The minimum echo-time for hyperpolarized 13C echo-planar imaging can be reduced with partial sampling along the blipped direction in k-space. To investigate the extent to which echo-time shortening can improve signal-to-noise ratio, we’ve employed an experimental design that toggles between two different spatial encoding strategies during a time-resolved hyperpolarized [1-13C]pyruvate acquisition. Using clinically approved hardware with a pre-clinical animal model, we compared symmetric with asymmetric echo-planar imaging. Considerable signal-to-noise ratio gains for asymmetric vs symmetric sampling were observed without artifacts. On the basis of this study, our group will employ asymmetric sampling in our forthcoming human trials.
Hyperpolarized (HP) 13C magnetic resonance imaging and spectroscopy with dynamic nuclear polarization (DNP) substrates is an emerging technology for monitoring cellular metabolism in vivo. Due to favourable DNP properties and metabolic significance in diseases such as heart failure and cancer, studies involving the substrate [1-13C]pyruvate are driving current progress in translating HP 13C MRI for clinical use, and several human trials have been reported.1-10
Some of our group’s forthcoming HP [1-13C]pyruvate trials focus on imaging solid tumours in a variety of patient cohorts, and utilize spectrally and spatially selective radio frequency (ssRF) excitation followed by 3D echo-planar imaging (EPI) readouts (2D EPI + phase encoding along the “slice” direction) to generate time-resolved, volumetric images of [1-13C]pyruvate and [1-13C]lactate.
In the ssRF + “fast-readout” regime, pulse durations can be on the order of tens of milliseconds, and this invariably extends the minimum echo-time (TE) in comparison to approaches that utilize non-selective RF excitation. Longer TE entails heavier T2* weighting and this can lead to degradation of signal-to-noise ratio (SNR). For a fixed RF pulse length, the minimum TE for ssRF + 3D EPI can be reduced considerably with partial sampling along the blipped direction in k-space.
To explore SNR performance of asymmetric vs symmetric k-space sampling in EPI, we’ve employed a previously described11 experimental design that toggles between two different spatial encoding strategies during a dynamic HP [1-13C]pyruvate acquisition. By interleaving the two readouts within a single injection rather than between successive experiments, potentially confounding variables such polarization, perfusion, injection timing and metabolism can be arguably eliminated. Here we present pre-clinical results comparing asymmetric vs symmetric EPI obtained with a HP 13C human breast imaging platform12.
Symmetric EPI gradient waveforms were designed for 96×12 cm2 field-of-view at 7.5 mm in-plane resolution. For asymmetric EPI, the dephasing and rephasing trapezoids for the blipped gradient were redesigned to shift the sampling window in k-space by the equivalent of either 5 or 6 lines, reducing minimum TE by 7.6 and 9.2 ms, respectively.
Phase encoding was used for 24 cm coverage with 10 cm slice thickness. A tailored 18.8 ms ssRF pulse13 was designed for excitation. Waveforms were incorporated into a custom 3D gradient-echo sequence and were toggled according to figure 1. Odd and even time points were encoded with symmetric and asymmetric EPI, respectively.
Imaging was performed on a GE MR750 3T MR scanner (GE Healthcare, Waukesha, WI) using the previously described 2-channel HP 13C breast imaging platform (figure 2). Two Sprague-Dawley rats were handled in accordance with our institutional animal care and use committee.
For rat A (480 g), the asymmetric EPI readout was shifted by 6 lines (~9.2 ms) and images were acquired at 10 mm isotropic resolution. For rat B (500 g), the readout was shifted by 5 lines (~7.6 ms) and the resolution was increased to 7.5×7.5×10 mm3.
Time-integrated [1-13C]pyruvate and [1-13C]lactate overlays are displayed in figure 3. In both experiments, SNR within the right kidney was apparently similar for either EPI encoding method; however, there is a striking difference within the left kidney, with elevated SNR in volumes encoded with asymmetric EPI. A region-of-interest (ROI) analysis for the kidney signal was performed and is illustrated in figure 4. Area-under-the-curve (AUC) summary metrics of the ROI analysis are tabulated in figure 5.
In experiment A, the left/right kidney averaged SNR increase for [1-13C]pyruvate and [1-13C]lactate was 36% and 45%, respectively, and for experiment B, the increase was 24% and 44%. If the difference in SNR was primarily due to T2* dephasing during the readout, then the benefit in asymmetric EPI vs symmetric EPI would be expected to be higher in experiment A than experiment B, and this was observed. The voxel size used in experiment A was approximately 78% larger than in experiment B, which, all else equal, would contribute to increased intra-voxel dephasing. Additionally, 6 rather than 5 lines were skipped, providing experiment A with a shorter TE than experiment B. Despite the significantly larger voxel size in A vs. B, the total SNR was comparable, which underscores the advantage of comparing sequences within a single injection.
Asymmetric EPI is a viable method for shortening minimum TE in order to mitigate T2* related SNR reductions in the ssRF regime. In the cases studied, the SNR advantage of asymmetric-vs-symmetric EPI was in the range of 20-50%. A potential drawback of asymmetric-vs-symmetric EPI sampling is ringing artifacts, which were not observed here. On the basis of this study, our group will be employing asymmetric EPI in our forthcoming human trials.
The authors are thankful for funding from the Canadian Breast Cancer Foundation Ontario and the Canadian Cancer Society Research Institute and for assistance from Jennifer Barry and Yiping Gu for animal handling.