Interleaved ³¹P MRS Imaging of Human Frontal and Occipital Lobes using Dual RF Coils Incorporated with Single-Channel Transmitter-Receiver and Dynamic B₀ Shimming

Abstract

In vivo ³¹P MRS provides a unique tool for noninvasively studying brain energy metabolism and mitochondrial function. Assessment of bioenergetic impairment in different brain regions is essential to understand the pathophysiology and progression of human brain diseases. This paper presents a simple and effective approach allowing interleaved measurement of ³¹P spectra and imaging from two distinct human brain regions of interest with dynamic B₀ shimming capability. A transistor-transistor logic controller was employed to actively switch the single-channel X-nuclear RF transmitter-receiver between two ³¹P RF surface coils enabling to interleaved acquisition of two ³¹P FIDs from human occipital and frontal lobes within the same repetition time. Linear gradients were incorporated into RF pulse sequence to perform the first-order dynamic shimming to further improve spectral resolution. The overall results demonstrate that the approach provides a cost-effective and time-efficient solution for reliable ³¹P MRS measurement of cerebral phosphate metabolites and ATP metabolic fluxes from two human brain regions with high detection sensitivity and spectral quality at 7 Tesla. The same design concept can be extended to acquire multiple spectra from more than two brain regions or can be employed for other MR applications beyond the ³¹P spin.

Graphical abstract

Illustration of dual-− coil technique enabling interleved ³¹P MRS imaging from human occipital lobe (Coil 1) and frontal lobe (Coil 2) at 7 Tesla with the capability of dynamic B₀ shimming for obtaining high quality 3D CSI data as demonstrated from single voxel spectra as well as the maps of PCr integral measured from the frontal lobe (left panel) and occipital lobe (right panel).

1 | INTRODUCTION

In vivo ³¹P magnetic resonance (MR) spectroscopy (MRS) provides a unique and non-invasive tool for assessing cerebral high-energy phosphate metabolites, phosphate lipid metabolism, intracellular pH and redox state of NAD (nicotinamide adenine dinucleotide) and neuroenergetics ^1–6. Moreover, the ³¹P MRS technique incorporated with the magnetization transfer (MT) preparation (³¹P-MT MRS) offers new utility enabling researchers to non-invasively examine the kinetics of key bioenergetic reactions catalyzed by the creatine kinase (CK) and adenosine triphosphate (ATP) synthase (ATP_ase) and their metabolic fluxes ^7,8. In this regard, in vivo ³¹P or ³¹P-MT MRS studies have shown the bioenergetic impairment specifically in the frontal lobe in psychiatric patients like Schizophrenia ⁹ and neurodegenerative diseases, like Alzheimer’s disease ^10,11.

With the increasing recognition of complex neurological disorders involving a behavioral and cognitive impairment commonly linking to dysfunction in specific brain regions ^12–16, examination of functional and metabolic impairments in multiple brain regions of interest is essential to advance the understanding of underlying pathophysiological mechanisms and to improve diagnosis and treatment of diseases.

To detect in vivo ³¹P spectra from multiple brain regions, one common approach is to employ a large size ³¹P radiofrequency (RF) coil in combination with three-dimensional (3D) chemical shift imaging (CSI) for measuring the ³¹P MRS signals across the whole brain ^17–19. However, the presence of static magnetic field (B₀) inhomogeneity in critical human brain regions, particularly the frontal and temporal lobes as a result of the large susceptibility difference between the brain tissue and surrounding air cavity, poses a technical challenge for B₀ field shimming and obtaining high-quality ³¹P MRS data. Another limitation is relatively low detection sensitivity because of the large volume coil size, resulting in a low signal-to-noise ratio (SNR) or spatial resolution of CSI in the cortical regions of the human brain as compared to the use of RF surface coil(s). An alternative approach to overcome the technical limitations is to use multiple surface coils incorporated with multi-channel RF receivers and split RF transmission power from a single channel transmit or use transmitter-receiver (transceiver) arrays or volume coil ^20–23. Despite the merits of high detection sensitivity and large imaging coverage, these approaches require advanced hardware equipment such as multi-channel X-nuclear receivers or transceivers, which are not commonly available even for the MR scanners equipped with the X-nucleus MRS capability. Moreover, it is challenging to perform regional dynamic B₀ shimming for those brain regions prone to susceptibility artifact since all ³¹P signals (or FIDs) from multiple RF receiver channels need to be acquired at the same time. For instance, strong B₀ inhomogeneity in the human frontal lobe causes a magnetic field distortion due to the susceptibility difference between air cavity and brain tissue, consequently a broader resonance linewidth and poor spectral quality ²⁴. Therefore, it is a challenge to simultaneously obtain high-quality in vivo ³¹P spectra or CSIs from distinct brain regions of interest for most MR scanners that are not equipped with advanced shimming technology ^25–28.

