moreover, thinning of the cerebral cortex is often region-dependent, stressing the importance of high quality highly-resolved images. Aside from its applications in pathology, the MP-RAGE sequence has become the current standard for anatomical reference images in functional studies such as in fMRI, PET, EEG and MEG.

In the quest for ever more detailed images of the human brain, increasingly higher main-magnetic field strengths are being explored to reap the benefits of increased signal and contrast (Ocali and Atalar, 1998). Already at 3 T, the RF wavelength corresponding to the proton Larmor frequency becomes comparable to the dimensions of some imaged human body parts. Consequently, non-uniformities arise in the transmit sensitivity (B1+), which result in zones of shade and losses of contrast distributed across the images of large organs such as the abdomen or thighs (Bernstein et al., 2006). Considering the application of the MP-RAGE sequence to human brain imaging, a combination of adiabatic inversion-pulses (Bernstein et al., 2004) and hard excitations can still be considered adequate to obtain high quality T1-weighted images at 3 T (Han et al., 2006). Moving up in field strength to 7 T and beyond, the increased Larmor frequency results in strong transmit sensitivity (B1+) variations throughout the volume of the human brain (Yang et al., 2002). Consequently, contrast artifacts form due to unwanted spatial variations of the flip-angle (FA). Coincidentally, the uniformity of the receive sensitivity (B1) deteriorates with the decrease in wavelength. This results in additional unwanted signal intensity variations (bias-field) throughout the image.

Various methods have been proposed to alleviate both the deteriorated T1 contrast and spatial B1 variation in MP-RAGE images. These include post-processing techniques exploiting the relatively low spatial frequencies in the bias-fields to remove unwanted signal intensity variations (Ashburner and Friston, 2005, Styner et al., 2000, Wald et al., 1995). Alternatively, an adaptation on the MP-RAGE sequence, referred to as the MP2RAGE sequence, has been proposed in the form of a second acquisition block following the inversion (van de Moortele et al., 2009). Combined with a suitable post-processing scheme, the images produced by the two acquisition blocks can be combined to produce high-quality bias-field-corrected T1-weighted images. Although this method is less sensitive to B1+ non-uniformities, the contrast may not be recovered in severely affected areas. At 7 T, depending on the RF coil in use, such areas may typically reside in the cerebellum, potentially extending out into the occipital or temporal lobes (Marques et al., 2010, Van de Moortele et al., 2009). Post-processing-based correction of these contrast artifacts induced by strong B1+ non-uniformities requires a priory knowledge of the anatomical structures, therefore defeating the general purpose of most T1-weighted imaging techniques (Wang et al., 2005). Considering the foreseeable continued increase in field strength, several 9.4 T scanners readily in operation and the development of an 11.7 T human MRI system at our site, increased B1+ artifacts may be expected, possibly limiting the applicability of the MP-RAGE and MP2RAGE approaches. Furthermore, the inclusion of the second acquisition block in the MP2RAGE significantly lengthens the TR (Marques et al., 2010). Thus high parallel imaging acceleration factors must be employed, at the expense of SNR, to achieve acceptable acquisition times (Griswold et al., 2002, Pruessmann et al., 1999, Sodickson and Manning, 1997).

Parallel transmission (pTx) was proposed to alleviate these limitations (Katscher et al., 2003, Zhu, 2004). This technique utilizes multiple independently-driven coil elements distributed around the subject. In its simplest form, referred to as RF-Shimming (Adriany et al., 2005), the B1+-fields from all coil elements are combined to optimize the B1+-distribution in a region of interest (ROI). Similarly to the conventional circularly-polarized (CP) mode utilized by most clinical systems today, a single pulse shape is transmitted at any particular time point in the pulse sequence. However, the RF-Shimming technique allows the amplitude and phase combination between coil-elements to be tailored to a specific ROI. Further generalization of this concept led to the introduction of Transmit-SENSE (Katscher et al., 2003, Zhu, 2004), exploiting the full potential of the transmit-array by tailoring the RF-waveforms to be applied to each of the individual coil-elements. This transmission generally occurs in concert with magnetic field gradients (usually used for spatial selectivity) so as to provide additional degrees of freedom to maximize the final excitation uniformity. Thus the RF pulses are played while pursuing a trajectory in k-space, where k is the spatial frequency vector: k(t) =  γ  tTG(τ), T being the RF pulse duration and G(t) the controlled linear magnetic field gradient at time t.

In the framework of Transmit-SENSE, whole-brain non-selective uniform excitations were recently demonstrated at 7 T with the kT-point method introduced by the authors (Cloos et al., 2012). This technique proposes a minimalist transmit k-space coverage in the vicinity of the k-space center to compensate for the smooth RF inhomogeneities without depositing high levels of energy. The few k-space locations where RF is transmitted are the so-called kT-points. The objective of this manuscript is to extend this work beyond the small-tip-angle regime (Pauly et al., 1989a) and to validate the applicability of kT-point pulses to T1-weighted whole brain imaging in the MP-RAGE sequence. This would demonstrate the ability of such pulses to retain the desired contrast in short-TR sequences (here we refer to the FLASH inter-echo time), where earlier attempts have shown their limitations (Boulant, 2009). In addition, the in-vivo demonstration of pTx-based inversion pulses would provide a prospective on their potential to replace SAR-intensive adiabatic pulses in UHF MRI.

