Synopsis

New hyperpolarized-¹³C MRI instrumentation and methods enabled phase II clinical studies on 17 prostate cancer patients using 3D dynamic CS-EPSI with 0.5cm³ spatial and 2s time resolution. New manufacturing processes for ¹³C-pyruvate provided more consistent polarization levels, pyruvate concentrations, radical filtrations for reliable and safe patient injections. Improved quantitative methods were designed and implemented for reproducible kinetic modeling, correction for receiver profile, and peak detection, accounting for B₀ inhomogeneity and noise characteristics. These studies have shown the ability to detect high pyruvate-to-lactate conversion rates in biopsy confirmed prostate cancer.

Introduction

Methods

Patient Studies: Seventeen prostate cancer patients were studied in this phase II HP-¹³C MRI research to this date, whose pathological grades range from low-grade Gleason 3+3 to the aggressive 4+5. GMP-grade pyruvic acid was polarized in a 5T SPINlab polarizer for 2.5-3 hours, yielding 242±13mM solution with 37.8±6.1% polarization, 0.9±0.3uM radical and 31.7±0.9°C temperature. The new manufacturing process of sterile ¹³C-pyruvate greatly improved reliability and efficiency, and significantly reduced cost. A 3D compressed-sensing CS-EPSI sequence provided 0.5cm³ volumetric and 2s temporal resolution with low-power, B1-insensitive RF pulses^1,3,4.

Improved Quantitative Methods: Two types of models were applied to estimate pyruvate-lactate conversion rate k_PL. Model 1 was a two-site exchange model with a Gamma-variate arterial input function⁵. Model 2 was an assumption-less model⁶. A coil sensitivity correction⁷ was applied to improve quantitative accuracy. The HP-¹³C maps were normalized by the reception B1 profile map of the endorectal coil to remove receiver profile weighting. The phased sum of spectrum used zero-order phase correction for each metabolite peak.

Results and Discussions:

Figure.1A shows pyruvate signal appearing at ~15s after the end of injection in this patient. To investigate the variability across the patient cohort, bolus deliveries were represented by selecting arterial voxels adjacent to the prostate. The arterial signal curve was corrected by the variable RF flip angles, generating an equivalent magnetization curve across time (Fig.1B). The mean injection duration for these patients was 10.7±0.7s, and the mean bolus time in the arterial voxel was detected at 26±6s post-injection (Fig.1C). This indicated that pyruvate bolus delivery was variable among the 17 patients of this study under the similar injection and acquisition timing.

To quantitatively analyze cancer metabolism, we endeavored to develop and test methods for the accurate and robust evaluation of k_PL. Model 1 and 2 produced similar k_PL maps within the prostate in this patient who had biopsy-confirmed prostate cancer (Fig.2A). Also, comparable fit curves and k_PL estimates (0.0235 vs 0.0216s^-1) were found in a voxel with high lactate signal (Fig.2B). This highlights the importance of choosing an appropriate model to represent the underlying pyruvate-lactate conversion dynamics.

Figure 3A illustrates a typical map of pyruvate area summed over time before and after reception profile correction⁷. The correction normalized the strong pyruvate signal seen in the high intensity regions near the receive ER coil. While such corrections do not affect k_PL estimation, it benefits the study of pyruvate perfusion and uptake^8,9, providing relative quantification across the FOV (Fig.3B).

Another important task is to quantify spectral peaks. On average, less than 0.5ppm of B₀ off-resonance shift was observed within the prostate FOV. Such shifts may slightly affect the sum area under peaks, but can easily be corrected using cross correlation between the spectrum of interest and a “reference spectrum (Fig.4A).” Peak detection under magnitude mode of the spectra suffers from noise bias, while a phase-sensitive peak detection using phase-corrected spectrum follows a Gaussian noise distribution, which is zero-mean and has well-understood statistical behavior (Fig.4B).

In summary, the improved acquisition strategies and data analysis allowed better characterization of prostate cancer metabolism in patients. These developments are important for future larger population studies correlating HP-¹³C MR data with pathological information from either biopsy or post-surgical whole-mount histopathology.

Conclusions

Acknowledgements

References

[1] Chen HY et al., Phase II Clinical Hyperpolarized 13C 3D-Dynamic Metabolic Imaging of Prostate Cancer using a B1-insensitive Variable Flip Angle Design. Proceedings of the 25th Annual Meetings of ISMRM 2017;

[2] Gordon JW et al., Human Hyperpolarized 13C MRI using a Novel Echo-Planar Imaging (EPI) Approach. Proceedings of the 25th Annual Meetings of ISMRM 2017;

[3] Hu S et al., 3D Compressed Sensing for Highly Accelerated Hyperpolarized C-13 MRSI With In Vivo Applications to Transgenic Mouse Models of Cancer. Magn. Reson. Med. 2010; 63(2):312-21.

[4] Larson PEZ et al., Fast Dynamic 3D MR Spectroscopic Imaging with Compressed Sensing and Multiband Excitation Pulses for Hyperpolarized C-13 Studies. Magn. Reson. Med. 2011; 65(3):610-9.

[5] Zierhut ML et al., Kinetic modeling of hyperpolarized 13C1-pyruvate metabolism in normal rats and TRAMP mice. J. Magn. Reson 2010; 202(1), 85-92.

[6] Maidens J et al., Optimizing flip angles for metabolic rate estimation in hyperpolarized carbon-13 MRI . IEEE Trans. Med. Imag. 2016; 35(11), 2403-2412.

[7] Noworolski SM et al., Post‐processing correction of the endorectal coil reception effects in MR spectroscopic imaging of the prostate. J. Magn. Reson. 2010; 32(3), 654-662.

[8] von Morze C et al., Simultaneous Multiagent Hyperpolarized 13C Perfusion Imaging. Magn. Reson. Med. 2014; 72(6), 1599-1609

[9] Chen HY et al., Assessing Prostate Cancer Aggressiveness with Hyperpolarized Dual-Agent 3D Dynamic Imaging of Metabolism and Perfusion. Cancer Res. 2017; 77(12), 3207-3216