To address these technical difficulties faced by most scanners, we herein present a simple and effective design that incorporates the transistor-transistor logic (TTL) controller and modified RF pulse sequence with an existing single-channel X-nuclear RF transceiver for rapid switching the RF transceiver connection and interleaved signal acquisition between two distanced ³¹P RF surface coils covering the human occipital lobe (OL) and frontal lobe (FL), respectively. This design enables simultaneous assessment of the high-energy phosphorous metabolism from two human brain regions of interest at 7 Tesla (T) and supports the dynamic B₀ shimming utility for further improving the spectral quality of ³¹P MRS or CSI acquired in the human FL: a difficult brain region to reach an optimal shimming. This method was validated using a phantom and in human brain.

2 | METHODS

All phantom and human experiments were carried out on a Siemens whole-body/90-cm bore 7T human scanner (MAGNETOM, Erlangen, Germany) equipped with single-channel broadband X-nuclear transceiver. In vivo brain ³¹P MRS-MT experiments were performed on three healthy male volunteers (mean ± SD = 20 ± 1 years old), and the human study protocol was approved by the Institutional Review Board of the University of Minnesota. Informed consent was obtained from the participants prior to the MR measurement.

2.1 | Dual-coil ³¹P MRS system

Figure 1 illustrates the configuration of the dual-coil ³¹P MRS system consisting of two single-loop ³¹P surface coils for acquiring ³¹P MRS or chemical shift imaging (CSI) from the human OL (Channel 1) and FL (Channel 2), respectively; two passively decoupled butterfly-shape ¹H coils for anatomic MRI and B₀ shimming; a X-nuclear transmit-receiver (T/R) switch connecting with a single-channel RF power transmitter (Tx) and a single-channel RF receiver (Rx); and a TTL controller. The size of the single-loop ³¹P surface coils (~ 5 cm diameter) was desired to ensure that the detected ³¹P signal was dominated from the targeted brain tissue in the OL or FL with an optimized RF pulse flip angle. A home-built TTL controller comprised lumped elements controlled by a digital logic circuitry and a PIN diode. A TTL pulsed signal controlled by the MR console and RF pulse sequence was sent to the TTL controller for toggling the T/R switch (thus, the Tx/Rx connections) between the two ³¹P surface coils for interleaved RF transmission and acquisition of two ³¹P FIDs from the OL and FL, respectively, within the same repetition time (TR). After placing the ³¹P coils at the target brain regions, each coil resonator was tuned and matched separately with minimal mutual interference due to the small coil size and the large distance between the two coils.

Figure 1.

Schematic illustration of the dual-coil ³¹P MRS system for simultaneous measurement or imaging of ³¹P-containing metabolites from two brain regions of interest. (A) The system consists of a ³¹P T/R switch with single-channel RF transmitter and receiver, two butterfly-shape ¹H RF coils, two single-loop ³¹P RF surface coils and TTL controller between the T/R switch and the ³¹P coils. With the existing single-channel RF transmitter-receiver configuration for X-nuclear application, the TTL switcher toggles the RF transmit power and signal reception between the two ³¹P coil channels: Channel (or Coil) 1 for detecting ³¹P signals from the OL, and Channel (or Coil) 2 for detecting ³¹P signals from the FL in human brain. (B) The size and geometry of each coil were determined based on the signal coverage and location of the targeted brain regions. D is the diameter of single-loop ³¹P surface coils, whereas H, W, and L represent the size of the butterfly-shape ¹H coils.

