Efficient spectroscopic imaging by an optimized encoding of pretargeted resonances

Abstract

Purpose: A "relaxation-enhanced" (RE) approach to acquire in vivo localized spectra with flat baselines and good sensitivity has been recently proposed. As RE MR spectroscopy (MRS) targets a subset of a priori known resonances, new possibilities arise to acquire spectroscopic imaging data in faster, more efficient manners. This is hereby illustrated by Spectroscopically Encoded Chemical Shift Imaging (SECSI).

Methods: SECSI delivers spectral/spatial correlations by collecting gradient echo trains whose timings are defined by the shifts of the resonances to be disentangled. Condition number considerations allow one to unravel these image contributions for various sites by a simple matrix inversion. The efficiency of the ensuing method is high enough to enable a sampling of additional spatial axes by means of their phase encoding in spin-echo trains.

Results: The one-dimensional (1D) spectral / 2D spatial SECSI acquisitions were implemented on phantom, ex vivo, and in vivo models. In all cases, quality site-resolved images were obtained. The experimentally observed enhancements were consistent with theoretical signal-to-noise ratio derivations.

Conclusion: While still bound by MRSI's sensitivity limitations, a novel spectroscopic imaging protocol exploiting a priori information, selective excitations and multiple echo encodings, was proposed and demonstrated. The method is promising when dealing with high T₂ / T2* ratios, sparse data, or hyperpolarization studies. Magn Reson Med 77:511-519, 2017. © 2016 International Society for Magnetic Resonance in Medicine.

Keywords: MRSI; brain metabolic imaging; selective spectral excitations.

Figure 1

Principles of RECESS MRSI. (a) Basic RE MRS sequence [10,13] incorporating a 90˚ SLR multiband selective pulse, a series of 180˚ pulses providing refocusing and 3D LASER localization, and a FID acquisition. (b) In a chemical shift imaging (CSI) extension, this localized spectroscopy is endowed with 2D spatial resolution by the addition of two phase-encoding gradients. (c) In RECESS MRSI the spectroscopic information is no longer retrieved by fast Fourier Transform (FT) of a FID, but via N refocused readout gradients whose echoes are timed at intervals τ that enable –following a FT of the echoes along their k-evolution axes– a stable matrix-based inversion delivering each of the metabolites’ images. A τ₁ delay is introduced to refocus the gradient-imposed evolution without disturbing these shift-encoding steps. (d) Given the short duration of the RECESS chemical-shift spectral decoding in (c), this scheme can be extended by a CPMG-based echo train that carries out a phase encoding of the remaining spatial dimensions (N_pe loops) in a single scan. In (b), (c) and (d), the spatial encoding of the MRS information is assumed along only two axes, leaving a need for a 1D LASER block to localize the remaining dimension.

Figure 2

Transverse of the (k_ro,k_pe,τ_n) space in the single-shot RECESS acquisition protocol, highlighting details of the chemical shift encoding and decoding modes. (a) Expanded acquisition block of the sequence, showing only the first and last acquisition gradient echoes for each phase encoding step. (b) Trajectories imposed on the spins’ evolution along the chemical shift encoding, the k_ro and the k_pe imaging dimensions, highlighting the positions of ten instants marked in the sequence (a). Green arrows represent k_ro-axis traces scanned during six hypothetical readout gradients for each chemical-shift-encoding train; the dotted green lines next to them represent the trajectories imposed by the corresponding refocusing gradients. The solid cyan lines show the jumps along k_pe imposed by the phase encoding gradient blips and by the refocusing pulses; notice in this case the alternating changes in k_pe values. In each phase encoding step the signals collected in the presence of positive readout gradients (k_pe ≤ 0; darker green colors) are modulated by {E_mn}, while those recorded during negative readout gradients (k_pe > 0; lighter green) are modulated by {E_mn*}. (c) Details of the chemical shift decoding procedure. The N×N_pe full k_ro-space echoes carrying the chemical shift encoded information, are separated into two matrices according to the sign of the readout gradient employed. For each phase-encoded value, the N echoes recorded in the presence of positive readout gradients are decoded by the inverse of the encoding matrix {E_mn}, while shift-encoded signals recorded while in the presence of negative readout gradients are decoded by the inverse of {E_mn*}. Each of the ensuing 2D k-space data sets is then fast FT for retrieving what in principle are identical spectrally-resolved images (procedure not shown); these can be co-added for the sake of improving sensitivity.

