3701

A Referenceless Workflow for Hyperpolarized 13C EPI
Jiazheng Wang1, Alan Wright1, Richard Hesketh1, De-en Hu1, and Kevin M Brindle1,2

1Cancer Research UK Cambridge Institute, University of Cambridge, Cambridge, United Kingdom, 2Department of Biochemistry, University of Cambridge, Cambridge, United Kingdom

Synopsis

We have developed a workflow for hyperpolarized 13C EPI phase correction that requires no reference scan. The workflow provides ghost-free images on phantoms with large or tight fields of view and where there are multiple signal sources. Dynamic images acquired from hyperpolarized [1-13C]pyruvate and [1-13C]lactate in a tumor showed comparable image quality to those corrected using a separate 13C reference scan.

Purpose

Development of a referenceless workflow for hyperpolarized 13C EPI.

Introduction

The development of DNP has enabled MRI studies of tissue metabolism, using hyperpolarized 13C-labelled substrates, at relatively high spatial and temporal resolutions 1. The transient nature of the hyperpolarization requires fast imaging. While EPI is an appealing technique because it can acquire a 2D image after a single excitation, it suffers from Nyquist ghosting. To correct for this a 13C reference scan can be acquired, but this takes up one frame of the dynamic data set. Alternatively a reference scan can be acquired using the 1H signal 2, but this can be suboptimal for 13C phase correction. Here we describe a workflow for hyperpolarized 13C EPI that renders reference scans unnecessary.

Methods

The workflow is illustrated in Figure 1, and is demonstrated using imaging of hyperpolarized [1-13C]pyruvate and [1-13C]lactate in vivo. An exhaustive search of the phase error correction coefficients is performed on the 13C EPI image to minimize signal in the ghost-containing background, which is readily obtainable from the proton image of tissue anatomy. As shown in Figure 2, a tumor ROI is selected from the proton image and mapped onto the [1-13C]lactate echo planar image, while regions outside the ROI in the phase encoding direction are treated as ghosting-containing background. The search is performed on the [1-13C]lactate image with the highest SNR and the correction coefficients obtained applied to all other images and for both metabolites.

Experiments

Experiments were performed on a 7T animal scanner (Agilent). Gradient-echo echo planar images were acquired from phantoms (diameters 7 mm and 18 mm, filled with 8.5M and 4M [1-13C]lactate respectively at thermal equilibrium) with a 4 cm FOV in order to examine performance of the method under conditions where there was a regular FOV, a tight FOV, and in the situation where there were multifocal signals. Image results from the workflow with 1D and 2D searches for the phase correction coefficients were compared with those corrected using a separate 13C reference scan and using 13C reference echoes 3. The workflow was also compared with the 1H reference scan method in the case of multifocal-signals, with a well-shimmed magnetic field and with a background gradient of 6mT/m2 in the XY direction. Imaging in vivo was conducted using the same parameters on female C57BL/6J EL4 lymphoma tumor-bearing mice following injection of hyperpolarized [1-13C]pyruvate, with alternating SpSp pulses on the [1-13C]pyruvate and [1-13C]lactate resonances and with a time resolution of 2s per metabolite.

Results

The proposed workflow resulted in phantom images of similar quality to those obtained using a separate 13C reference scan and using integrated 13C reference echoes (Figures 3A). The workflow also provided similar ghost-free images in vivo compared to those corrected with 13C reference scan (Figure 4). Quantitative measurements of residual ghosting in both phantom and in vivo images are shown in Figure 5A&5C. The workflow also showed an SNR benefit over the integrated echo method because of the prolonged TE with the latter method. The SNR was 1.7%, 3.0%, and 12.6% higher than the integrated echo method for a regular FOV, tight FOV, and with multifocal-signals respectively (Figure 5B). As shown in Figure 3B, the workflow gave better results than the 1H reference scan method (residual ghosting 1.3% versus 4.4%) under well-shimmed condition and this difference was increased in a de-shimmed condition (residual ghosting 1.6% versus 13.7%). The in vivo images obtained using the 1D search method and images corrected using 13C reference data were almost identical (SSIM≈1) through the whole dynamic imaging window for both [1-13C]pyruvate and [1-13C]lactate images (Figures 5D&5E).