The scheme of RF pulse sequence is designed for the interleaved acquisition of in vivo ³¹P MR spectra from two brain regions with the option for dynamic B₀ shimming (Fig. 2). The sequence enables to collect regional ³¹P FIDs from the brain regions defined by the RF magnetic field (B₁) profile of the ³¹P surface coil without additional spatial localization, or to acquire 3D C SI data by adding 3D phase encoding gradients between the RF excitation pulse and analog-to-digital converter (ADC) acquisition. Two short TTL trigger pulses were inserted at the beginning of the first and second half of the TR period, respectively, for toggling the ³¹P spin excitation and FID detection between the two ³¹P coils: one (Channel 1) measured the FID from the OL and another (Channel 2) from the FL within the same TR. The acquired ³¹P FIDs were alternatively saved in the same ADC and separated through the post data processing. The B₁ profiles of the two ³¹P surface coils were not electronically coupled (the isolation was < −27 dB between the two coils), and the RF pulse delivered by one ³¹P coil only affected the targeted brain region. Therefore, there was no signal saturation effect from the RF excitation pulse applied to another ³¹P coil during the magnetization recovery, and the true repetition time for each channel’s signal acquisition was the same as TR, thus, to maximize SNR.

Figure 2.

Schematic illustration of the dual-coil ³¹P-MT pulse sequence with 3D-CSI capability. This sequence is designed for acquiring either non-localized MRS FIDs or 3D chemical shift imaging data in the absence or presence of 3D phase encoding gradients, respectively. ³¹P spectra from two channels are acquired alternatively by switching each RF coil channel at the time of TR/2 by sending the TTL pulse from the MR scanner. For further improving B₀ field homogeneity and ³¹P spectral quality, the adjusted linear gradients are applied to the 2^nd TR to perform dynamic B₀ shimming and optimize the ³¹P spectral quality for the FL. The BISTRO pulse train can be turned on prior to the RF excitation pulse for selectively saturating the γ-ATP resonance and performing the ³¹P-MT MRS study.

The isolation between the two ¹H coils was −25 dB. The ¹H coils were driven as transceivers with adjustable RF amplitude and phase to optimize ¹H B₁ field and MRI signal intensities. Prior to ³¹P MRS experiments using both ¹H coils, global B₀ shimming covering a rectangular box (with an oblique) including both the FL and OL regions was first carried out with a dual echo in the steady-state MRI pulse sequence incorporated with high-order shimming (up to 3^rd order shim). Subsequently, the linear shimming in x, y and z directions was manually adjusted and optimized only for the FL region using the frontal ¹H RF coil to further improve spectral quality. The first-order shimming values were converted to the linear gradient strengths for before (G_x, G_y, G_z) and after (G_f,x, G_f,y, and G_f,z) the manual shimming, and the offset of linear shimming gradients (δ G_x= G_x−G_f,x δ G_{y =}= G_y − G_f,y, and δ G_z = = G_z − G_f,z) were turned on throughout the duration of the 2^nd half of TR for dynamic shimming and acquiring optimized ³¹P FID from the FL (see Fig. 2). The carrier frequencies of RF pulse and receiver were adjusted and optimized for the two ³¹P coil channels, independently.

2.2 | ³¹P-MT MRS measurement

A single-pulse-acquire sequence was carried out for studying cerebral high-energy phosphorous metabolisms in the human FL and OL. The ³¹P-MT MRS technique was implemented using the B₁ insensitive selective train to obliterate signal scheme (BISTRO) ²⁹ to selectively saturate the γ-ATP resonance (−2.5 ppm from the phosphocreatine (PCr) resonance); The crusher gradients (isosceles trapezoidal shape, ramp up/down time = 1 ms, top base time = 4 ms, amplitude = 40 mT/m, and a random order of either positive or negative gradient sign) were inserted between the RF saturation pulses (hyperbolic secant RF pulse waveform, pulse length = 51.2 ms, saturation pulse bandwidth = 160 Hz or 1.3 ppm for ³¹P MRS at 7T). Each BISTRO pulse train consisted 8 RF pulses with varied voltage amplitude and a scaling factor of 0.02, 0.04, 0.07, 0.14, 0.27, 0.49, 0.82, and 1, respectively; and the same BISTRO pulse train was repeated three times. The RF pulse power of the BISTRO trains was adjusted for optimal saturation of γ-ATP resonance. Total magnetization saturation duration was 1.37 s, and the pre-saturation delay time was 1.63 s for each RF coil channel, which could lengthen the TR time owing to the manner of interleaved acquisition. The control spectrum was also acquired in the absence of γ-ATP saturation.

Using a series of RF pulse input voltages, the reference RF power for nominal 90-degree excitation pulse flip angle (FA) for each ³¹P coil was determined by the voltage which reaches the maximum signal intensity of PCr measured under full relaxation condition (TR = 20 s, FA = 90°, single average, and 500 μs hard pulse) and then reset to achieve an Ernst flip angle for γ-ATP for the desired TR. The ³¹P–MT MRS data were acquired alternatively from the OL and FL using the acquisition parameters (TR = 3 s, spectral receiver bandwidth (BW) = 5 kHz, excitation (hard) pulse length = 300 μs, Ernst flip angle = 84.6°, number of complex FID points = 800, and number of signal average = 320).

