Synopsis

Keywords: Liver, Spectroscopy

Multinuclear spectroscopy requires the use of free breathing acquisitions to achieve acceptable SNR. Here proton spectroscopy is used as an analogue for multinuclear MRS along with dynamic phase imaging to characterise the effect of respiration. Consistent agreement was seen between the respiratory bellows trace and the timeseries of the water peak frequency and imaging field offset, with overlapping primary frequencies on respective Fourier transforms. On free breathing a maximum frequency shift of 31Hz was measured, this equates to a 0.24ppm shift which could cause significant line broadening in the spectra of multinuclear ³¹P and ¹³C liver experiments.

Introduction

Multinuclear MRS of the liver is increasingly used to understand the body’s metabolic process and characterise diseases such as Type-II-Diabetes^1,2. Due to the low natural abundance of MR visible nuclei, multiple averages are needed, so spectra are usually acquired during free-breathing^1,2,3. However, the impact of free breathing on the spectral line width has not been studied in detail. Here, proton (¹H) spectroscopy is used as an analogue for X-nuclei to allow high SNR spectral data to be obtained for the assessment of the effects of respiration on spectral peaks. The results are corroborated with B₀ field maps and extrapolated to X-nuclei applications.

Methods

Acquisition: Six participants (23-30yrs, BMI 21-30kg/m³) were scanned on a Philips Achieva 3T. To model the effect of unlocalized surface coil acquisitions (as is common in multinuclear MRS scanning), proton data were collected using a butterfly ¹H quadrature surface coil (Pulseteq, Surrey) placed centrally over the liver. A respiratory belt was placed over the diaphragm to record the respiratory trace.

Data collected comprised: 1) Dynamic free breathing non-localised ¹H spectroscopy (TR=350ms) of 1800 spectra); 2) Free-breathing single slice single echo FFE scheme images (FFE TE/TR=2/4ms, image acquisition time 532 ms to encode the breathing cycle, 1200dynamics) with magnitude and phase data saved to dynamically study respiratory-induced field offsets; 3) Dual-echo FFE (TE/DTE =2.3/2.3ms, 12 slices) scanner-generated B₀ maps and magnitude data acquired on inspiration and expiration breath-holds, these data sets were collected using both the ¹H decoupling coil of the ¹³C surface coil and the Q-body coil.

Spectral Analysis. Spectra were phase corrected in JMRUI and AMRES was used to quantify the water peak frequency. This was extracted and imported into MATLAB (MathWorks). A time series of the water peak frequency was generated (f_peak(t)) and this was overlayed on the signal from the respiratory bellows; both signals were Fourier transformed.

Image Analysis: The pipeline is shown in Fig.1. The free-breathing time-series of phase images $$$\phi \left ( r,t \right )$$$ were unwrapped^4,5. The magnitude image time series was normalised and masked to remove adipose fat under the coil and voxels below the noise floor (mask generated in MRIcron). These normalised images indicated the weighting of signals from different parts of the liver in spectral signals collected with the same coil,$$$\omega\left ( r,t \right )$$$.A weighted phase evolution time series was then computed $$$ \left ( \omega\left ( r,t \right ) \phi \left ( r,t \right )\right ) $$$. A stack of line profiles through the liver across dynamics was created to demonstrate the effect of respiration. The maps of weighted phase evolution were then converted to a time-series of weighted field offsets $$$B_{0}\left ( t \right )= \frac{\sum \omega \left ( r \right )\phi \left ( r \right )}{TE\sum \omega \left ( r \right )} $$$which were overlaid on the respiratory bellows; as above, both signals were Fourier transformed. This field offset provides an imaging comparator to the frequency shift measured from the spectral analysis.

In addition, for the breath-held B₀ maps collected using the ¹H decoupling coil, the weighted field offset was computed between inspiration and expiration. The magnitude images from the Q-body breath-held B₀ maps were segmented to determine lung volume (3-class segmentation in FAST, FSL with lung cluster identification).

Binning the spectral f_peak(t) and imaging B₀(t) datasets into inspiration and expiration phases: First any point greater than two standard deviations above the mean was discarded. Next peaks/troughs that were >30% above/below the mean, which from the comparison with the bellows data corresponded to inspiration/expiration respectively, were used for moving subtractions of (inspiration – expiration) for f_peak(t) and B₀(t) to compute the mean shifts during free breathing.

Results

Fig.2 shows the agreement between the bellows trace and the time-series of the water peak frequency f_peak(t) across participants, confirmed to oscillate at the same primary frequency by the Fourier transform. Fig.3 shows analogous results for the imaging field offset time-series B₀(t). Mean free breathing spectral shifts ranged from 10-31Hz, and imaging field offsets ranged from 6-31Hz across subjects (Fig.4). The B₀ maps collected on inspiration and expiration showed consistently larger field shifts, ranging from 16-13Hz, with changes linked to the change in lung volume between inspiration and expiration (Fig.5).

Discussion

Many multinuclear spectroscopy experiments utilise free breathing to collect multiple acquisitions and increase SNR. This study investigated this approach by using the increased SNR of proton MRS to study the impact of free breathing on the spectral shift of the water peak and compare this to field offsets measured using imaging. Shifts in spectral resonant frequency and field offsets followed the respiratory cycle with good agreement, with a maximum shift during free breathing of 31Hz (0.24ppm). The field offsets were mostly smaller than the spectral frequency shifts, a likely cause being the single-slice dynamic field maps; slightly larger field shifts may occur in a different region.

At 3T, ³¹P metabolite linewidths range from 0.34-0.73ppm⁶, meaning shifts of 0.24ppm during free breathing could cause significant line broadening in some experiments. This work provides key information needed to design dynamic shim systems to overcome respiratory-induced field shifts in the liver; in particular, the spatial variations in the field maps can be used to design dynamic shim gradient coils.

Acknowledgements

Thank you to the MRC IMPACT DTP for funding this project.