0725

Phase II Clinical Hyperpolarized 13C 3D-Dynamic Metabolic Imaging of Prostate Cancer using a B1-insensitive Variable Flip Angle Design

Hsin-Yu Chen^1,2, Peder E.Z. Larson^1,2, Jeremy W. Gordon¹, Robert A. Bok¹, Marcus Ferrone³, Mark van Criekinge¹, Lucas Carvajal¹, Rahul Aggarwal⁴, James B. Slater¹, Ilwoo Park¹, Eugene Milshteyn^1,2, Sarah J. Nelson^1,2, John Kurhanewicz^1,2, and Daniel B. Vigneron^1,2

¹Department of Radiology and Biomedical Imaging, University of California, San Francisco, San Francisco, CA, United States, ²UCSF/UC Berkeley Graduate Program in Bioengineering, University of California, San Francisco, San Francisco, CA, United States, ³Department of Clinical Pharmacy, University of California, San Francisco, San Francisco, CA, United States, ⁴School of Medicine, University of California, San Francisco, San Francisco, CA, United States

Synopsis

The 3D compressed-sensing echo-planar spectroscopic imaging (CS-EPSI) sequence has enabled 5-dimensional HP-¹³C imaging of human prostate cancer with full-prostate coverage and high spatial and temporal resolution. A new variable flip angle design substantially desensitizes the sequence to B₁ variations for improved quantitative analysis of pyruvate-to-lactate rate constant (k_PL) measurements. Pre-clinical animal and clinical patient studies demonstrate that the resulting sequence enabled improved quantitative accuracy while maintaining excellent metabolite SNR.

Purpose

The 3D compressed-sensing echo-planar spectroscopic imaging (CS-EPSI) sequence enables 5-dimensional (3-spatial, 1-temporal, 1 spectral) dynamic HP-¹³C imaging of human prostate cancer with high spatiotemporal resolution¹. A variable flip angle scheme has been applied to improve the metabolite SNR, but the quantitative accuracy of pyruvate-to-lactate conversion can potentially be affected by B₁⁺ inhomogeneity (ΔB₁⁺) with this scheme. Calibration of B₁⁺ is challenging due to the limited natural abundance of ¹³C in vivo, and can also vary across larger imaging volumes. In this study a new variable flip angle scheme was developed to reduce the sensitivity to ΔB₁⁺ variations across the transmit-coil volume. In vivo HP-¹³C studies and simulations were conducted to explore the sensitivity of quantitative k_PL estimation to ΔB₁⁺, as well as changes in total [1-¹³C]pyruvate and [1-¹³C]lactate SNR.

Methods

Sequences: The CS-ESPI sequence enables 3D dynamic imaging with high spatiotemporal resolution¹. Spectral-spatial excitation pulses for varying flip angles between metabolites were designed using an open-source SS-RF toolbox². Sequence parameters include TR = 150ms, TE = 4.0ms, spatial resolution: 0.36cm³ (rats), or 0.5cm³ with FOV=9.6x9.6x12.8cm (humans), temporal resolution = 2s, acquisition window = 42s.

MRI experiments: GMP [1-¹³C]pyruvic acid was polarized in a 5T SpinLab polarizer for 2.5 hours, yielding 35+% polarization of 245mM pyruvate and met all pharmacy specifications prior to injection for the clinical trial study conducted in accordance with the approved IND protocol. A healthy Sprague Dawley rat was used for animal studies.

Simulations: HP-¹³C pyruvate and lactate signals were generated and fit using a simple and robust two-site exchange model⁴. B₁ inhomogeneity translates into an effective scaling in a given flip angle scheme, denoted by Δθ_n. Potential variations of ΔB₁⁺= ±20% was assumed for clinical environment.

Results and Discussions

One way of desensitizing the sequence to B₁ inhomogeneity is to adjust the effective T₁ design factor (T_1,design) used to generate the lactate flip angles⁵. At earlier timepoints, higher amount of magnetization is present, and is thus more susceptible to flip angle variation due to ΔB₁⁺. Increasing T_1,design reduces the flip angle for early timepoints (Fig.1A), thereby making the design less sensitive to ΔB₁⁺. Deviating from the “optimal” design parameters comes at the cost of reduced total lactate SNR. Simulations suggest that while the original T_1,design =30(s) generates maximum signal, a choice of T_1,design =100(s) significantly reduces B₁-sensitivitiy (Fig.2A, Δk_PL from 12.5% to 4.3% for ΔB₁⁺= -20%, 13.7% to 7.1% for ΔB₁⁺= 20%) without substantially reducing SNR (Fig.1B, 2.5% lactate tSNR loss).

Additionally, preventing the final flip angle from increasing to 90° helps to mitigate slice profile artifacts⁶. Such thresholding does not have a significant impact on k_PL errors in terms of B₁ inhomogeneity, and a 60° threshold was selected based on past simulations⁶ (Fig.2B). The resulting new spectral-spatial RF pulses using this variable flip angle scheme were 4.0 (ms) long with peak B₁ of 0.492(G), 37% shorter and 17% reduced in peak power than our previous clinical design¹.

HP-¹³C imaging of rat kidney measured substantially lower Δk_PL using the new flip angle scheme optimized for B₁ insensitivity compared to a previous scheme (T_1,design =30(s)) in the presence of ΔB₁⁺ (Fig.3, Δk_PL = 22% vs 63% with ΔB₁⁺= -20%), which agrees with simulation.

