Repeatability of B₁⁺ Inhomogeneity Correction of Volumetric (3D) Glutamate CEST via High-Permittivity Dielectric Padding at 7T

Abstract

Purpose:

Ultra-high field MR imaging lacks B₁⁺ inhomogeneity due to shorter RF wavelengths used at higher field strengths compared to human anatomy. Chemical Exchange Saturation Transfer (CEST) techniques tend to be highly susceptible to B₁⁺ inhomogeneities due to a high and uniform B₁⁺ field being necessary to create the endogenous contrast. High-permittivity dielectric pads have seen increasing usage in MR imaging due to their ability to tailor the spatial distribution of the B₁⁺ field produced. The purpose of this work is to demonstrate that dielectric materials can be used to improve glutamate weighted CEST (gluCEST) at 7T.

Methods:

GluCEST images were acquired on a 7T system on six healthy volunteers. Aqueous calcium titanate pads, with a permittivity of approximately 110, were placed on either side in the subject’s head near the temporal lobes. A post-processing correction algorithm was implemented in combination with dielectric padding to compare contrast improvement. Tissue segmentation was performed to assess the effect of dielectric pads on gray and white matter separately.

Results:

GluCEST images demonstrated contrast enhancement in the lateral temporal lobe regions with dielectric pad placement. Tissue segmentation analysis showed an increase in correction effectiveness within the gray matter tissue compared to white matter tissue. Statistical testing suggested a significant difference in gluCEST contrast when pads were used and showed a difference in the gray matter tissue segment.

Conclusion:

The use of dielectric pads improved the B₁⁺ field homogeneity and enhanced gluCEST contrast for all subjects when compared to data that did not incorporate padding.

Introduction

Magnetic Resonance (MR) imaging at 7T field strengths allows for significant increases in spatial resolution and signal-to-noise ratio (SNR) relative to lower field strength counterparts, such as 1.5 or 3T, especially in neuroimaging^1,2. However, at 7T, due to the shortened radiofrequency (RF) wavelengths in comparison to human anatomy, inherent B₁⁺ field inhomogeneities are a more significant problem than at lower field strengths. This is primarily due to the wave-like behavior of the RF magnetic field within the tissue volume of interest that disrupts the homogenous field distribution.

Chemical Exchange Saturation Transfer (CEST) techniques can image specific labile protons by saturating the chemical species of interest, with a spectrally selective saturation pulse and allowing that saturation to spontaneously transfer to water over time. This process of saturation transfer is continuously repeated, leading to a buildup of saturation in water magnetization, which can then be directly measured^3–6. However, since this process relies on the B₁⁺ field to create endogenous but transient contrast, CEST imaging is significantly more vulnerable to B₁⁺ inhomogeneities, like those that exist at ultra-high B₀ fields, in comparison to other MR imaging sequences^7,8. Glutamate-weighted CEST (gluCEST) is particularly challenging in this regard, due to the fast exchange of its protons with surrounding water. It requires both an ultra-high B₀ field (≥7T) and a high and uniform B₁⁺ strength across the tissue to generate adequate contrast. As data acquisition has moved from single-slice (two-dimensional: 2D) to volumetric slabs (three-dimensional: 3D), the absolute spread of B₁⁺ amplitude over the field of view increases and, because one now needs to collect more points in k-space per saturation pulse, the minimum B₁⁺ amplitude required to maintain glutamate derived contrast at an acceptable level increase as well^9,10.

One of the primary ways in which B₁⁺ inhomogeneity has been mitigated recently has been through the use of parallel transmit (PTX). There have been many studies that have employed PTX based approaches to correct for B₁⁺ inhomogeneity^11–13, even in the case of CEST imaging, however, it has not been shown that PTX can maintain a high absolute B₁⁺ power over the entire field of view. This would suggest that a multi-facetted correction approach may be necessary to better utilized other techniques in combination with PTX. B₁⁺ correction via post-processing algorithms is another way of compensating for inhomogeneities. In previous work, “calibration” gluCEST data is acquired and fit to a phenomenological function to parametrize correction surfaces over space (B₁⁺, T₁)¹⁴. In another study, similar “calibration” data is acquired at coarsely sampled B₁⁺ values across measured B₁⁺ maps. In this way separate correction curves could be derived for separate gray and white matter gluCEST segmentations¹⁵.

