This work aimed to use cardiac 3T 1H-MRS to quantify creatine concentration (“[Cr]”) in the human heart, because creatine plays a key role in cardiac energy metabolism. We implemented a cycled water-suppression scheme (WC) into the PRESS and STEAM pulse sequences on a Siemens 3T Prisma MRI scanner. Compared to the vendor’s product PRESS sequence, we found that PRESS-WC was able to measure [Cr] more than 2 times more reproducibly and in a shorter scan. Using PRESS WC in a 10-breathhold protocol, we were able to detect a decrease in creatine concentration in 5 patients suffering from aortic stenosis.
All measurements were performed on a 3T scanner (Prisma, Siemens) using an 18-channel body-array coil (Siemens) in combination with a spine coil array (Siemens).
Phantom experiment
Data were acquired from phantoms containing 6 different [Cr] (0, 5, 10, 20, 40 and 80mM) using three sequences:
In each case, 50 measurements were obtained, followed by 3 non-water-suppressed measurements. TR was 4000ms throughout. Complex weights for coil combination were calculated from the non-water-suppressed data. The residual water peak was used for phase-correction and frequency alignment before signal averaging.2 Spectra were fitted using the OXSA toolbox.4
Healthy controls
10 healthy volunteers (3 females, mean age = 29.3 ± 4.0 years) were scanned in two sessions on the same day. Table 1 describes the protocol. After each sequence, a non-water-suppressed data set was obtained (1 breath-hold (BH), 3 measurements). All MRS data were obtained at end expiration using ECG triggering from a 12.6cm3 voxel centred on the mid-interventricular septum.
For direct comparison, 12BH were randomly selected and analysed from the STEAM WET data (giving 60 measurements to match the STEAM WC data). SNR and Cramér-Rao lower bounds (CRLB) of the Cr-CH3 peaks were compared between the 4 sequences. [Cr] was obtained by correcting the Cr-CH3/water peak ratios for relaxation and number of protons while fat fraction was computed with the lipid CH2 peak area. Reproducibility data were analysed for 30BH STEAM WET and 10BH PRESS WC.
Patient group
5 patients with aortic stenosis (2 females, mean age = 74.8 ± 6.6 years) due to undergo valve replacement surgery were recruited. Data from 10BH PRESS WC acquisitions were analysed as described above. In 4 cases, intra-operative surgical biopsies of left ventricular endocardium were flash frozen in liquid nitrogen for biochemical analysis. 2 samples are still awaiting analysis.
Phantom experiment
Figure 1 shows that both sequences are sensitive over the typical physiological range of [Cr].
Healthy controls
Figure 2a depicts a set of representative spectra from the compared sequences. Although there was a slight decrease in SNR, STEAM WC improves the fitting of the spectra (lower CRLB value) compared to the product STEAM WET. This allows better Cr quantification (Figure 2b and Figure 2c). This fitting is further improved with PRESS WC implying a more accurate measurement of myocardial Cr.
The [Cr] and fat fraction range obtained with both sequences agrees with literature values.5,6 Figure 3 compares the reproducibility of the sequences. PRESS WC shows an increased reproducibility for both measurements, even when using only 10BH. A simple power calculation of the sample size needed to detect a 15% change in [Cr] decreases from 20 for STEAM WET to 3 for PRESS WC. This is of great benefit for clinical studies.
Patient group
All patients tolerated the exam well. One patient had both [Cr] and fat fraction more than 8 times the mean values. We suspect that the voxel had moved away from the septum during the water reference scan and subsequently excluded this data set. Figure 4 compares the spectra, [Cr] and fat fraction between healthy controls and patients. The increase in fat fraction is also consistent with previous studies.7