PURPOSE

Most MRS methods for quantification of cardiac lipid content do not account for differences in relaxation rates between water and fat or biological variations in the relaxation rates. Single breath-hold multi echo MRI methods recorded with enough echo times may yield enough time points per pixel for fitting T₂* of water and fat, but the T₂ can only be obtained from multiple spin-echoes. Using ¹H-MRS T₂ were determined for water and lipid signals in the human heart to derive relaxation corrected proton density fat fractions (PDFF). The T₂s of water and fat human myocardium can provide a basis for models to be used in MRI water-fat mapping methods with iterative fitting of proton densities.

METHODS

Eleven healthy subjects (age 40±8, range 28-53, ten female) were scanned in a 3T Siemens Verio. A 16-channel phased array coil over the chest was used for all subjects. After short axis and 4 chamber (4ch) scout images a 4ch T1 MOLLI¹ and a 4 echo VARPRO water-fat image² were recorded. Then a 20x6x35mm volume was prescribed on the ventricular septum for MRS. After B0 field mapping and shim correction a series of PRESS³ localized spectra was collected with cardiac gating. The first fully relaxed shot was recorded with TE 24ms, followed by five spectra with TE 24, 36, 53, 80 and 120ms. Depending on R-R intervals the number of signals averaged per TE was four (when R-R< 0.8s) or two ( R-R > 0.8s). The first spectrum served to estimate the saturation of the water resonance; the other spectra were used to estimate the T₂ of the water signal and summed lipid CH₂ and CH₃ signals. Spectra were analyzed with AMARES⁴ time domain fitting in jMRUI. The log of the signal intensities was fitted against TE in linear regression to derive T₂ and T₁ relaxation corrected spin densities for water (w) and fat (f) at TE=0. PDFF were then calculated as PDFF= f/(w+f) %.

RESULTS

With MOLLI the tissue T₁ were 1225±33 ms. A sample T₁ map is shown in Figure 1. Saturation factors for the water signals in MRS were 0.72±0.11 (N=11, median 0.75, range 0.52-0.85) with R-R intervals 0.88±0.13 s (range 0.72-1.06 s). The saturation factors correlated poorly with the MOLLI T₁ results. Probably because the average R-R interval recorded for the scan by scanner software was not a reliable basis for an estimate of the T₁s. The T₂ regression fit for the water signals showed good linearity: the correlation coefficients R² were 0.93±0.10 ms (N=11, median 0.98 range 0.66-1.00) and the mean T₂ found for water was 35±7 ms (N=11, median 33 range 26-47ms). In hearts with very low fat content and hence low fat signals, the longer TE spectra had insufficient SNR for reliable quantification and thus the correlation coefficients R² lower 0.70±0.27 ms (N=11, median 0.70 range 0.23-1.00). The T₂ for the lipid signal was 45±20 ms (N=10, median 40 ms range 17-92ms). In one subject the fit resulted in a negative T₂ for the lipid signal and this result was omitted. In this series of observations the T₂ of water and fat are not significantly different. The PDFF from water-fat MRI in the ROI defined by the cross section of the MRS volume with the 4ch images slice was 2.7±1.0%, (N=11, median 2.3% range 1.6-4.7%) This was higher than the T₂ corrected PDFF found in the breath-hold MRS TE series: 0.93±0.61% (N=11, median 0.88% range 0.14-2.02%) the difference was significant in paired two-tailed t-test at p< 0.001.

DISCUSSION

The measurement of water and lipid T₂s in the human heart in a single breath-hold was possible with five TE values and signal averaging. The T₂ relaxation corrected PDFF measured from the series was lower than measured with water fat imaging. MRI did not record a heart PDFF < 1.5% in any of the subjects. The water signal in these MRS acquisitions is partially saturated so the low PDFF may still be an overestimate. Measuring the T₂ of lipids proved to be problematic in subject with normal low fat content.

CONCLUSION

With favorable SNR and R-R intervals the T₂ corrected fat fraction of the healthy human heart can be measured in a single breath-hold at 3T, but for a more reliable result the acquisition of a TE series should be recorded over multiple breath hold. This however requires the breath-holds to be reproducible. The saturation of the water signal in single R-R interval repetitions is about 25% large enough to require a T₁ correction, which could be based on a T₁ mapping with MOLLI.

Acknowledgements

We would like to thank the NIH and NIDDK for providing the framework and funding for this research and also thank Jatin ‘Raj’ Matta, and Marissa Schoepp for their very essential efforts and enthusiasm.