In recent years, there has been an increasing application of high-permittivity dielectric materials in MR imaging to tailor the spatial distribution of the B₁⁺ field and improve SNR within the tissue volume of interest. Neuroimaging studies, both at 3T and 7T, having shown improvement in B₁⁺ homogeneity and SNR include surrounding the head with water pads¹⁶, positioning aqueous calcium titanate padding near the temporal lobes and cerebellum^17,18, and creating a dielectric helmet for whole cortex enhancement¹⁹. Other applications include using dielectric padding composed of aqueous manganese salts to correct for abdominal imaging artifacts²⁰ and the use of barium titanate padding to improve SNR for imaging of the inner ear²¹, temporomandibular joint²² and femoral arteries²³. In the present work, high-permittivity dielectric pads, composed of aqueous calcium titanate (7TNS, Multiwave Imaging, Marseille, France), were used to augment the B₁⁺ field amplitude during the image acquisition process in conjunction with a post-processing image correction algorithm and hence gain enhanced gluCEST contrast. While dielectric pads have been adopted in several 7T imaging studies, it is believed that this is the first time that they have been used to improve brain gluCEST imaging at ultra-high field strengths.

Theory

CEST theory

Chemical exchange saturation transfer (CEST) is an MR imaging technique used to image certain solute compounds of interest via the exchange of protons with the bulk water protons present in the tissue volume³. In the imaging sequence, a frequency-selective RF pulse is used to saturate exchangeable protons in the solute, where there is then a proportional decrease in the water signal because of the saturated protons accumulating in the bulk water pool of the tissue. The amine proton on glutamate, which resonates at 3ppm downfield from water, exchanges at an ideal rate to be imaged using this technique (gluCEST) and can be further improved using 7T and higher field strength MR systems^5,7.

Bloch simulations can allow for an understanding of why sufficient B₁⁺ power is necessary for CEST imaging given the proton exchange rate for a metabolite of interest. Figure 1 illustrates that, in contrast to the more widespread technique of amide proton transfer (APT) imaging, accurate gluCEST imaging requires a high and constant B₁⁺ power. Figure 1A shows the simulated CEST contrast from saturation at 3.5ppm as a function of B₁⁺ strength for different concentrations of amide protons. This illustrates that CEST at 3.5ppm will detect differences in [APT] for a broad range of B₁⁺ power, starting at approximately 0.7μT. This sensitivity is quantified in Figure 1C, where the slope of the difference in CEST signal from two concentrations is plotted against the B₁⁺ power used. Figure 1B, analogous to Figure 1A, shows CEST contrast at 3.0ppm (gluCEST) and illustrates its ability to detect differences in the concentration of glutamate. In sharp contrast to APT CEST measurements, B₁⁺ strengths below 1.5μT are insensitive to glutamate; they almost exclusively reflect differences in water and lipid properties. Figure 1D illustrates that optimal gluCEST contrast can be achieved using the highest possible B₁⁺. However, currently, B₁⁺ values above 4μT cannot be used in human experiments due to specific absorption rate (SAR) limitations.

Figure 1.

a) Simulated CEST contrast from saturation at 3.5ppm (targeting protein amide protons, APT) as a function of B₁⁺ strength for different concentrations of amide. The plot illustrates that CEST at 3.5ppm will detect differences in [APT] comparably for a broad range of B₁ power. This sensitivity is quantified in c), where the slope difference between CEST from a 40mM source is plotted against the B₁. b) CEST contrast at 3.0ppm (gluCEST). d) Optimal gluCEST signal is only possible with high B₁. This figure was adapted from Cember et al¹⁴.

Dielectric Padding

Dielectric materials with high permittivity values can augment existing B₁⁺ values via two main mechanisms: 1) The addition of materials to reduce the elliptical shape of the tissue volumes can disperse the B₁⁺ fields more evenly and 2) displacement currents that are created within the dielectric materials create a secondary magnetic field which can add constructively to the primary field from the RF coil²⁴.

