INTRODUCTION

Hyperpolarized ¹³C MRI has shown potential for imaging cardiac metabolism of the in-vivo human heart¹. However, the quantification of metabolic alterations is limited by the low SNR due to the rapid relaxation of the hyperpolarized substrate in blood. Hence, the achievable SNR depends both on the T₁ value and on the time-to-peak of the bolus.
Furthermore, the majority of hyperpolarized MRI sequences is T₂^*-weighted, which is a function of T₂ and field inhomogeneity $$$\gamma\Delta{}\textrm{B}_0$$$:
$$\textrm{T}_2^*=\frac{1}{\textrm{T}_2}+\frac{1}{\gamma\Delta{}\textrm{B}_0}$$
To predict these effects, reliable values of relaxation parameters are required.
While pyruvate T₁ values in blood have been reported at 7T², existing T₂ values are derived from in-vivo measurements at 3T^3,4 and 9.4T⁵. Given the expected long T₁ and T₂ values in blood in comparison to substrate circulation times, a controlled setup is preferred. Additionally, relaxation parameters of hyperpolarized ¹³C pyruvate at fields <3.0T have not been reported yet. In the present work, T₁ and T₂ relaxation parameters are determined using ¹³C MRS for different pyruvate concentrations in venous blood at 1.5T and 3T.
Heart rate dependency of the achievable SNR is examined using simulations and exemplarily shown on human in-vivo data of the heart.

METHODS

The experimental setup of T₁ and T₂ relaxometry is shown in Fig-1. Different concentrations of neutralized [1-¹³C]pyruvate polarized to >40% in a SpinLab hyperpolarizer (GE Healthcare) were dissolved in venous pig blood (12.5mM-250mM) and examined on clinical 3T and 1.5T scanners (Philips Healthcare). Identical in-house built transmit-receive ¹³C volume saddle coils (diameter=3cm) were used. Neat samples (250mM) were additionally measured in a 0.58T benchtop MRI (Pure Devices). All experiments related to blood samples were approved by the Cantonal Veterinary Office (Zurich, Switzerland) and all human experiments were approved by the Cantonal Ethics Committee and Swissmedic.
T₁ relaxation was determined using dynamic pulse-acquire spectroscopy (FA=5°,TR=10s, ten dynamics). For T₂, a custom slice-selective Carr-Purcell-Meibloom-Gill (CPMG) sequence⁶ was implemented. To account for long T₂ times, 1024 echoes with an echo-spacing of 10ms were acquired. For the benchtop system, the parameters were the following: FA=10°, TR=10s for T₁, and 2000 echoes, TE=2.5ms for T₂.
T₁ and T₂ relaxation times were fitted monoexponentially from the averaged first three FID samples of each dynamic, and the spectral maximum, respectively. For T₂, even and odd echoes were fitted individually and subsequently averaged to obtain a final T₂ estimate.
To investigate heart rate dependency on the achievable SNR, a two-stage bolus timing model as shown in Fig-2A was developed: stage 1 from dissolution until injection (T₁=70s) and stage 2 from injection until myocardial bolus peak (variable, concentration-dependent T₁). The time-to-peak of the myocardial bolus (in heart beats) was determined empirically from 25 in-vivo pig datasets as shown in Fig-2B. The times between dissolution and injection (20s/90s for animal/human application) are based on experimental experience. Relative achievable SNR was determined as a function of heart rate and T₁ in blood.
Exemplary in-vivo data measured in patients with heart failure (preserved ejection fraction (EF)>=50% and mid-range EF=40-49%) are presented to illustrate the effect of heart rate on the SNR in hyperpolarized ¹³C metabolic imaging.

RESULTS

Fig-3 depicts the estimated T₁ and T₂ values for pyruvate concentrations in blood of 12.5mM, 25mM, 50mM, 100mM, 150mM and 250mM (neat) at different field strengths (3T, 1.5T and 0.58T).
Fig-4A shows the expected resulting SNR at myocardial bolus peak time as a function of heart rate and T₁. For realistic pyruvate-blood concentrations with corresponding T₁ values according to Fig-3, the achievable SNR is illustrated in Fig-4B.
Fig-5 shows exemplary human in-vivo data for high (81bpm), medium (67bpm) and low (45-55bpm) heart rates demonstrating improved SNR for higher heart rates. Based on the bolus timing according to Fig-2, this range of heart rates (81bpm-45bpm) corresponds to a time span between 25-45s for the time-to-peak of the myocardial bolus, which has an significant impact on the resulting base SNR in case of small T1 values.

DISCUSSION

To predict data quality in hyperpolarized MRI, reliable estimates of relaxation parameters T₁ and T₂ under in-vivo conditions as well as for different pyruvate concentrations in blood are required.
In addition to the bolus concentration, the achievable SNR depends on the time-to-peak in the organ of interest, and therefore on the heart rate.
Even for low pyruvate concentrations in blood, the estimated T₂ relaxation times (>5.9s) remain relatively long and therefore, T₂^* is primarily determined by field inhomogeneities including tissue susceptibility and shim settings.
While our current data demonstrate trends, further measurements are warranted to substantiate average and variation of T₁ and T₂ values in-vivo at different field strengths.
Using experimental data and simulations, we have demonstrated that the achievable base SNR in hyperpolarized MRI is strongly dependent on the heart rate of the subject. The underlying model relies on bolus timings determined from in-vivo data measured in pigs, which had relatively high heart rates (85$$$\pm$$$9bpm), while clinical patients mostly present with lower heart rates (64$$$\pm$$$9bpm). Given the significant impact of heart rates on data quality, heart rate management of patients using stressors is considered an important aspect in future study designs of experimental and human trials.

Acknowledgements

The authors would like to acknowledge Miriam Weisskopf for animal handling and care.