Figure 3

Optimizing the condition number of a chemical shift encoding matrix E_mn = {e^i(Ω_mnτ)}, assuming τ_n = nτ, n = 1,…N. (a) Phantom sample scenario involving the excitation of three chemical sites at 4.03, 2.99 and 1.86 ppm, showing how condition numbers change as a function of τ and N. Although equally good encodings appear for a variety of combinations, an encoding based on N = 6 gradient echoes equispaced by τ = 1.002 ms provides both low condition number (1.08, red) and timings that can be accommodated by the gradient echo train without being overtly short or long. (b) Idem but for a hypothetical mouse brain sample, involving the excitation of residual water plus Cho, Cre and NAA methyl peaks at 4.80, 3.09, 2.93 and 1.92 ppm. A suitable condition number = 1.84 (red) is then reached for N = 8 and τ = 1.52 ms; although a better conditioning is achieved by N = 10, the increased number of gradient echoes coupled to the longer τ = 1.62 ms intervals would lead to unnecessary T₂-derived losses. (c) Examining the robustness of E’s conditioning, with ten cases where the frequencies of the four resonances targeted in (b) were varied randomly over 20 Hz, for N fixed at 8 gradient echoes.

Figure 4

Comparing the spectroscopic imaging results from a phantom sample involving water, methanol (I), acetone (II) and cyclohexane (III), arising from various versions of relaxation-enhanced sequences. (a) Reference spin-echo multi-shot image. (b) Reference spectrum acquired by PRESS [30], delivering the spectral information required to design the SLR pulse (the small unlabeled resonance at ~5.5 ppm corresponding to the untargeted –OH methanol site). (c,d) Spectra acquired using the spatially-localized RE MRS sequence in Fig. 1a, in well shimmed and in poorly shimmed magnetic fields, respectively. A 40 ms multiband linear-phase equiripple pulse was used to excite the (I, II and III) resonances; the line widths of resonance III were 12 Hz in (c) and 75 Hz in (d). (e) Spectrally resolved images of the three components arising from a Relaxation-Enhanced but otherwise conventionally phase-encoded CSI sequence like the one introduced in Fig. 1b, collected in the well-shimmed field. (f,g) Spectrally resolved images of the three targeted components arising from the multi-scan shift-encoded RECESS sequence introduced in Fig. 1c, collected in a well-shimmed and in an inhomogeneous field, respectively. (h,i) Idem but for the single-scan multi-echo RECESS MRSI sequence shown in Fig. 1d. Regions of interests for SNR and for residual cross-talk level calculations were marked with squares in (a). The SNR for each image was measured by dividing the average of the corresponding component signals within the marked green/magenta/blue squares by the standard deviation of the noise in the red-marked square. The same markers were also used to calculate residual levels in panels (e)-(i), by dividing the targeted signal by the residuals arising at the positions of the remaining two components. Common scan parameters: FOV = 40 × 40 mm², slice thickness = 2 mm, two dummy scans, echo time = 50 ms, TR = 5 s, averages nt = 1. For the RECESS experiments, the shift-encoding parameters were τ = 1.002 ms, N = 6. The matrix size for (e) was 64 × 64 and its total scan time is 5 h 41 min. Matrix sizes for (f), (g), (h) and (i) were 128 × 128; total scan times for (f) and (g) were 10 min 40 sec, while scan times for (h) and (i) were only 20 sec.

Figure 5

Comparison of ex-vivo mouse brain results arising from RECESS and from a RE-based CSI sequences. (a) Reference SEMS image. (b) Reference PRESS spectrum acquired on a 16 × 16 × 4 mm³ voxel using 32 averages. (c) RE MRS spectrum acquired on the same voxel and scan numbers as (b), but using the sequence in Fig. 1a with a 40 ms two-band pulse exciting the total Cholines (tCho) and total Creatines (tCre) in one band, and N-Aceytl Aspartate (NAA) in another. (d-f) Cho, Cr, NAA maps (overlaid on anatomical T₁-weighted images) arising from the execution of the CPMG-based RECESS sequence in Fig. 1d. (g-i) Idem but for acquisitions based on a RE CSI sequence incorporating independent phase-encoding loops. Common scan parameters: FOV = 16 × 16 mm², slice thickness = 4 mm, matrix size =32 × 32, echo time = 50 ms, TR = 2 s. RECESS parameters: τ = 1.524 ms, N = 8, number of averages = 2048. The number of averages for the RE CSI sequence was 8 per phase-encoding value. The experimental time of the RECESS MRSI experiment was 2 h 16 m (including the reference scans) while the phase-encoded CSI acquisition took 4 h 33 m.

Figure 6

In vivo fat/water separation capabilities of the RECESS sequence in Fig. 1d, as applied to abdominal mouse investigations at 7T. (a, c) Multiscan spin echo references involving fat suppression (a) and water suppression (c). (b, d) Water- (b) and fat-tissue (d) images separated by single-scan RECESS acquisition. Common imaging parameters: FOV =32 × 32 mm², TR = 2 sec, TE = 14 ms, slice thickness = 2 mm. The reference image matrix size was 128×128; due to the short T₂s the RECESS image size was 64×64, which explains the lower resolution. Total scan time for each spin-echo multi-scan image: 4 min 16 sec. Total RECESS acquisition time (including a reference navigator scan): 4 sec.