2950

Chemical Shift Encoded (CSE) Image Reconstruction for Spectral Selection in Fluorine-19 MRI
Kai D. Ludwig1, Diego Hernando1,2, Nathan T. Roberts2,3, Ruud B. van Heeswijk4, and Sean B. Fain1,2,5

1Medical Physics, University of Wisconsin - Madison, Madison, WI, United States, 2Radiology, University of Wisconsin - Madison, Madison, WI, United States, 3Electrical and Computer Engineering, University of Wisconsin - Madison, Madison, WI, United States, 4Radiology, Lausanne University Hospital (CHUV) and University of Lausanne (UNIL), Lausanne, Switzerland, 5Biomedical Engineering, University of Wisconsin - Madison, Madison, WI, United States

Synopsis

In preclinical applications, the high specificity of quantitative 19F MRI may be compromised by non-negligible signal contributions from fluorinated anesthetics (e.g. isoflurane). Here, we demonstrate the feasibility of chemical shift encoding (CSE) with multi-resonance fluorine signal modeling and least-squares estimation image reconstruction for 19F MRI. We optimize noise performance (NSA) and use a 3D spoiled gradient-echo acquisition to separate signal contributions from perfluoro-15-crown-5-ether (PFCE) and isoflurane. The method is tested in mixed PFCE/isoflurane phantoms showing effective signal separation. The CSE reconstruction removes isoflurane signal contributions in 19F MR images of PFCE in vivo, potentially reducing errors in 19F concentration quantification.

Introduction

Fluorine-19 (19F) MRI is a highly specific technique for quantitative cellular and molecular detection. Fluorinated anesthetics (e.g. isoflurane) are commonly used in preclinical applications, but introduce substantial background signal1 complicating the interpretation of 19F images. For these reasons, removal of isoflurane signal from 19F images is highly desirable. Importantly, chemical shift encoded (CSE) image reconstruction2 may effectively isolate spectral components of contaminating signal from unique 19F resonances. Therefore, the purpose of this work is to demonstrate the feasibility of CSE in 19F MRI using a 3D spoiled gradient-echo (SPGR) acquisition and to optimize noise performance for separation of signal contributions from isoflurane and perfluoro-15-crown-5-ether (PFCE), a common 19F contrast agent.

Methods

Proposed CSE technique: A multi-TE CSE acquisition is proposed. The model used for CSE reconstruction (Equation 1) was solved using a non-linear least-squares estimation to determine the signal contributions ($$$ s $$$) at each k-space position ($$$ k_x,k_y,k_z $$$) from isoflurane ($$$ ρ_I $$$) and PFCE ($$$ ρ_P $$$).

$$ s_{TE} (k_x,k_y (t),k_z,t;ρ_P,ρ_I,f_B )=∭_{x,y,z}e^{i2πf_B (x,y,z)(TE+t)} e^{i2πxk_x} e^{i2πyk_y (t)} e^{i2πzk_z} [ρ_I (x,y,z) ∑_{m=1}^M(α_m e^{i2πf_m (TE+t)} ) +ρ_P (x,y,z) e^{i2πf_P (TE+t)} ]dx dy dz $$

The resonant frequencies of the PFCE ($$$ f_P $$$) and $$$ m $$$ isoflurane peaks ($$$ f_m $$$) with relative signal amplitude $$$ α_m $$$, the frequency shifts ($$$ f_B $$$) due to local B0 field inhomogeneity and phase evolution during the imaging readout ($$$ k_y(t) $$$, where $$$ t $$$ is the time relative to the $$$ TE $$$ during the readout) are included in the model.

Noise performance optimization: Cramér-Rao Lower Bound (CRLB) analysis3 was performed in Matlab 2014b (MathWorks, Natick, MA) to determine the effective number of signal averages (NSA) at different combinations of initial echo time (TEinit), echo spacing (ΔTE), and number of echoes. NSA was normalized to the number of echoes.

Phantom experiment: All MR data were acquired on a 4.7T preclinical MRI system (Agilent Technologies, Santa Clara, CA). The spectral separation between PFCE and isoflurane was measured in phantoms of PFCE (Exfluoro, Round Rock, Tx) and isoflurane (Piramal, Bethlehem, PA) using a FID collected with a 90° global excitation, rBW=10kHz, and NSA=1. We demonstrated the feasibility of the technique in vitro with mixed phantoms of PFCE and isoflurane prepared in 0.5mL tubes, and MR imaged with a 3D SPGR repeated with multiple echo times (TR=10.0ms, TEinit=2.3ms, ΔTE=0.3ms, 12 TEs, 1.0mm3 isotopic resolution, rBW=50kHz, NSA=1) using a home-built 19F quadrature volumetric RF coil. CSE image reconstruction was performed in Matlab.

In vivo experiment: To demonstrate in vivo feasibility of CSE image reconstruction, MR data was collected on one healthy male C57BL/6 mouse anesthetized with 1.5% isoflurane and maintained at 37°C with a temperature probe and hot-air blower. This pilot study complied with institutional animal care and use committee regulations. The 19F MRI data was acquired with a 3D SPGR using the optimized TE and ΔTE from the CRLB analysis and phantom experiments. (TR=200.0ms, TEinit=2.3ms, ΔTE=0.3ms, 6 TEs, 1.6x1.6x4.0mm3 resolution, rBW=18kHz, NSA=8) prior to and after intraperitoneal injection of 45mM of a PFCE emulsion. The PFCE emulsion was synthesized as a kinetically stable, oil-in-water nanoemulsion loaded with PFCE. Anatomic 1H data was acquired with a T1-weighted 3D SPGR (TR=4.35ms, TE=2.19ms, and 0.31mm2 in-plane resolution.