All ³¹P-MT spectra were processed using non-linear least squares fitting algorithm of the AMARES ³⁰ using jMRUI 5.2 software program ^30,31 with the prior knowledge and fitting constraints including Lorentzian fitting, linewidth (Hz) and chemical shift ranges. Post-processing of the ³¹P MR spectra included zero-filling of FIDs to 2048 data points. A spectral linewidth reported herein was measured in the phase mode after spectral phase correction, and determined by the full-width at half-maximum (FWHM) of the resonance peak without the use of line broadening.

The concentrations of ATP, PCr, and inorganic phosphate (Pi) metabolites in the human brain were determined from the control ³¹P spectra (in the absence of γ-ATP saturation) after correcting a partial saturation factor due to the use of a short TR. Absolute concentrations of PCr and Pi were then calibrated using the cerebral ATP concentration of 2.8 mM as an internal reference ⁵. The forward reaction rate constants (k_f) under partially relaxed acquisition condition were calculated according to the following equation:

$${\mathrm{M}}_{\mathrm{c}}/{\mathrm{M}}_{\mathrm{s}}\approx 1+k_{f·}·{{\mathrm{T}}_1}^{\text{nom}}$$ [1]

where M_c and M_s are control and γ-ATP saturated magnetization signal; T₁^nom is the nominal longitudinal relaxation time (T₁) that depends on the intrinsic T₁ values of ATP, PCr and Pi at a given magnetic field strength, TR and FA parameters ^7,32, and it was simulated and determined in this study. The forward reaction fluxes can be calculated by F_f,ATPase=k_f,ATPase [Pi]_intra for the ATP_ase reaction and F_f,CK=k_f,CK [PCr] for the CK reaction ^6,32, where [Pi]_intra is the intracellular Pi concentration.

2.3 | pH Measurement

The intracellular and extracellular pH values were determined from the chemical shift difference of Pi (δ_Pi) relative to PCr based on the following equation using intracellular Pi (P_iⁱⁿ) and extracellular Pi (P_i^ex) resonance signals, respectively ^33,34:

$$\text{pH}=6.77+{log}_{10}[({\mathrm{\delta }}_{\text{Pi}}-3.29)/(5.68-{\mathrm{\delta }}_{\text{Pi}})].$$ [2]

2.4 | ³¹P CSI measurement

With the application of 3D phase encoding gradients as shown in Fig. 2, 3D ³¹P CSI was performed to explore the sensitivity for assessing the brain high-energy phosphate metabolites and to access the signal profile and coverage of the single-loop ³¹P coils in the FL or OL. The ³¹P MRS imaging was realized using the 3D Fourier series window (FSW) CSI technique; in which weighted k-space filtering was applied according to the Fourier coefficients for a predetermined voxel shape following an optimum determination of the Fourier series ³⁵. All ³¹P FSW-CSI data for both RF coil channels were acquired using the same acquisition parameters: TR = 1.2 s, spectral BW = 5 kHz, excitation RF (hard) pulse length = 300 μs, Ernst flip angle (γ-ATP) = 67.1°, field of view = 120 × 120 × 90 mm³, 3D phase encoding step = 7 × 7 × 5, total k-space scan number = 112 per CSI volume, number of complex FID points = 800, number of signal average = 8. The nominal CSI voxel (3D cylindrical shape, and circular shape on the CSI plane) size was 5.3 ml, and the total CSI acquisition time was 18 minutes. Post-processing of the ³¹P MRS or FSW-CSI voxel spectral data included zero-filling of FIDs to 2048 data points with a line broadening of 10 Hz for enhancing SNR.

Figure 3.

Comparison of in vivo ³¹P spectra (TR = 3 s, 320 signal averages, 16 minutes of acquisition time per spectrum) in the healthy human brain. In vivo ³¹P spectra were acquired in the OL (A, the PCr resonance outlined by the black dotted line) and in the FL with (C, blue dotted line) and without adjustment of the 1^st order gradients (B, red dotted line). (D) Comparison of PCr resonance spectral quality shows a significant improvement in the FL using the dynamic B₀ shimming approach. Piⁱⁿ: intracellular inorganic phosphate; Pi^ex: extracellular inorganic phosphate; GPC: glycerophosphocreatine; GPE: glycerophosphoethanolamine; PC: phosphocholine; PE: phosphoethanolamine.