The RF energy control is an important issue, as it will directly impact the SAR level in the patient. To facilitate our demonstration, an approach to SAR management for parallel-transmission will be presented in the next section, which while conservative, is less restrictive than the commonly adopted worst-case approach (Brunner et al., 2009, Collins et al., 2007).

Experimental results obtained at 7 T with both the conventional and proposed pTX methods were compared to images acquired at 3 T without pTX. Subsequently, the potential impact of the attained image quality on volumetric and morphological studies is quantified by analysis of the outcome from automated tissue classification and evaluation of the cortical ribbon.

Section snippets

Setup

Four volunteers were scanned at 7 T to evaluate the applicability of the proposed methods in the MP-RAGE sequence. For each subject, the first 30 min was reserved for low resolution (5-mm isotropic) transmit and receive sensitivity measurements as detailed in (Cloos et al., 2012). This was followed by a low-resolution (3D) T2*-weighted FLASH acquisition to define, with the aid of the brain extraction tool from the FSL software package (Smith, 2002), the three-dimensional ROI on which pulse design

Pulse design

Based on the B1+-maps measured in all eight volunteers, full Bloch simulations were performed to evaluate the theoretical performance of the inversion pulses (Fig. 3a). On average, the 7-ms hyperbolic secant adiabatic inversion applied to the conventional CP-mode resulted in a 38 ± 6% NRMSIE (cerebrum only: 22 ± 8%). Similar to what is achieved at 3 T (Fig. 4a), typically good inversion is established in the parietal and frontal lobes. However, extremely poor results are obtained in the cerebellum,

Pulse design

Due to the low-energy demand of the kT-point RF pulses, the absence of relaxation-related contrast deterioration (Boulant, 2009) confirms their applicability as excitations in short-TR echo trains such as used in the MP-RAGE sequence. Furthermore, their extension beyond the small-tip-angle domain provided inversion pulses with excellent spatial uniformity throughout the human brain at UHF, not only outperforming the conventional adiabatic hyperbolic secant pulses (Silver et al., 1985) in terms

Conclusions

We have demonstrated the soundness of the kT-point method in the framework of parallel transmission applied to T1-weighted 3D-imaging of the human brain at 7 T. The derived inversion pulses provided excellent spatial uniformity throughout the human brain (NRMSIE ~ 5%), outperforming the adiabatic pulses played in the conventional CP-mode (NRMSIE ~ 38%) and in subject-specific RF-shim (~ 18%), while simultaneously reducing the cumulative energy deposition. Thus the proposed method showed that the

Acknowledgment

The authors thank U. Fontius, H-P. Fautz and A. Vignaud from Siemens for the helpful discussions contributing to this work. This work was funded by the Iseult/Inumac French-German project.

References (56)

  • M.A. Bernstein et al.

    Imaging artifacts at 3.0 T

    J. Magn. Reson. Imaging

    (2006)
  • N. Boulant et al.

    Method for monitoring safety in parallel transmission systems based on channel-dependent average powers

  • D.O. Brunner et al.
  • M.A. Cloos et al.

    Specific absorption rate monitor for in-vivo parallel transmission at 7 Tesla

  • M.A. Cloos et al.

    Local SAR reduction in parallel excitation based on channel-dependent Tikhonov parameters

    J. Magn. Reson. Imaging

    (2010)
  • M.A. Cloos et al.

    K(T)-points: short three-dimensional tailored RF pulses for flip-angle homogenization over an extended volume

    Magn. Reson. Med.

    (2012)
  • C.M. Collins et al.

    A conservative method for ensuring safety within transmit arrays

  • G. Eichfelder et al.

    Local specific absorption rate control for parallel transmission by virtual observation points

    Magn. Reson. Med.

    (2011)
  • Ferrand, G., Luong, M., France, A., 2010: Résonateur Linéaire d'une Antenne Haute Fréquence pour Appareil d'Imagerie...
  • H.W. Fischer et al.

    Nuclear relaxation of human brain gray and white matter: analysis of field dependence and implications for MRI

    Magn. Reson. Med.

    (1990)
  • B. Fischl et al.

    Measuring the thickness of the human cerebral cortex from magnetic resonance images

    Proc. Natl. Acad. Sci. U. S. A.

    (2000)
  • W.A. Grissom et al.

    Spatial domain method for the design of RF pulses in multicoil parallel excitation

    Magn. Reson. Med.

    (2006)
  • W.A. Grissom et al.

    Joined design of dual-band large-tip-angle RF and gradient waveforms in parallel excitation

  • M.A. Griswold et al.

    Generalized autocalibrating partially parallel acquisition (GRAPPA)

    Magn. Reson. Med.

    (2002)
  • R. Gur et al.

    Gender differences in age effect on brain atrophy measured by magnetic resonance imaging

    Proc. Natl. Acad. Sci. USA

    (1991)
  • International Electrotechnical Commission

    International standard, medical electrical equipment — part 2: particular requirements for the safety of magnetic resonance equipment for medical diagnosis

    (2010)
  • U. Katscher et al.

    Transmit SENSE

    Magn. Reson. Med.

    (2003)
  • K. Kuroda et al.

    Temperature mapping using water proton chemical shift obtained with 3D-MRSI: feasibility in vivo

    Magn. Reson. Med.

    (1996)
  • Cited by (0)

    Grant sponsor: Iseult/Inumac French-German Project.

    View full text