Diffusion weighted spectroscopy (DWS) is limited by a low SNR, especially for high b-values^1,2. Hence, averaging over many measurements is inevitable, which prolongs the measurement time and, thus, leads to a high susceptibly for motion and line broadening artifacts³. The usage of a non-water-suppressed (nWS) sequence could alleviate some of the problems allowing the use of the water signal as inherent reference, not only for optimal frequency-, eddy current-, and phase-correction as in previous applications⁴, but also for compensation of motion-related signal drops. This work aims to demonstrate that the measurement of diffusion coefficients (ADC) of brain metabolites substantially profits from using water as internal reference for signal correction in MRS with metabolite cycling (MC) instead of water presaturation.

A DWS sequence based on STEAM⁵ was extended to include: (i) adiabatic MC pulses^4,6 implemented in the TM period and (ii) adiabatic inversion recovery for FLAIR CSF suppression. The sequence is illustrated schematically in Fig. 1. The sequence was tested in a “braino” phantom with aqueous solutions of typical brain metabolites and in gray matter (GM) of three healthy volunteers (VOL). Acquisition parameters included: TE/TM/TR=37ms/150ms/2500ms; diffusion gradient length δ=11ms; diffusion time Δ=168.2ms and maximum amplitude G_max=38mT/m. The crusher gradient strength and duration was optimized to avoid interfering spurious echoes and to minimize eddy currents. Maximum b-value in a single direction was 5236s/mm² with 32-128 acquisition per spectrum depending on the b-value. Spectra were acquired on a 3T Siemens TRIO scanner using a multi-channel headcoil. Single shots were stored, though already combined into a single FID (phase-corrected based on 1^st data point). Spectral fitting was done in a 2D-fashion in FiTAID⁷ (simulated metabolite basis sets, experimental macromolecule base spectrum), including ADC determination with a mono-exponential model.

Motion-compensation: Since spectra were recorded without triggering, single acquisitions are affected by various degrees of brain motion, evidenced by a signal drop. This drop was quantified in the non-suppressed water signal (using the median of the top quartile of all shots as reference level) and, thus, all single acquisitions were scaled up to this level before frequency alignment and signal averaging. Only acquisitions that increased the overall SNR were added in. The resulting data was used to determine the water and metabolite signals by adding or subtracting the up- and downfield inverted acquisitions, followed by eddy current correction⁴.

Fig. 2 summarizes the results obtain in the phantom, in particular the fact that ADCs for metabolites and water are obtained simultaneously – with all components showing mono-exponential decay as expected and the ADCs agreeing well with the literature (Fig. 2b,c and 4). In the in-vivo study, the water signal could be used as reference line for spectral correction and motion-related intensity compensation up to the highest b-value. The effect of eliminating the scans with largest motion-distortion and of intensity scaling is illustrated in Fig. 3. Besides the increase in signal intensity at high b, it can be appreciated from Fig. 3a,b that the lineshape becomes more symmetric and that ghosting artifacts are reduced. Fig. 3c illustrates that the motion-corrected water signal exhibits a non-mono-exponential decay. However, the metabolic attenuation is almost mono-exponential up to b=5236s/mm² (cf. Fig. 3c inset). Estimated ADCs for our initial 3 subjects are smaller after the correction for motion-related signal drops (Fig. 4).In DWS, even weak motion causes enhanced signal attenuation at high b-values, which can easily be misinterpreted as faster diffusion. Using the co-recorded high SNR peak of water, this attenuation can be quantified in single shots and a correction applied to the spectra. As seen from Fig. 3a the signal attenuation for the motion-corrected spectra is thus much lower at high b-values than for the non-motion-corrected case. Furthermore, the fact that the lineshape becomes more symmetric with elimination of outliers (and also spectral reference-based corrections like frequency-alignment) allows for better resolution of metabolites with multiplet patterns like mI or Glu and thus improved model fitting. The current fitting model is preliminary, possibly accounting for apparent intersubject differences.It is demonstrated, that DWS in conjunction with MC can provide spectra with equivalent quality to those with common water suppression techniques. In addition to eliminating any influences of water-exchange, it is shown, that the information provided by the water signal can be used for motion-correction, not only in terms of easy zero-order phasing, spectral re-alignment and eddy-current correction, but also compensation for motion-related signal drop. Overall, this leads to improved spectral resolution, more stable fitting, and determination of ADC values that are closer to the true values, and hence better physiological interpretation.