For human studies, the dynamic 3D CS-EPSI acquisition with the improved variable flip angle scheme enabled coverage of the entire prostate gland with good spatiotemporal resolution and high SNR (Fig.4A&5A) with spectroscopic information (Fig.4B). SNR was 68.5 for pyruvate and 44.4 for lactate 41 seconds from injection. The new variable flip angle scheme not only retained excellent pyruvate and lactate SNR, but also preserved the signal timing characteristics compared to our prior studies¹. The location of high k_PL (Fig.5C) corresponded with a large volume of biopsy-proven Gleason 4+5, 4+4 and 4+3 cancer involving a majority of the left lobe of the prostate and extending into the right peripheral zone.

Conclusions

The 3D CS-EPSI sequence with an improved variable flip angle scheme enabled quantitative analysis of pyruvate-to-lactate conversion with far less sensitivity to B₁⁺ variations. Moreover, the clinical research studies demonstrated the potential to acquire new information in patient exams while preserving the desirable SNR efficiency of the previous design. Improved ΔB₁⁺ insensitivity is beneficial for all HP-¹³C applications, where calibration of this field is challenging due to the limited natural abundance of ¹³C in vivo. Furthermore, the proposed variable flip angle schemes will be particularly valuable for human studies over large volumes, e.g. brain and liver, where substantial ΔB₁⁺ will likely be present across the FOV.

Acknowledgements

The authors acknowledge funding from R01EB017449, R01EB013427, and P41EB013598.

References

[1] Chen H-Y et al., “3D Dynamic Hyperpolarized ¹³C-Pyruvate MR Metabolic Imaging of Human Prostate Cancer”, ISMRM. 2016.

[2] Larson PE, et al., “Investigation of tumor hyperpolarized [1-¹³C]-pyruvate dynamics using time-resolved multiband RF excitation echo-planar MRSI,” MRM. 2010; 63(3):582-591.

[3] Nelson SJ et al., “Metabolic imaging of patients with prostate cancer using hyperpolarized [1-¹³C]pyruvate,” Sci Transl Med. 2013;5:198ra108

[4] Bahrami N et al., “Kinetic and perfusion modeling of hyperpolarized ¹³C pyruvate and urea in cancer with arbitrary RF flip angles,” Quant Imaging Med Surg. 2014; 4(1): 24-32.

[5] Nagashima K et al., “Optimum pulse flip angles for multi-scan acquisition of hyperpolarized NMR and MRI,” JMR. 2008; 190: 183-188.

[6] Gordon JW et al., “Mis-estimation and bias of hyperpolarized apparent diffusion coefficient measurements due to slice profile effects,” MRM. 2016

Figures

Figure1.One way of adjusting the B₁ sensitivity profile is changing the effective T₁ used to generate the lactate flip angles(T_1,design)⁵.A) Longer T_1,design reduces the early lactate flip angles, and Δk_PL resulting from the range of error ΔB₁⁺ was depicted in this simulation.B) Deviating from the “optimal” maximum SNR flip angle design parameters comes at the cost of reduced total lactate SNR. This simulation shows impact on lactate tSNR(A.U.) of different T₁,_design. While the original T_1,design=30(s) generates maximum signal, our choice of T_1,design=100(s) significantly reduces B₁-sensitivitiy at the cost of 2.5% lactate tSNR loss. Δk_PL was shown for 20% B₁⁺ error.

Figure 2. Simulated signal curves and variation of k_PL estimates A) between our original and the proposed B₁-insensitive flip angle schemes subjected to different ΔB₁⁺. The impact of ΔB₁⁺ on amplitude of lactate signal is more benign using the new scheme. B) A threshold was set to limit the maximum flip angle for improved slice profile. k_PL Error was simulated for ΔB₁⁺ = -20%. The threshold does not have a significant impact on k_PL estimation in terms of ΔB₁⁺.

Figure 3. HP-¹³C imaging of a healthy Sprague Dawley rat using our original and the new B₁-insensitive flip angle scheme. The injections were ~15 minutes apart. A) Estimated k_PL of rat kidney, which is relatively homeostatic during the course of study. When subjected to variation ΔB₁⁺ = -20%, the new scheme shows substantially less k_PL drift than the original one. (old- Δk_PL = 63%, new- Δk_PL = 22%). B) Overlaid k_PL map on bSSFP anatomical reference for the same data as in A) .

Figure4. The new flip angle design enables acquisition of strong pyruvate and lactate signal in this patient with biopsy-proven Gleason 4+5, 4+4 and 4+3 cancer involving a majority of the left lobe of the prostate and extending into the right peripheral zone.A) 3D dynamic CS-EPSI allows spatial coverage of the entire prostate, from base to apex for HP-¹³C biomarkers (AUC through time). High lactate signal correlates with location of tumor.B) Spectrum through time for a selected voxel in tumor region. Pyruvate signal appears at around 10secs, lactate was observed later at 25secs, consistent with previous findings. At 41sec, SNR_pyruvate=68.5, SNR_lactate=44.4.

Figure 5. A) Temporal dynamics of HP-¹³C pyruvate and lactate in a slice that contains prostate cancer. B) Signal from clinical prostate tumor and peripheral vasculature. The new sequence not only maintains good pyruvate and lactate SNR, but preserves the timing characteristics of our previous design. Moreover, it potentially enables access to more information by the slightly longer (42s) acquisition window. C) Estimated pyruvate to lactate conversion rate k_PL map corresponded with a large volume of biopsy-proven Gleason 4+5, 4+4 and 4+3 cancer involving a majority of the left lobe of the prostate and extending into the right peripheral zone.

Proc. Intl. Soc. Mag. Reson. Med. 25 (2017)

0725