Fast 3D chemical exchange saturation transfer (CEST) imaging of the human brain

Abstract

Chemical exchange saturation transfer magnetic resonance imaging can detect low-concentration compounds with exchangeable protons through saturation transfer to water. This technique is generally slow, as it requires acquisition of saturation images at multiple frequencies. In addition, multislice imaging is complicated by saturation effects differing from slice to slice because of relaxation losses. In this study, a fast three-dimensional chemical exchange saturation transfer imaging sequence is presented that allows whole-brain coverage for a frequency-dependent saturation spectrum (z-spectrum, 26 frequencies) in less than 10 min. The approach employs a three-dimensional gradient- and spin-echo readout using a prototype 32-channel phased-array coil, combined with two-dimensional sensitivity encoding accelerations. Results from a homogenous protein-containing phantom at 3T show that the sequence produced a uniform contrast across all slices. To show translational feasibility, scans were also performed on five healthy human subjects. Results for chemical exchange saturation transfer images at 3.5 ppm downfield of the water resonance, so-called amide proton transfer images, show that lipid signals are sufficiently suppressed and artifacts caused by B(0) inhomogeneity can be removed in postprocessing. The scan time and image quality of these in vivo results show that three-dimensional chemical exchange saturation transfer MRI using gradient- and spin-echo acquisition is feasible for whole-brain chemical exchange saturation transfer studies at 3T in a clinical time frame.

FIG. 1

Sequence diagram of 3D CEST imaging with GRASE readout. The sequence consisted of 4 block saturation pulses (200 ms each, 2 μT), a frequency modulated lipid suppression pulse, and 3D GRASE image acquisition with TSE in the y (phase) direction and EPI in the z (slice) direction. The dotted lines in the Gx direction are the Maxwell gradients. TSE factor = 22, EPI factor = 7, SENSE factor = 2 × 2, TR = 2.5 s.

FIG. 2

Geometric prescription of 3D CEST sequence in the brain. a: 3D imaging box (yellow) covering the entire brain, parallel to the AC-PC line. Shimbox (green) covering nearly the entire cerebral area. b: Cerebellum area manually included in lower slices of B₀ maps for shimming. c: Scanner’s center frequency determined on water signals in shimmed regions only, excluding sinus region.

FIG. 3

Volumetric CEST imaging on a homogenous eggwhite phantom using multi-slice SE and 3D GRASE acquisition. a,b: z-spectra from 4 of 12 slices using multi-slice SE (a) and 3D GRASE (b). c,d: Corresponding MTR_asym plots for a and b, respectively. e: MTR_asym(3.5ppm) values measured by the two methods as a function of the acquired time order. 3D GRASE shows a more uniform CEST response across slices than multi-slice SE. f: Slice locations in eggwhite phantom.

FIG. 4

Determination of the frequency offset for the FM lipid suppression pulse and the effect of lipid suppression in the 3D GRASE sequence. a: MR spectra at five different frequency offsets and the integral of residual lipid signals as a function of the offset (dashed plot). The dashed plot reflects the sharp edge of the suppression pulse. At an offset of −300 Hz (pink spectrum), which was used in this study, the lipid resonance is sufficiently suppressed (pink dot in the dashed plot) while the water resonance is not affected significantly (shaded area). b,c: The saturated images at ±3.5 ppm and MTR_asym(3.5ppm) images for three middle-top slices (in three columns) without and with lipid suppression. Without lipid suppression, ring-like hypointensities (white arrow) appear in the MTR_asym(3.5ppm) images.

FIG. 5

Unsaturated (S₀) images, calculated water center-frequency offset maps, uncorrected MTR_asym maps at 3.5 ppm and corresponding corrected MTR_asym maps for six typical slices. B₀ inhomogeneity (solid black arrow) near air-tissue interfaces (sinus, ear) caused visible large artifacts (open arrow) in the MTR_asym maps at 3.5 ppm. With lipid suppression and B₀ correction, all MTR_asym(3.5ppm) maps (bottom row) appear reasonably homogeneous.

FIG. 6

Comparison of the experimental results of 3D CEST imaging in different brain regions on healthy subjects (n = 5). a: The SNRs at three ROIs (cerebellum, cerebral WM and cerebral GM, each with about 40 voxels). b: z-spectra in these three ROIs. c: Corresponding MTR_asym plots. MTR_asym(3.5ppm) is significantly higher in cerebellum than in cerebrum (GM, p = 0.002; WM, p = 0.02).