Results

The relative chemical shifts between PFCE and isoflurane's CF3 and CHF2 spectral groups were 1.8kHz (9.6ppm) and 0.45kHz (2.4ppm) at 4.7T, respectively (Figure 1). CRLB analysis plots indicated optimal data acquisition parameters (TEinit, and ΔTE, and number of echoes) needed to maximize NSA for CSE reconstruction of PFCE and isoflurane (Figure 2). CSE reconstructed 19F images showed spectral separation of PFCE and isoflurane in mixed phantoms (Figure 3). The in vivo experiments demonstrated that fluorinated gas contamination can be removed from 19F MR images (Figure 4).

Discussion

We have demonstrated that a chemical shift encoding approach using multi-spectral fluorine signal modeling with least-squares estimation can be used to remove fluorinated anesthetic signal contributions from 19F MR images, potentially reducing errors in 19F contrast agent concentration quantification. The large spectral separation between fluorine species causes significant phase accumulation during the acquisition trajectory, which is corrected using k-space modeling.4 Others have demonstrated the utility of 'multi-color' 19F MRI to image different fluorine contrast agents simultaneously using chemical shift imaging (CSI)5,6 or chemical shift selective7 techniques. The CSE methodology proposed here provides flexibility for applications that require spectral separation or selective quantification to detect multiple fluorine agents.

Conclusion

We have demonstrated the feasibility of a CSE approach for separation of PFCE and isoflurane in 19F MRI.

Acknowledgements

The authors thank our collaborators and colleagues. We thank Alexa Barres and Dr. Sandro Mecozzi for providing the PFCE emulsion for in vivo studies. We gratefully acknowledge NIH awards UL1TR000427 and TL1TR000429, UW School of Medicine and Public Health, UW Carbone Comprehensive Cancer Center, and GE Healthcare for funding support.

References

1Gaudet JM, Ribot EJ, Chen Y, Gilbert KM, Foster PJ. Tracking the fate of stem cell implants with fluorine-19 MRI. PLoS One. 2015;10(3):e0118544.

2Reeder SB, McKenzie CA, Pineda AR, Yu H, Shimakawa A, Brau AC, et al. Water-fat separation with IDEAL gradient-echo imaging. J Magn Reson Imaging. 2007;25(3):644-52.

3Pineda AR, Reeder SB, Wen Z, Pelc NJ. Cramer-Rao bounds for three-point decomposition of water and fat. Magn Reson Med. 2005;54(3):625-35.

4Brodsky EK, Holmes JH, Yu H, Reeder SB. Generalized k-space decomposition with chemical shift correction for non-Cartesian water-fat imaging. Magn Reson Med. 2008;59(5):1151-64.

5Temme S, Jacoby C, Owenier C, Grapentin C, Wang X, Schubert R, et al. Assessment of Thrombus Stage by 'Multicolor' 19F MRI. Proc Int Soc Magn Reson Med. 2016; 24.

6Basse-Lüsebrink TC, G. Ladewig, Kampf T, Melkus G, Haddad D, Bauer WR, et al. Multi-color 19F CSI: Simultaneous detection of differently labeled cells in vivo. Proc Int Soc Magn Reson Med. 2009;17.

7Jacoby C, Temme S, Mayenfels F, Benoit N, Krafft MP, Schubert R, et al. Probing different perfluorocarbons for in vivo inflammation imaging by 19F MRI: image reconstruction, biological half-lives and sensitivity. NMR Biomed. 2014;27(3):261-71.

Figures

Figure 1: The relative fluorine-19 (19F) chemical shifts of perfluoro-15-crown-5-ether (PFCE) and isoflurane at 4.7T. The molecular structures of isoflurane and PFCE are color-coded according to their respective resonances. Isoflurane has two resonance structures from the CF3 (red) and CHF2 (green) spectral groups while PFCE contains a singular peak (blue).

Figure 2: Noise performance for chemical shift encoded separation of PFCE and isoflurane is dependent on the echo time combination (number of echoes, TEinit, and ΔTE). In this work, 6 echoes were acquired with TEinit=2.3ms and ΔTE=0.33ms resulting in a normalized effective number of signal averages (NSA) for PFCE imaging of 0.99 (maximum normalized NSA is 1).

Figure 3: 19F MR images of mixed isoflurane (Iso) and PFCE phantoms reconstructed with either a conventional iFFT (A), listed here as a ‘source’ image, or a chemical shift encoded (CSE) image reconstruction (B,C). Ghosting from chemical shift artifact is noted in the source image (white arrow). The CSE reconstruction using a multi-resonance fluorine signal model with least-squares estimation demonstrates signal recovery in a separate PFCE-only image (B) and Iso-only image (C). Six echoes were used in the CSE reconstruction (TEinit=2.3ms and ΔTE=0.33ms). The PFCE:Iso volume ratio is noted on the source image.

Figure 4: In vivo feasibility of CSE image reconstruction of PFCE is demonstrated in a mouse model with removal of background isoflurane signal. Anatomical 1H MR images (gray scale) with overlaid 19F MR images (color scale) taken before (A) and after (B) intraperitoneal (i.p.) injection of 45mM of a PFCE emulsion, showing substantial isoflurane background signal in both source images pre- and post-PFCE injections. CSE image reconstruction allows for the spectral separation of signals to create PFCE-only and Iso-only images. Six echoes were used in the CSE recon (TEinit=2.3ms and ΔTE=0.33ms).

Proc. Intl. Soc. Mag. Reson. Med. 25 (2017)
2950