The PCr and γ-ATP resonance integral images were generated from the 3D FSW-CSI dataset using a home-built program coded in Matlab v14.0 (MathWorks, Natick, MA). After phase correction of the ³¹P FSW-CSI voxel spectra in the frequency domain, the integrals of PCr and γ-ATP resonance peaks were estimated and employed to reconstruct high-resolution images and spatial profiles using the established interpolation algorithm as well as the ratio maps between the PCr and γ-ATP resonance integrals.

3 | RESULTS

Figure 3 shows the comparison results of in vivo ³¹P spectral data from the human OL and FL acquired with and without the use of first-order dynamic B₀ shimming at 7T. In the absence of the dynamic B₀ shimming, the ³¹P MR spectrum of the FL showed a broader linewidth and asymmetric shape of PCr resonance (Fig. 3B and 3D), resulting in a relatively low resonance intensity, thus, low SNR (=100). In contrast, the application of the calibrated offset linear gradients (ΔG ≤ ±10 μT/m) for optimal B₀ shimming of the FL region significantly improved the line shape and narrowed the linewidth (Fig. 3C, spectral linewidth or FWHM of PCr = 11.3 Hz or 0.09 ppm at 7T; SNR=130), reaching comparable quality of the ³¹P spectra as in the OL (Fig. 3A, FWHM of PCr = 10.1 Hz, SNR = 140) for this representative subject.

We also compared the spectral linewidths of ³¹P spectra acquired from the FL and OL regions from the same subject as shown in Fig. 3 under two acquisition conditions: dual-coil configuration with dynamic shimming (11.3 Hz in FL and 10.1 Hz in OL) versus regional shimming within the lobe of interest using a traditional single ³¹P coil approach (10.3 Hz in FL and 9.0 Hz in OL). Therefore, the difference was small (about 1 Hz) between the two-coil (global plus dynamic shimming) and single-coil (optimal shimming focused on a single lobe of interest) configuration. The proposed method provides an effective approach for interleaved acquisition of ³¹P spectra from two brain regions with high spectral resolution and quality.

Figure 4 displays typical in vivo ³¹P-MT spectra in the absence (control) and the presence of γ-ATP saturation (i.e., MT) preparation (pointed by red arrows) acquired from the human OL (Fig. 4A and 4C) and FL (Fig. 4B and 4D) from two representative subjects. Overall, in vivo ³¹P control spectra showed excellent quality and high sensitivity allowing for the reliable detection and quantification of the PCr, ATP, Pi and other phosphorus metabolite signals in healthy human brains. The ³¹P-MT spectra with γ-ATP saturation indicated sufficient saturation efficiency at given MT preparation duration with optimal RF saturation powers for each ³¹P RF coil channels.

Figure 4.

Representative in vivo ³¹P-MT spectra simultaneously detected from the OL and FL from healthy human subjects. Top panels (A and C) show the respective ³¹P control spectra and ³¹P-MT spectra (with γ-ATP saturation pointed by the red arrows) acquired from the OL, while the bottom panels (B and D) show the control and ³¹P-MT spectra from the FL. Overall, the ³¹P MR spectra from the two brain regions show excellent spectral quality and adequate MT saturation efficiency in two representative subjects (TR = 3 s, 320 signal averages, 16 minutes of acquisition time per spectrum).

The PCr concentrations measured in the OL and FL were 4.12 ± 0.11 mM and 4.08 ± 0.21 mM, respectively. The intracellular P_iⁱⁿ concentrations for the OL and FL were 1.05 ± 0.04 mM and 0.94 ± 0.05 mM and corresponding extracellular Pi^ex concentrations were 0.34 ± 0.01 mM and 0.40 ± 0.03 mM, respectively, after the correction of the T₁ saturation effect for PCr (apparent T₁ = 3.37 s) and Pi (apparent T₁ = 3.19 s) ³⁶.

