Optimized (31)P MRS in the human brain at 7 T with a dedicated RF coil setup

Abstract

The design and construction of a dedicated RF coil setup for human brain imaging ((1)H) and spectroscopy ((31)P) at ultra-high magnetic field strength (7 T) is presented. The setup is optimized for signal handling at the resonance frequencies for (1)H (297.2 MHz) and (31)P (120.3 MHz). It consists of an eight-channel (1)H transmit-receive head coil with multi-transmit capabilities, and an insertable, actively detunable (31)P birdcage (transmit-receive and transmit only), which can be combined with a seven-channel receive-only (31)P array. The setup enables anatomical imaging and (31)P studies without removal of the coil or the patient. By separating transmit and receive channels and by optimized addition of array signals with whitened singular value decomposition we can obtain a sevenfold increase in SNR of (31)P signals in the occipital lobe of the human brain compared with the birdcage alone. These signals can be further enhanced by 30 ± 9% using the nuclear Overhauser effect by B1-shimmed low-power irradiation of water protons. Together, these features enable acquisition of (31)P MRSI at high spatial resolutions (3.0 cm(3) voxel) in the occipital lobe of the human brain in clinically acceptable scan times (~15 min).

Keywords: 31P-MRS; 31P-MRSI; 7 T; B1 shimming; RF coil; WSVD; array coil; multi-transmit; spectroscopic imaging; ultra-high field.

Figure 1

Overview of the complete setup, with detailed representations of the newly designed, detunable ³¹P BC insert and ³¹P Rx array. (A) Schematic representation of the eight‐rung high‐pass BC coil; included are details of the tank and detune circuits (red box). Capacitor values used to tune and match the BC coil were C _m = 15.6 pF, C _t = 13.3 pF, C _c = 27 pF and C _p = 10 pF. (B) The octagonal‐shaped 8‐CH ¹H head coil was used as the basis for the insertable, detunable ³¹P BC coil, which could host the additional 7‐CH Rx array coil. (C) Schematic configuration of the 7‐CH Rx array. (D) A pictorial overview of a single element of the 7‐CH Rx array, where each adjacent element is decoupled by overlap and by pre‐amplifier decoupling.

Figure 2

(A–C, E–G) Simulation of proton magnetic field (A–C) and phase (E–G) distribution of the 8‐CH head coil, with and without the ³¹P BC inserted into the 8‐CH setup. (A, E) The field and phase distributions of the 8‐CH head only, (B, F) the distribution when the BC is inserted (in this case the legs had a width of 12.5 mm), (C, G) the distribution when the width of the legs is reduced to 4.5 mm. (D, H) ¹H images of the same phantom with the BC present in the setup; the image in D corresponds to the simulation result as presented in C, hence showing equal field distribution. In H the field distribution is shown with the tank circuits inserted into the BC.

Figure 3

Results of the validation measurements using the field probes. (A) Two‐dimensional visualization of the coil setup surrounding the head–shoulder phantom as it was used to determine the magnitude of the |E| and |H| fields with the field probes. Included is a map of the proton |H| field as it was measured with the 8‐CH head coil only. The phantom was filled with tissue‐simulating fluid, having a conductivity of 0.98 S/m for ¹H and 0.78 S/m for ³¹P, and a permittivity of 56.3 for ¹H and 76.5 for ³¹P. (B–E) Profiles for each individual |E|‐ and |H|‐field measurement (these profiles were taken along the dashed lines (A)) with (B, D) ¹H and (C, E) ³¹P. These measurements were made (i) for the 8‐CH head coil only, (ii) for the 8‐CH head coil with ³¹P BC inserted and (iii) for the complete setup with receive coil. These measurements were corrected for input power and hence are directly comparable to those of Reference 25. Note that |E| and |H| fields decrease when more components are present.

Figure 4

Noise correlation matrix of the 7‐CH Rx array. All channels are properly decoupled, because noise does not correlate significantly. The average noise correlation over all elements was 10 ± 7%, with its maximum of 23% found between elements 3 and 6.

Figure 5

Interpolated SNR images of a spherical phantom containing 30 mM inorganic phosphate. The images were obtained with the BC coil (top row) and with the local Rx array (middle row). The gain in SNR is more than sevenfold close to the receive array. Also note the uniform B ₁ field when data was solely acquired with the BC.

Figure 6

(A–C) Results of in vivo experiment using the complete setup showing different B ₁ ⁺ maps of the ¹H field: (A) in CP− mode, (B) homogenized for the complete brain, and (C) locally optimized for occipital lobe. Showing the ability to use B ₁ shimming with the coil setup. Background images were acquired with the 8‐CH ¹H head coil driven with global B ₁ optimization, showing fairly homogeneous field distribution. (D–F) ³¹P spins were excited and signal was acquired with the BC (D) without and (E) with NOE enhancement of PCr; (F) the global enhancement map. (G, H) An example of ³¹P spectra taken from the same voxel (G) without NOE enhancement and (H) with NOE enhancement. Spectra were obtained in 7 min 48 s with an approximate voxel size of 38 cm³. The metabolites present in these spectra are (1) phosphoethanolamine (PE), (2) phosphocholine (PC), (3) P_i, (4) glycerophosphoethanolamine (GPE), (5) glycerophosphocholine (GPC), (6) PCr, (7) γ‐ATP, (8) α‐ATP and (9) NADH.

Figure 7

High‐resolution 3D MRSI using all available options to increase SNR. The signals were enhanced using NOE and were received with the local receive array. Note the exquisite quality of the spectra. Top left, a sagittal view of the human brain overlapped with ³¹P spectra as acquired with the receive array. Top right, a transversal view of the human brain also overlapped with a spectral map covering the occipital lobe. Bottom two rows, metabolite maps of seven different ³¹P compounds. The colors show the distribution of the metabolites (scaled 0–100%) represented by the fitted integral. Note that in only two voxels were low‐SNR metabolites (P_i and GPC) fitted incorrectly within the FOV of the array coil. Moreover, note the difference in spatial distributions of the different metabolites, which arise not only from the B ₁‐receiving profile (anterior–posterior), but also from differences between gray and white matter (left–right). A spectrum with high quality (3 cm³ voxel), as acquired in 15 min in the occipital lobe (red box), is shown too. Not shown in this spectrum is the β‐ATP, as the bandwidth of the pulse to properly excite this metabolite was insufficient. (1) PE, (2) PC, (3) P_i, (4) GPE, (5) GPC, (6) PCr, (7) γ‐ATP, (8) α‐ATP and (9) NADH.