In reference to the second mechanism, the displacement current density that exists within the dielectric material can be quantified via the equation below [EQN 1]. This equation shows the total current density (J_T) within a material is comprised of conduction currents (J_C) and displacement currents (J_D). It can be seen the conduction current is dependent on the conductivity (σ) value of the material, while the displacement current depends on the relative permittivity value (ε_r), frequency (ω), and permittivity of free space (ε_o). This implies that at a sufficiently high ω or with sufficiently high ε_r, the displacement current can become a significant contributor to the main magnetic field created by the RF coil²⁵.

$$J_T=J_C+J_D= \sigma E+i\omega {\epsilon }_o{\epsilon }_{r}E$$ (1)

These high permittivity materials are typically manufactured as a powder, with the internal grain size being an important factor in the macroscopic field altering properties. These powders can be mixed into an aqueous solution (usually with water or deuterium oxide), done here with deuterium oxide, to reduce the background MR signal.

Methods

Dielectric pads

Two 18 × 18 cm² dielectric pads, having a composition of calcium titanate (CaTiO₃) suspended in deuterium oxide (7TNS, Multiwave Imaging, Marseille, France) with an estimated ε_r of 110¹⁸, were utilized in this study.

MR Imaging data acquisition

All images were acquired on a 7T system (MAGNETOM Terra, Siemens Healthcare, Erlangen, Germany) using a single-volume transmit 32-channel receive phased-array head coil (Nova Medical, Wilmington, MA, USA). Six healthy volunteers were scanned with informed consent under an approved institutional regulatory board protocol. All scans were performed with and without dielectric padding placed on either side of the subject’s head near the temporal lobes, as seen in Figure 2. Five repeat imaging experiments were also performed on a single-subject, with and without dielectric padding, to show intra-subject repeatability. 3D gluCEST data was acquired using a modified 3D turbo fast low angle shot (FLASH) sequence, as previously described in Cai et al.⁵, with TR = 3.5ms, TE = 1.79ms, and shot TR = 6000ms. The 3D slab had a thickness of 24mm (12 slices with each slice being 2mm thick), with an acquired in-plane resolution of 1 × 1mm² and matrix size of 240 × 196. Each slice was acquired in three shots. Generalized auto calibrating partially parallel acquisitions (GRAPPA) was applied with an acceleration factor of 2. Magnetization preparation was achieved using a B_1rms of 3.06μT, and saturation durations of 800ms (eight 99ms Hanning shaped pulses with 1ms interpulse delay) were applied at offset frequencies of (±1.8, 2.1, 2.4, 2.7, 3.0, 3.3, 3.6, 3.9, 4.2ppm) relative to water.

Figure 2.

Image showing the placement of the two dielectric pads (white arrows) inside the head coil on either side of the subject’s head near the temporal lobes.

Water saturation shift referencing (WASSR) images²⁶ were collected from 0 to ±1.5ppm (step-size of 0.15ppm) with a saturation pulse of B_1rms = 0.29μT and a 200ms duration. These images were used to compute B₀ maps and had identical imaging parameters as those used for the gluCEST sequence described above. An MP2RAGE sequence, which is a 2D multi-slice Siemens product sequence, was used with the same spatial parameters as described for the gluCEST sequence to generate T₁ maps. Relative B₁⁺ maps were generated using square preparation pulses with flip-angles of 20°, 40° and 80°²⁷ and acquired with the same spatial parameters as the 3D gluCEST sequence.

Image post-processing and statistical analysis

GluCEST-weighted images were corrected for the B₀ field distribution using the acquired WASSR²⁶ and relative B₁⁺ field maps²⁷. A further post-processing correction strategy, as described in Cember et al.¹⁴, was utilized to correct gluCEST contrast and quantitative accuracy in images acquired with and without dielectric padding. This post-processing correction was utilized as a standard step when analyzing gluCEST data in later sections. A tissue segmentation analysis between gray and white matter was also performed to determine the influence of dielectric padding and post-processing on separate tissue types²⁸.

Mixed effect statistical models were performed to estimate the effect of padding on the change in B₁⁺ and gluCEST contrast enhancement with random effect for subject/scans to account for repeated measures.

All image analyses were performed in MATLAB (The Mathworks, Inc., Natick, MA, USA) and statistical analyses were performed in R.