Under partially relaxed acquisition condition at given nominal excitation RF pulse flip angle and TR, estimated T₁^nom values for PCr and P_iⁱⁿ were 1.91 s and 1.55 s, which were determined by the slopes of the solid lines in Fig. 5. The forward rate constants of k_f,CK (for PCr→ATP reaction) were 0.389 ± 0.002 s⁻¹ and 0.32 ± 0.01 s⁻¹ for the OL and the FL, respectively, which are in good agreement with previously reported values ³². The forward rate constants of k_f,ATP (for Pi→ATP reaction) were 0.14 ± 0.06 s⁻¹ for the OL, and 0.18 ± 0.02 s⁻¹ for the FL. Accordingly, the forward CK reaction fluxes were 87.5 ± 1.7 μmol/g/min and 71.9 ± 1.6 μmol/g/min, and the forward ATP_ase reaction fluxes were 8.1 ± 3.7 μmol/g/min and 9.4 ± 1.3 μmol/g/min for the OL and FL, respectively, after the unit conversion using the brain tissue density of 1.1 g/ml ⁶.

Figure 5.

Simulated results of M_c/M_s versus forward reaction rate constant (k_f) plot for CK (A) and ATP_ase (B) reactions of the human brain at 7T under partial relaxation condition (TR = 3 s, FA = 84.6^o, pre-saturation time delay (d1) = 1.63 s). Intrinsic T₁s for PCr, γ-ATP, and Pi are 4.86, 1.35, and 3.77 s ³². Estimated nominal T₁ values (T₁ ^norm, the slope for each solid line plot) for PCr and Pi were 1.91 and 1.55 s, respectively. The simulated T₁ ^norm values were employed to calculate the k_f values for CK and ATP_ase reactions according to Equation [1].

The measured intracellular pH was 7.03 ± 0.01 identically for the OL and FL, and the pH result is consistent with the value of 7.03 in the human gray matter dominated brain regions ¹⁹. The extracellular pH was 7.45 ± 0.04 and 7.47 ± 0.01 for the OL and FL, respectively.

Figures 6 displays in vivo 3D ³¹P CSI results from three representative image slices (Figs. 6B–6D; in coronal orientation) detected in the OL from one subject and corresponding maps of PCr (Figs. 6F–6H) and γ-ATP resonance integrals (Figs. 6I–6K), respectively. The large integral variation reflects the inhomogeneous B₁ and signal intensity profiles of the ³¹P surface coil. In contrast, the ratio map between the PCr and γ-ATP integrals shows relatively uniform distribution in the central brain region covered by the surface coil even without the correction of relaxation saturation effect because of the lack of information regarding the spatial distribution of RF pulse flip angle in this study. Figure 7 shows the similar results detected in the FL from the same subject. Given the proximity of the ³¹P RF coils, each ³¹P CSI showed a sufficient signal coverage and excellent detection sensitivity for the desired cortical lobes without signal overlapping between the two ³¹P coils. In conjunction with the linear dynamic B₀ shimming, in vivo ³¹P spectra taken from representative CSI voxels indicate comparable spectral quality with narrow resonance linewidth and excellent SNR for both the OL (Fig. 6E, SNR=60) and FL (Fig. 7E, SNR=45) regions.

Figure 6.

Representative in vivo ³¹P 3D CSI data, PCr and γ-ATP maps, and their ratio map from the human occipital lobe overlaid on the coronal anatomical images. Panel A displays the location of a single-loop ³¹P coil (Channel 1, blue bar) placed near the OL, the field of view, and slice positions of the 3D ³¹P-CSI; (B–D) displays the ³¹P MRS profiles from three selected CSI slices; (E) A single-voxel ³¹P spectrum denoted in yellow circle in (C) shows excellent detection sensitivity at the given spatial- and temporal-resolution; (F–H) shows the corresponding normalized PCr maps (by setting the maximum peak integral of PCr to 1); (I–K) shows the corresponding normalized γ-ATP maps; (L) shows the ratio map between the PCr and γ-ATP in the region of interest (ROI) as indicated by the white box in (C). ³¹P CSI data acquired with a single-loop coil clearly demonstrate a sufficient signal coverage (reaching ~ 5.4 cm depth).

Figure 7.

Representative in vivo ³¹P 3D CSI data, PCr and γ-ATP maps, and their ratio map from the human frontal lobe overlaid on the coronal anatomical images. Panel A displays the location of a single-loop ³¹P coil (Channel 2, blue bar) placed near the FL, the field of view, and slice positions of the 3D ³¹P-CSI; (B–D) displays the ³¹P MRS profiles from three selected CSI slices; (E) With dynamic B₀ shimming, the single-voxel ³¹P MR spectrum (the yellow circle in (C)) shows excellent spectral quality and detection sensitivity; (F–H) shows the corresponding normalized PCr maps; (I–K) shows the corresponding normalized γ-ATP maps; (L) shows the ratio map between the PCr and γ-ATP in the ROI as indicated by the white box in (C). 3D ³¹P CSI data demonstrate the brain regions with the highest signal detected by the ³¹P coil was approximately 2.5 cm from the surface of the frontal lobe (Slice 2), which covers critical brain regions including medial frontal gyrus and cingulate gyrus.

