Cardiac proton MR spectroscopy is a challenging technique to perform in a reliable manner(1). The aim of this study is to optimize the reproducibility of cardiac measurements with respect to RF power optimization, B0 shimming, the measurement point within cardiac cycle, the method of respiratory motion correction and the choice of water suppression technique. Finally, we use these optimized parameters to measure the intra- and inter-session reproducibility in healthy volunteers.

Data acquisition parameters were optimized using forty-one healthy subjects (Table1).

1. Point resolved spectroscopy (PRESS) spectra were acquired using a global- or local power optimization. The global power optimization determines the optimal power setting over a transverse slice while local power optimization uses a PRESS sequence with a tip angle sweep of the two slice-selective refocussing pulses to determine the optimal power setting locally. The SNR of the water signal was calculated and compared for the two techniques using a Wilcoxon signed-rank test;

2. PRESS spectra were acquired using B0 shimming during free breathing and or within breath-holds. The full-width-at-half-maximum (FWHM) of the water peak was determined and compared using a Wilcoxon signed-rank test;

3. Eight different measurement times were defined throughout the cardiac cycle and PRESS (sixteen averages) spectra were acquired for each of the measurement times. The standard deviation (SD) of the water signal amplitude for the averages was calculated and plotted in a graph;

4. The water signal amplitude SD was calculated from PRESS spectra (sixteen averages) without respiratory compensation, with breath-holds and with navigator triggering (2). The amplitude variation was compared using Friedman`s ANOVA test and post-hoc Wilcoxon signed rank test;

5. The residual water signal was measured for frequency selective excitation, chemical shift selective (CHESS) and Multiply Optimized Insensitive Suppression Train (MOIST) water suppression(3,4). The residual water signal was compared using a Friedman`s ANOVA test and post-hoc Wilcoxon signed rank test;

6. Intra- and inter-session reproducibility was measured using a PRESS sequence (TE 35ms, unsuppressed 6 averages, water suppressed 32 averages) with the aforementioned optimized parameters. The myocardial triglyceride content (MTGC) was calculated for the intra- and inter-sessions. The Spearman correlation coefficient was calculated and Bland-Altman plots were constructed.

1. The spectra of thirteen of the fifteen subjects show a higher SNR using a local power optimization compared to the global power optimization, while for the remaining two subjects SNR did not differ. The mean SNR (±SD) of the water signal was 1787 (±787) using the global power optimization vs. 2173 (±626) for the local power optimization (p=0.0002).

2. The average linewidth without breathing compensation (±SD) was 19 Hz (±3.2) compared to 17 Hz (±4.6) with breath-holds (p=0.15).

3. In figure 1 the SD of the water signal averages is plotted against the cardiac trigger delay. For all subjects the SD was at its minimum with a trigger delay around 200 ms, corresponding to the highest signal stability in this region which was in line with a previous study(5). The optimal region was only minimally affected by the heart rate of the subjects.

4. In figure 2 the amplitude variation for the three respiration compensation techniques is shown. The mean amplitude variation was significantly lower for free-breathing vs breath-holds (p=0.03) and for free-breathing vs navigator triggering (p=0.03) as well as for breath-hold vs navigator triggering (p=0.02).

5. In figure 3 the residual water signal for water suppression techniques are shown. The mean (±SD) residual water signal was 7.0% (±4.8%), 1.1% (±0.9%) and 0.7% (±0.4%) for frequency selective excitation, CHESS and MOIST water suppression respectively. The mean residual signal using CHESS (p=0.01) or MOIST (p=0.01) water suppression was significantly lower than using excitation water suppression.

6. The mean (±SD) MTGC (in %) of the intra-sessions was 0.55 (±0.40) and 0.59 (±0.42) and had good correlation (r=1.000, p<0.0001) with a mean difference of 0.04 with 95% limits of agreement of -0.11 to 0.04 (fig 5A). The mean (±SD) MTGC (in %) of the inter-sessions was 0.60 (±0.42) and 0.63 (±0.45) with also good correlation (r=1.000, p=0.0004) and a difference of 0.03 with 95% limits of agreement of -0.15 to 0.9 (fig 5B).

The optimal protocol to perform cardiac MR spectroscopy was found to be a PRESS localized scan using a separate suppressed and unsuppressed acquisition, local power optimization, pencil beam B0 shimming, cardiac trigger delay of 200 ms, MOIST water suppression and pencil beam navigator based respiratory compensation. When all these parameters are used it is possible to get good intra-(r=1.000, p<0.0001) and inter-session (r=1.000, p=0.0004) correlation of MTGC and a reproducibility with small limits of agreement.