Results

Intersubject reproducibility was established by acquiring B₁⁺ and gluCEST data across six healthy volunteers with and without pads. This can be illustrated in Figure 3 where B₁⁺ maps are presented for each subject showing an increase in signal in the temporal lobes resulting in greater homogeneity across the image. The resulting gluCEST images can be seen in Figure 4 across padding conditions and with and without post-processing. The resulting gluCEST image can be seen in Figure 4 with raw uncorrected data in column A (without dielectric pads) and column B (with dielectric pads); post-processing correction in column C (without dielectric pads) and column D (with dielectric pads). All subjects qualitatively showed an increase in gluCEST contrast, especially in the lateral temporal lobes, when dielectric padding and post-processing were implemented separately. Further enhancement was observed when both dielectric padding and post-processing were used in combination. It should be noted, uncorrected gluCEST data is shown here only as an example to illustrate the effect of only the dielectric pads being present. All remaining gluCEST data presented and analyses performed will only incorporate post-processing across pad conditions. Histograms presenting the distribution and fits of the different subject data can be seen in Figure 6. Figure 6A shows a narrowing B₁⁺ distribution, indicating an increase in homogeneity while Figure 6B shows a slight narrowing of the processed gluCEST contrast values. A mixed effect model to test for statical significance across the different subject data across pad conditions. The results showed no statistically significant differences present in the B₁⁺ values (p = 0.861) across pad conditions. However, suggestive evidence for significance was present for the post-processed gluCEST contrast values (p = 0.064) across pad conditions.

Figure 3.

Intersubject B₁⁺ reproducibility acquisitions performed on six subjects with and without dielectric padding. These images show an increase in relative B₁⁺ strength in the temporal lobes and a slight reduction in the center magnitude of the image, indicating an overall increase in homogeneity.

Figure 4.

Reproducibility gluCEST acquisitions performed on multiple subjects. GluCEST data acquired on six subjects without (first column) and with (second column) dielectric pads. A post-processing correction method was implemented on the same data acquired without (third column) and with (fourth column) dielectric padding. These images show substantial contrast improvement in the temporal lobes of all subjects when pads were used compared to data acquired without pads. A similar trend can be observed in the data with post-processing applied, in which temporal lobe contrast improvement can be seen when pads were used.

Figure 6.

Histogram and fit comparisons for B₁⁺ (a), total gluCEST contrast (b), gray matter segment gluCEST contrast (c), and white matter segment gluCEST contrast (d). Histogram distribution narrowing can be seen in the relative B₁⁺, reflecting an increase in homogenization, and in the total gluCEST and gray matter tissue segment gluCEST, reflecting an enhancement in contrast when pads are used. The histogram comparison for the white matter tissue segmentation shows no major difference when pads are used. These histograms incorporate all data acquired from the intersubject reproducibility scans.

A white/gray matter tissue segmentation analysis of the processed intersubject gluCEST data was performed and can be seen in Figure 5. Gray matter tissue showed a substantial improvement in contrast in the temporal lobes when pads were used in comparison to no pads. In the white matter tissue, dielectric pads did not have as great of an effect as in gray matter. A histogram comparison of the segmented gray and white matter gluCEST contrast distributions can be seen in Figure 6C and 6D respectively across pad conditions. Distribution narrowing can be observed in the gray matter data when pads are used while no major difference can be seen in the white matter distribution. The mixed effect model was used to test for significance and showed a statistically significant difference in the gray matter data across pad conditions (p = 0.006) while in the white matter data no significance was observed (p = 0.223). Average and standard deviation values are presented in Table 1 for relative B₁⁺ strength and gluCEST contrast for all subjects scanned, across padding conditions, and between gray and white matter tissue segmentation.

Figure 5.

GluCEST contrast maps showing the effect of dielectric padding on segmented gray and white matter tissue. The left-most column shows the gray and white matter tissue segments overlaid on an anatomical image acquired of the same region. The gray matter tissue segment (top row) shows an increase in gluCEST contrast in the temporal lobe regions when pads are used. The white matter tissue segment (bottom row) shows less enhancement than the gray matter tissue segment when pads are used.

Table 1.

Calculated overall and tissue segmented mean and standard deviation values for relative B₁ and gluCEST data acquired for all subjects under different padding conditions. Overall values were calculated from the relative B₁ and gluCEST contrast values across the whole brain slice while tissue segmented gray matter (GM) and white matter (WM) were calculated separately.

