Diffusion-weighted MR spectroscopy has recently been shown to help elucidate the microstructure of brain tissue and the use of oscillating gradients in MRI has provided the option to study local features on a smaller scale. Here, we aim at combining both techniques to investigate human skeletal muscle in vivo on a clinical scanner. Oscillating and pulsed gradient schemes with a large range of diffusion times were applied with otherwise identical acquisition settings. Initial results are shown, but also the challenges faced for muscle where physiologic pulsation and gradient-related artifacts may be prominent.
Acquisition: Siemens Prisma 3T (maximum gradients of 80 mT/m); multichannel transmit/receive extremity coil; in vitro tests using "braino" phantom (GE medical), in vivo data from 4 human subjects with ROI in tibialis anterior muscle (ROI size of 11-14 [cm3]) with leg parallel to B0. DW spectra acquired with a semiLaser localization sequence extended by optional diffusion weighting elements either in the form of pulsed trapezoidal gradients or oscillating gradient11 shapes placed on both sides of the last adiabatic refocusing pulse as illustrated in Fig. 1; sequence details and brain applications presented elsewhere10; in vivo spectra recorded with a repetition time of 2.5 s and a long echo time (TE) of 170 ms to accommodate enough diffusion weighting ($$$b_{max}$$$ of $$$\sim$$$ 800 s/mm2 at TD 7.5 ms) for the case of oscillating gradients. For pulsed gradients, the same TE was used with modified TD (39-139 ms). At long TE, no water suppression is needed. Diffusion-weighting gradients were applied approximately along (z) and perpendicular (x,y) to the muscle fibers.
Data processing and fitting: Spectra processed in jmrui12 and matlab. Simultaneous fitting of the whole dataset from each volunteer performed in FiTAID13 making use of lineshape information taken from the isolated water peak in each spectrum, where the water peak had been parameterized as a superposition of 25 arbitrary Voigt lines spaced by 2-5 Hz. Spectra were modeled to fit resonances from water, IMCL, EMCL, creatines, and taurine (no choline signals visible at this TE). To correct for motion/vibration-based signal attenuation, the IMCL amplitude from each spectrum was used as scaling reference. IMCL doesn’t show any appreciable diffusion attenuation5 or temporal change during the scans, while the EMCL signal can readily change with slight patient motion.
Schematic diagram of the semi-Laser sequence used with two alternatives for imprinting diffusion-weighting along individual or multiple spatial directions. The oscillating diffusion gradient shape formed by modified cosine oscillations including a stretching exponent is shown in A). The effective diffusion time (derived from the gradient modulation spectrum) can be chosen within allowed limits of slew rate and stimulation thresholds (see Ref. 11). The pulsed trapezoidal gradients shown in B) are arranged symmetrically around the last adiabatic 180° pulse with an effective diffusion time TD=(Δ-δ/3).
Fig. 2: Spectra illustrating quality in single scans (48 acquisitions) of tibialis anterior with diffusion-weighting of $$$b=$$$800 s/mm2 along one direction perpendicular to fibers. Spectra contain resonances from water (unsuppressed, strong T2-weighting), creatine/phosphocreatine (dipolar-split peaks; triplet: 3ppm; negative doublet: 3.9ppm), taurine (mostly negative peak-pattern at 3.2-3.5ppm due to J- and dipolar coupling), Intra- (IMCL) and Extra-Cellular Lipids (EMCL) with main peaks at 0.9, 1.3, 5.3ppm for IMCL and shifted by ~0.2ppm for EMCL.
Top: pulsed gradients (TD=130 ms, δ=10 ms, G=29.7 mT/m)
Middle: oscillating gradients (TD=7.5 ms11, G=41.3 mT/m)
Bottom: average of 5 scans without additional diffusion weighting.
Fig. 3: Sample spectra illustrating resulting spectra when applying the diffusion-weighting along different directions. The lower half shows raw spectra contaminated by eddy-current artifacts that vary according to diffusion direction (x/y; negative/positive amplitude, particularly evident from the lineshape distortion of the main creatine resonance at 3ppm). Equivalent changes are seen for water and IMCL motivating the use of A) reference lineshapes defined from the water peak and B) reference amplitude taken from the IMCL peak that is not diffusion-weighted in area because of close-to-zero ADC. The upper part illustrates eddy-current-correction using the fitted water line.