Discussion and Conclusion

The referenceless workflow resulted in ghost-free hyperpolarized 13C echo planar images of phantoms and in vivo. The resultant image quality was comparable to that obtained using a 13C reference scan or integrated 13C reference echoes and was consistently better than that obtained using 1H reference data, especially under suboptimal conditions. It also showed an SNR benefit over the integrated 13C reference echo method, and this benefit is expected to be enhanced in vivo because of the shorter T2*. The 1D search showed higher efficiency than the 2D search with similar ghost-removal capability and much reduced processing time. The workflow is compatible with 3D EPI and parallel imaging, and is readily translatable to clinical applications given the sparse nature of hyperpolarized 13C metabolite data, which will nearly always provide a ghost-containing background.

Acknowledgements

No acknowledgement found.

References

[1] Brindle KM. Imaging metabolism with hyperpolarized 13C labeled cell substrates. J Am Chem Soc 2015; 137:6418-6427

[2] Gordon, JW, Vigneron DB, Larson PEZ. Development of a symmetric echo planar imaging framework for clinical translation of rapid dynamic hyperpolarized 13C imaging. Magn Reson Med 2016; doi: 10.1002/mrm.26123

[3] Miller JJ, Lau AZ, Teh I, Schneider JE, Kinchesh P, Smart S, Ball V, Sibson NR, Tyler DJ. Robust and high resolution hyperpolarized metabolic imaging of the rat heart at 7 T with 3D spectral-spatial EPI. Magn Reson Med 2016; 75: 1515-1524

Figures

Figure 1: Flow diagram outlining the proposed workflow for hyperpolarized 13C dynamic MRI with a 2D echo-planar trajectory. The phase correction coefficients α and β are searched for from the image set with the highest SNR. The individual slice phase correction coefficients are then applied to the corresponding slices in the other image sets in the dynamic acquisition.

Figure 2: Selection of ghost-containing image areas, where the signal should be minimized using the proposed workflow. The tumor region is drawn from the proton image (A), and the resulting ROI is applied to the [1-13C]lactate image (B). Bands falling outside of the ROI in the phase encoding direction (shaded bands in B) are considered as ghost-containing background areas.

Figure 3: 13C images of [1-13C]lactate-containing phantoms at thermal equilibrium, phase corrected using the proposed work flow (1D and 2D searches) and the indicated methods (13C reference scan, 13C reference echoes, 1H reference scan). (A) Images acquired using a regular FOV, a tight FOV, and from two separate phantoms, and phase corrected using the indicated methods. (B) Images acquired from two separate phantoms with a well shimmed magnetic field and in the presence of magnetic field inhomogeneity, where a 6 mT/m2 XY gradient was applied using the XY shim, and phase corrected using the indicated methods.

Figure 4: Representative images acquired following injection of hyperpolarized [1-13C]pyruvate into an EL4-tumor bearing mouse. Images (A) to (D) show a slice reconstructed without phase correction (A), phase corrected with 13C reference scan data (B), corrected with a 1D search (C), and corrected with a 2D search (D). Panels (E) and (F) show dynamically acquired [1-13C]pyruvate (E) and [1-13C]lactate images (F), which were corrected using the phase error coefficients obtained from (C). Only data acquired from the first 25 seconds is shown. Note that images acquired at 23 and 24 seconds were used as reference scans (for the 13C reference scan method) and are not shown.

Figure 5: A) Residual ghost and B) SNR were measured for each phase correction method in three different imaging conditions. C) Residual ghosting in in vivo datasets. The similarities between images corrected using the 13C reference scan method and images corrected using a 1D search were measured for both the [1-13C]pyruvate (D) and [1-13C]lactate (E) images taken from a representative in vivo dataset. The SSIM assessment suggests that the images corrected using the proposed workflow were very similar to those corrected using a reference scan (SSIM indices ≈ 1).

Proc. Intl. Soc. Mag. Reson. Med. 25 (2017)
3701