	Relative B1 Strength	gluCEST Contrast with Post-processing (%)
No Pads	0.63 (0.15)	7.27 (4.37)
No Pads GM	0.71 (0.31)	8.14 (4.04)
No Pads WM	0.60 (0.34)	7.19 (4.33)
Pads	0.63 (0.13)	7.42 (4.29)
Pads GM	0.65 (0.31)	7.85 (4.32)
Pads WM	0.59 (0.33)	7.04 (4.28)

Intrasubject repeatability was established by acquiring B₁⁺ and gluCEST data on the same subject for five scans. Results presented in Figure 7 show B₁⁺ maps (top row) and post-processed gluCEST maps (bottom row) for one of the same subject repeat scans. These results show an increase in B₁⁺ homogeneity and gluCEST contrast in the temporal lobes when pads were used versus absent. Histograms presenting the distribution and fit of B₁⁺ values show a narrowing of the distribution reflecting the increase in homogeneity. However, in comparison, the gluCEST contrast value histogram does not show the same degree of distribution change. Figure 8 shows all gluCEST image slices (apart from first and last slices due to the presence of phase artifacts) from one of the same subject repeat scans when pads were present versus absent. It can be seen qualitatively that when pads were present, the temporal lobes saw an increase in contrast. It should be noted that this contrast enhancement diminished slightly for lower slices. Average and standard deviation values are presented in Table 2 for relative B₁⁺ strength and gluCEST contrast across repeat scans and padding conditions. The same mixed effect model used previously, was performed on the same subject repeat data comparing padding conditions. The results showed no statistically significant differences present in the B₁⁺ values (p = 0.338) across pad conditions. However, a significant difference was present for the post-processed gluCEST contrast values (p = 0.0006) across pad conditions.

Figure 7.

An example same subject repeatability acquisition showing the effect of dielectric padding on the B₁⁺ homogeneity and gluCEST contrast. The left column shows data acquired with padding showing lack of B₁⁺ homogeneity and gluCEST contrast in the temporal lobes. The center column shows data acquired with padding showing an improvement in homogeneity and contrast. The upper right panel shows a histogram and fit comparison for relative B₁⁺ strength. The lower right panel shows a histogram and fit comparison for gluCEST contrast. These histograms incorporate all data from the intrasubject repeatability scans.

Figure 8.

An example multi-slice comparison from one intrasubject repeatability acquisition with and without padding. In this example, slice 2 corresponds to areas lower in the brain while slice 11 is 20mm higher. The greatest amount of temporal lobe correction can be seen in the upper slices with a slightly diminished effect for the lower slices. The first and last slices were omitted due to the presence of phasing artifacts.

Table 2.

Calculated mean and standard deviation values for all relative B₁ and gluCEST data for five separate scans on a single subject with and without dielectric padding. Mean and standard deviation values were calculated from relative B₁ and gluCEST contrast values across the whole brain slice.

		Relative B1 Strength	CEST Contrast with Post-Processing (%)
No Pads	Scan 1	0.63 (0.14)	7.42 (4.06)
	Scan 2	0.62 (0.16)	7.29 (4.28)
	Scan 3	0.62 (0.16)	7.28 (4.28)
	Scan 4	0.62 (0.16)	7.28 (4.29)
	Scan 5	0.61 (0.15)	7.35 (4.16)
Pads	Scan 1	0.61 (0.13)	7.77 (4.29)
	Scan 2	0.63 (0.13)	7.82 (4.31)
	Scan 3	0.63 (0.12)	7.54 (4.56)
	Scan 4	0.62 (0.13)	7.45 (3.95)
	Scan 5	0.61 (0.13)	7.29 (4.06)

Discussion

The aim of this work was to show that dielectric padding can be used to correct B₁⁺ inhomogeneities found in gluCEST images by altering the spatial distribution of the B₁⁺ fields via secondary magnetic field generation. These images were specifically acquired in a slice in line with the hippocampus as there tends to be significant signal drop off in this region. The data presented here have shown that gluCEST contrast can be improved using carefully placed calcium titanate dielectric padding. The relative B₁⁺ maps are reflective of this, showing not only an increase in signal in the temporal lobes but also a decrease in the center B₁⁺ intensity. This dielectric corrective effect was clearly seen across the images acquired across multiple subjects and on repeat scans of the same subject. Images where only the post-processing correction was performed showed that even though the image did show improvement, it could not compensate for all inhomogeneities present, particularly in the temporal lobe regions. Both dielectric pads and the post-processing correction had the effect of shifting the distribution of gluCEST values and narrowing the distribution of B₁⁺ values. The use of post-processing was seen to result in major distribution shifts while the dielectric pads had more of a distribution narrowing effect. When both were used in combination, the dielectric pads allow more of the acquired gluCEST data to fall into a “correctable” regime, leading to better final images. Based on Cember et al.¹⁴ the correctable B₁⁺ regime that the post-processing algorithm can accommodate for falls between 1.3 to 3.1μT. With gluCEST contrast being seen at nominal B₁⁺ saturations of 0.7μT (or about 40% nominal value), this post-processing can’t effectively correct for the lower B₁⁺ saturations seen in the uncorrected temporal lobes. Dielectric padding has proven useful in this case by creating a secondary B₁⁺ field to add to regionally lower magnitude primary field. So, while the dielectric pads were of great benefit in correcting inhomogeneities, they do not eliminate the need for adequate post-processing, as clearly illustrated in the work that originally presented this type of post-processing correction method¹⁴.

