The creatine kinase enzyme regulates the rate of creation of adenosine triphosphate (ATP) from phosphocreatine (PCr) in myocardial myofibrils.¹ The creatine kinase forward rate constant k_f^CK is a sensitive biomarker for heart failure.²

Phosphorus magnetic resonance spectroscopy (³¹P-MRS) at 1.5 and 3T has previously been used to measure k_f^CK in the human myocardium.³ At 1.5T and 3T ³¹P-MRS suffers from low SNR, so k_f^CK had to be measured with a low spatial resolution, e.g. using 1D chemical shift imaging (CSI).

The SNR for cardiac ³¹P-MRS increases by 2.8x at 7T compared to 3T.⁴ Our aim is to use this increased SNR to obtain the first 3D-localised measurements of k_f^CK in the human heart. Higher resolution decreases confounding signal contamination, and is the first step towards k_f^CK measurements in localised myocardial disease.

Saturation transfer (FAST³, TRiST⁵ and TwiST⁶) and progressive saturation⁷ methods have been implemented for human cardiac k_f^CK measurements. The FAST method, with no restraint on the sequence T_R, is the most extensible to 3D localisation.

FAST makes two dual-angle T₁ measurements: one while saturating the γ-ATP peak, and one with control saturation. Each measurement comprises identical acquisitions, with different flip angles: α (15^o) and β (60^o).⁸ The original 1D-CSI FAST implementation, using BIR-4 pulses for uniform excitation, comprised 4 acquisitions.

After measuring T₁s, with and without saturation (T₁’ and T₁), the corresponding magnetisations with and without saturation (M₀’ and M₀) are calculated. k_f^CK is given by: $$k_{\mathrm{f}}^{\mathrm{CK}}=\frac{1}{T_{1}'}(1-\frac{M_{0}'}{M_{0}}).$$

Unfortunately, the $$$B_{1,\mathrm{peak}}^{+}$$$ on a Siemens Magnetom 7T scanner falls below the adiabatic onset of BIR-4 pulses. Therefore, we replaced BIR-4 excitation with conventional pulses and added a $$$B_{1}^{+}$$$-mapping step to the protocol, using a ³¹P-MRS Bloch-Siegert method.⁹

The precision of the k_f^CK measured by FAST was estimated, using Monte Carlo simulations, in the presence of errors in flip angle. The error from the uncertainty in the B₁ maps was compared with the error from imperfect excitation by BIR-4 pulses.¹⁰

This “FAST+Bloch-Siegert” method was validated in one subject’s calf muscle, which has high SNR and homogeneous tissue, at 3T, with a volume coil for uniform excitation, and at 7T, with a surface coil. The 10-cm loop transmit-receive surface coil was also used for the cardiac protocol.

Four healthy subjects (male, 29±6yrs and 75±7Kg) underwent the cardiac protocol. Subjects were positioned and localizers and calibration scans were acquired as previously described.⁴ The protocol comprised 3D-CSI ³¹P-MRS acquisitions with a 16x16x8 acquisition weighted CSI matrix (8 along the heart long-axis), 240x240x200mm³ FOV, 6000Hz bandwidth and shaped excitation pulse.⁴

The scan protocol and processing is described in Figure 1. In the Bloch-Siegert prepared acquisitions, a Fermi pulse ⁹, at ±2000Hz, was added after the excitation. The $$$B_{1}^{+}$$$ is used to scale the following, four constituent, FAST acquisitions’ excitation flip angles, to 15^o for α and 60^o for β in the interventricular septum. Selective saturation was achieved by DANTE (T_R=330μs, T_P=100μs). The total ³¹P-MRS acquisition time was 80 min. Data was processed and fitted using AMARES in Matlab.

The error in k_f^CK from the uncertainty in $$$B_{1}^{+}$$$ maps was computed as being <±20% which is comparable to that from BIR-4 pulses (6.1-23.4%).

The k_f^CK measured in skeletal muscle was 0.27±0.10 s^-1 (3T: all purely muscle containing voxels), and 0.31±0.07 s^-1 (7T: all voxels within 10% of target flip angles).

The k_f^CK in the target myocardial voxels of the 4 subjects was 0.39±0.15 s^-1, 0.31±0.17 s^-1, 0.39±0.21 s^-1 and 0.35±0.09 s^-1 (value ± SD). The study average was 0.37±0.05 s^-1.

k_f^CK values recorded using the "FAST + Bloch-Siegert" method fall within the range of literature values for cardiac and skeletal muscle. The values in this study have been obtained from voxels centred on the myocardium, with voxel sizes (FWHM of the point-spread-function) <50% those obtained in any previous study.

Simulations show that the precision and accuracy of k_f^CK drop to <50% if the flip angles deviate from their target values by more than 20%. The inhomogeneous flip angles produced by surface coils mean that the target flip angles are only obtained across a few voxels in the left ventricle. A coil that produces more uniform excitation, e.g. a larger loop or a volume coil, would allow more voxels to be quantified simultaneously.

The acquisition time is limited by the SNR of the method. Receive array coils offer a 1.8x increase in SNR for cardiac ³¹P-MRS at 7T; equivalent to a 3.2x shorter scan duration.¹¹