A promising tool to investigate human heart disease is cardiac ¹H MR spectroscopy (MRS)¹. Particularly the intramyocellular lipid (IMCL) concentration might be of interest as a biomarker for heart disease, as it represents an important energy depot of muscle cells². Therefore the independent quantification of IMCL and extramyocellular lipids (EMCL) is desirable. However, quantification of ¹H spectra remains challenging, as different metabolite signals may overlap and signal to noise ratios of in vivo MRS are rather small. This problem is aggravated in cardiac spectroscopy, due to quality limitations of spectra imposed by cardiac and respiratory motion, fluctuating and inhomogeneous B₀ fields, as well as finite scan times.

This work investigates, whether quantification of metabolite signals within cardiac MRS improves by the use of prior knowledge about the dependence of chemical shifts on the muscle fiber orientation in the quantification process.

Although IMCL and EMCL have the same chemical composition, the signals of the different compartments experience a different chemical shift, due to bulk susceptibility effects³, as was demonstrated in skeletal muscle studies. While the signal frequency of the IMCL signal remains constant, the chemical shift of EMCL exhibits an angular dependency on the muscle fiber orientation with respect to the main magnetic field (B₀)². Assuming reproducible voxel positioning with respect to cardiac fiber orientation, the EMCL frequency shift ($$$Δω_{EMCL}$$$) can be expected to be equivalent for four different angles, between the voxel and B₀ (fig. 1). Hence, $$$Δω_{EMCL}(α)$$$ might be described by equation (1) (see fig. 2). Furthermore, taurine and creatine are subject to dipolar coupling due to restricted tumbling motion^2,4. The signal of the three chemically equivalent spins of the creatine methyl group (CH₃) splits into a triplet-structure with chemical shift and amplitude of the satellite peaks depending on the average angle between the interaction direction and B₀ ($$$Θ$$$). This coupling effect is proportional to $$$3\cdot\cos^{2}(Θ)-1$$$. Hence, the amplitudes of the satellite peaks ($$$A_{Cr28}(α)$$$) of the CH₃ group at 3.03ppm can be described by equation (2) (fig. 2).¹H cardiac spectra were acquired from the interventricular septum (fig. 3a,b) of 10 healthy female volunteers (bmi: ∅21.1kg/m²) using the combination of image-based B₀-shimming^5,6, ECG-triggering and navigator-gating along with retrospective frequency-alignment and phase-correction of metabolite-cycled⁷, non-water-suppressed MRS^8,9,10. All measurements were performed at a 3T Achieva system with a 32 channel cardiac coil (Philips Healthcare, Best, NL). Spectra were postprocessed with MRecon (Gyrotools, Zurich, CH) as detailed in [10], before they were fitted with LCModel¹¹ (fig.3c). The frequency shift of the EMCL signal at 1.5ppm (EMCL15) with respect to the IMCL signal at 1.3ppm (IMCL13) was extracted from the LCModel fit results and their dependency on $$$α$$$ was investigated. The metabolite concentrations calculated by LCModel, are referenced against creatine. However, the concentrations of the satellite peaks of the CH₃-triplet are not considered. This leads to an overestimation of metabolite/Cr ratios, which depends on the angle. Hence, the Cr28/Cr ratio, which is the concentration ratio of the CH₃ satellite peaks at 2.8ppm to the main creatine peak, was taken from the LCModel fits and analyzed with respect to $$$α$$$. Metabolite concentrations were corrected by taking the Cr28/Cr ratio from LCModel into account. Additionally metabolite concentrations were corrected by $$$2\cdot A_{Cr28}(α)$$$, as the CH₃ resonance splits into a triplet and the satellite peak amplitudes depend on $$$α$$$.In fig. 4 $$$Δω_{EMCL}$$$ is plotted versus $$$α$$$, together with a fit of equation (1). Two data points were excluded due to insufficient data quality. It can be seen that $$$Δω_{EMCL}$$$ follows the angular dependency in equation (1) closely. The results are in good accordance with the results from previous studies in skeletal muscle². Fig. 5a plots Cr28/Cr ratio over $$$α$$$ and from the fit of equation (2) can be seen that the data points follow the suggested mathematical relation. In case of correlating (Pearson) unsaturated IMCL (I21) with the volunteers BMI (fig.5b) the creatine correction schemes lead to a higher regression coefficient R². While such a linear correlation of unsaturated IMCL and BMI seems plausible, further studies are required to establish if this correlation is indeed to be expected in healthy subjects and whether correlation changes could arise in disease.Prior knowledge on the angular dependence of the chemical shift of EMCL signals and the signal splitting of dipolar coupled spins allow for prediction of EMCL frequencies and the amplitude of the satellite peaks of the CH₃ signal. This information improves accuracy and precision of the quantification if used in correction schemes. The feasibility of separate quantification of EMCL versus IMCL in the myocardium is confirmed.