Although two ³¹P coils as employed in this study had the same coil design and size, the coil planes to conform the human head shape were different between the frontal and occipital lobes. The occipital-lobe ³¹P coil plane was approximately parallel to the B₀ direction (or z axis, see Fig. 6) and it produced an optimal distribution of B₁ field (i.e., the transverse magnetic field component B_xy) for covering the occipital lobe. In contrast, the frontal-lobe ³¹P coil plane was slightly tilted from the z axis for most subjects in order to conform the front head shape (see Fig. 7), resulting in a reduction in B_xy (accompanying by an increase in B_z), thus, reducing the detection sensitivity and ³¹P signal in the frontal lobe. The degree of signal reduction varied between inter subjects depending on the front head shape and RF coil setup.

4 | DISCUSSION

We demonstrate a novel design of dual-coil ³¹P MRS system for interleaved measurement and imaging of the regional high-energy phosphorous metabolism in two human brain regions of interest. Using an existing single channel X-nuclear RF transmit amplifier and receiver, ³¹P MRS data were successfully collected from the frontal and occipital lobes within the same TR, providing a time-efficient way to reduce total acquisition time to half compared to the single acquisition approach for two brain regions, thus, for improving SNR by the square root of 2. Notably, with further adjustment of the linear gradients for dynamic shimming, the in vivo ³¹P spectrum of the frontal brain region showed high-quality with better spectral line-shape compared to that without adjustment. This merit enables reliable assessment and quantification of the cerebral phosphorus metabolites in the frontal lobe which plays an essential role in high cognitive brain function. In addition, the 3D ³¹P CSI results demonstrate sufficient signal coverage and good B₀ homogeneity achieved with the dual-coil ³¹P MRS system, resulting in excellent spectral quality and phosphate metabolite images in distinct two brain regions.

As aforementioned, the putative presence of complex neurological symptoms commonly linked to neurodegenerative diseases have prompted the need for re-evaluating classic diagnostic/treatment approaches to the neurodegenerative disease. For instance, there has been a high association between clinical manifestations of motor restlessness in the upper motor system and cognitive impairments in the frontal areas in amyotrophic lateral sclerosis ^12,13. In addition, the central viewpoint of upper motor dysfunction in Parkinson’s disease has been revised due to frequent reports of non-motor symptoms encompassing cognitive dysfunction in the frontal lobe ^14,16 and visual dysfunction in the occipital lobe ^37,38. In this regard, the proposed method in this work capable of simultaneous assessment of the bioenergetic abnormalities in the target brain regions of interest could provide insights into the pathophysiology underlying neurodegenerative diseases.

Detection sensitivity or SNR is essential for in vivo MRS measurement as it significantly influences the accuracy and precision for assessing and quantifying cerebral metabolites not only for X-nuclei but also ¹H MRS ^39–42. Moreover, low concentrations (a few millimolar) of the phosphate metabolites in the human brains become a daunting challenge for achieving high spatial or temporal resolution of ³¹P MRS imaging. Given the known lower sensitivity of RF volume coil, in vivo ³¹P MRS with application of two single-loop ³¹P surface coils in the present work shows higher detection sensitivity and excellent SNR in the desired cortical regions of interest in the human brain at 7T; thus, it could provide a valuable solution for addressing scientific questions demanding high SNR or imaging spatial resolution in the cortical regions of interest. The lack of physical proximity between the two RF surface coils inherently minimized unwanted interactions between the brain regions and allowed for discrete acquisition of in vivo ³¹P MRS data with comparable spectral quality within the same sampling time. Accordingly, the combination of the dual ³¹P surface coils with CSI method as demonstrated in this work bears considerable potential for further investigation of the correlation of high-energy phosphorous metabolism between different human brain regions or for functional imaging of neuroenergetic changes between rest and activated brain states with rich matrices of physiological measures including the high-energy phosphorous metabolites of ATP and PCr, F_f,ATPase and F_f,CK, intra- and extra-cellular pH, intracellular free [Mg⁺²], metabolic precursors for phosphorous lipid metabolisms.