Also, while largely effective across the whole brain, a difference in dielectric pad effectiveness was observed in tissue segmentation analysis between gray and white matter regions. In combination with inherent B₁ inhomogeneities present at ultra-high fields, a portion of the present B₁ drop off was due to the oblique slice taken in alignment with the hippocampus. In this oblique slice, the anterior brain region was physically closer to the edge of coil where a lower B₁ field is present. The gray and white matter regions in this anterior region showed the most contrast improvement when dielectric pads were placed in proximity. However, in this case, the gray matter region showed a much greater improvement in contrast between pad and post-processing conditions than white matter. Within this work, reproducibility was demonstrated between subjects as well as repeatability between subjects with and without padding. This is in strong agreement with work done by Nanga et al.²⁹ showing the reproducibility of 2D gluCEST across six healthy volunteers and furthermore, GM and WM segmentations of this data showed a high level of reproducibility. This is an important point as the addition of dielectric pads in the present work did not alter gluCEST reproducibility that has been previously demonstrated by Nanga et al. Additionally, when evaluating the current gluCEST maps against an optimal standard, comparisons should made between the B₁⁺ maps and not the gluCEST maps themselves. This is due to slight regional variations in glutamate concentration being present across different brain regions and between subjects, whereas the B₁⁺ maps are a result of the RF coil and could in theory be further homogenized to a theoretical constant across the tissue volume. A good example of this can be seen in Figure 8, where the hippocampal region of the brain expectedly shows a much higher gluCEST contrast than some of the immediate surrounding areas.

As B₀ strength is increased in the future, the problem of B₁ inhomogeneity will only become further exacerbated. Therefore, robust correction methods, like those presented here, will only become more important. One technique that is expected to contribute to more effective correction techniques, but was not explored here, is the use of PTX. With PTX being on the cutting edge of MR technological development, this type of sequence integration would be a valuable addition to dielectric padding and post-processing corrections.

A common criticism of dielectric padding is the lack of flexible form factor brought on by the materials used. With the most used materials being calcium titanate (ε_r=110) and barium titanate (ε_r=150), large volume fractions of these materials must be used in aqueous solution to effectively produce secondary magnetic fields¹⁷. If a material with a much higher permittivity value was used instead, the volume fraction and overall amount of material, could be reduced while preserving the corrective effect. With a smaller volume fraction of material used thinner pads would be achievable resulting in a potential increase in subject comfort during scans. However, when exploring the usage of any new dielectric material careful consideration must be given to accurate characterization of the potential frequency dependence, especially at the higher frequencies used in ultra-high field imaging. Lastly, while repeatability and reproducibility of dielectric pad usage for gluCEST was established, one difficulty in using these pads is the lack of standard placement. The large size pads used here cover a sizable area of the head and can be used effectively with general placement. However, custom smaller padding with different materials may need more exact positioning. In the future, standard positioning coil inserts could accommodate mounted dielectric material pouches for increased accuracy in placement for not only head studies but also abdominal and skeletal muscle studies.

Conclusion

In this work, high-permittivity dielectric padding, composed of a calcium titanate solution, were used to augment the B₁ field amplitude and resulted in gluCEST contrast enhancement. A post-processing correction, implemented in combination with the dielectric padding, showed a further improvement in B₁ field uniformity and hence gluCEST contrast. In the future, continued use of dielectric materials with a higher relative permittivity value used in combination with parallel transmit technology and post-processing correction methods, could result in larger improvements in SNR and image quality.