To homogenizing the B₀ field is an essential prerequisite to enhance spectral quality and resolution for all in vivo MRS applications. For this reason, several advanced B₀ shimming techniques have been proposed ^25–28. Despite the substantial improvement of B₀ homogeneity, the availability of those methods is still limited to most conventional MR scanners. In addition, the current-state-of-the-art technology employing either multiple-channel receivers or transceivers acquires the FIDs from the multiple RF receiver coils at the same time, thus, is difficult to perform alternative and dynamic B₀ shimming for individual RF coils at a different time within a single TR period. Most of conventional MR systems including the scanner employed in the present study do not allow real-time updating of high-order shim currents owing to the eddy currents induced by fast switching shim currents and slow decays, leading to the degradation of spectral quality. The present study was limited to the first-order dynamic shimming using the linear gradients. Nevertheless, the concept of the alternative B₀ shimming among multiple RF coils within the same repetition time is not limited to the first-order shim, and it can be incorporated with other advanced dynamic shimming techniques enabling first- and higher-order dynamic shimming.

Early attempts at dynamic shimming showed the improved image quality by updating the linear shim terms dynamically in multi-slice imaging acquisition ^43,44 as well as in single-voxel ¹H MRS ⁴⁵. We implemented this approach by adjusting linear gradients for first-order shimming of the FL region, providing the perceptible improved quality of in vivo ³¹P spectra in the FL comparable to that of the OL that is less problematic for achieving satisfactory B₀ shimming. On the other hand, two different shimming offsets could be readily incorporated into the pulse sequence (Fig. 2) and employed alternatively to the two ³¹P coils for further improving the spectral resolution in both brain regions.

In principle, in vivo MR spectra from multiple voxels could be acquired using a large (or volume) RF coil incorporated with one of single-voxel localization MRS methods (e.g., ISIS: Image Selected In Vivo Spectroscopy ⁴⁶; PRESS: Point Resolved Spectroscopy ⁴⁷; or STEM: Stimulated Echo Acquisition Mode ⁴⁸) and the option of dynamic shimming if the selected voxels are not spatially overlapped in a Cartesian coordinate system. Therefore, this approach is limited only to the oblique arrangement of multiple voxels and not suitable for the case if one is interested (for instance as investigated in the present study) in the human frontal and occipital lobes, which are aligned on the same sagittal plane. Although the voxel localization using multiple frequency-selective RF pulses could be not spatially overlapped between multiple voxels, the RF power delivered from all RF pulses employed to acquire multiple voxels will be absorbed by all brain regions covered by the RF coil. The average specific absorption rate (SAR) of the brain tissue approximately equals to the multiply of voxel numbers and the SAR required to collect a single voxel MRS. This could pose a safety concern especially at ultrahigh fields.

5 | CONCLUSION

We demonstrate that the new dual-coil ³¹P MRS approach as described in this work could offer a cost-effective and time-efficient solution for interleaved in vivo ³¹P MRS measurements and imaging from multiple desired brain regions based on an existing single-channel transmitter-receiver configuration with optimal SNR and spectral resolution. It could provide a valuable metabolic imaging tool for examining the cerebral energy metabolisms in healthy subjects and impairments in patients in desired and multiple brain regions. Moreover, the same approach and design can be: 1) extended to more than two RF coil channels, thus, covering more brain regions; 2) adapted to other X-nuclei or ¹H MRS; and 3) potentially used for functional MRI studies from multiple brain regions of interest with higher sensitivity and spatiotemporal resolution.

ABBREVIATIONS

3D

three-dimensional

ADC

analog-to-digital converter

ATP

adenosine triphosphate

ATP_ase

ATP synthase

B₀

static magnetic field

CK

creatine kinase

CSI

chemical shift imaging

FA

flip angle

FID

free induction decay

FL

frontal lobe

FSW

Fourier series window

MR

magnetic resonance

MRS

MR spectroscopy

MT

magnetization transfer

NAD

nicotinamide adenine dinucleotide

OL

occipital lobe

PCr

phosphocreatine

Piⁱⁿ

intracellular inorganic phosphate

Pi^ex

extracellular inorganic phosphate

GPC

glycerophosphocreatine

GPE

glycerophosphoethanolamine

PC

phosphocholine

PE

phosphoethanolamine

RF

radiofrequency

SNR

signal-to-noise ratio

T

Tesla

TR

repetition time

TTL

transistor-transistor logic