Accurate MR Thermometry by Hyperpolarized 129Xe
Le Zhang1,2, Alex Burant2,3, Andrew McCallister2,3, Karl Koshlap4, Simone Degan5, Michael Antonacci2,3, and Rosa Tamara Branca2,3

1Department of Applied Physical Sciences, University of North Carolina at Chapel Hill, Chapel Hill, NC, United States, 2Biomedical Research Imaging Center, University of North Carolina at Chapel Hill, Chapel Hill, NC, United States, 3Department of Physics and Astronomy, University of North Carolina at Chapel Hill, Chapel Hill, NC, United States, 4Eshelman School of Pharmacy, University of North Carolina at Chapel Hill, Chapel Hill, NC, United States, 5Center for Molecular and Biomolecular Imaging, Duke University, Durham, NC, United States

Synopsis

A new thermometry method based on the temperature dependence of lipid-dissolved 129Xe was proposed, while its accuracy was assessed by direct comparison with Proton Resonance Frequency (PRF) based MR thermometry methods. The temperature dependences of chemical shifts of lipid-dissolved 129Xe, water and methylene spins were first measured in vitro with high accuracy on various fat-rich tissues. The results were then used to obtain relative temperature maps in vivo in mice acclimated at different temperatures. Lipid-dissolved 129Xe based MR thermometry demonstrated superior accuracy in both in vivo and in vitro results when compared to PRF based MR thermometry in fatty tissues.

Purpose

Most MRI thermometry methods take advantage of the linear temperature dependence of the Proton Resonance Frequency (PRF)1 to measure temperature in vivo. However, the effect of temperature on PRF can be easily overshadowed by field drift as well as macroscopic and microscopic magnetic susceptibility gradients2. In fatty tissues, to partially remove the effect of field drift and macroscopic field inhomogeneities, the temperature-independent resonance frequency of methylene protons is often used as an internal reference3. However, as water and methylene spins reside in different tissue compartments, microscopic susceptibility gradients cannot be completely removed and can still produce large inaccuracies4. Recently, a strong linear temperature dependence of the chemical shift of Lipid-Dissolved 129Xe (LDX) was found in mice5. Being highly lipophilic, LDX spins reside in the same tissue compartment as methylene spins. As a result, methylene-referenced LDX based MR thermometry should provide more accurate temperature measurements. In this study, the accuracy of referenced and unreferenced PRF and LDX was evaluated both in vitro and in vivo.

Methods

In vitro high-resolution NMR experiments were performed on fatty tissues excised from rats and human cadavers on a Varian Inova 500MHz NMR Spectrometer by using a broadband probe. Samples were degassed and filled with isotopically enriched 129Xe (>85%) at desired pressures. A Varian biopack temperature control unit, calibrated using a 100% methanol sample, ensured a temperature accuracy of 0.1°C. For in vivo experiments, four obese (ob/ob) mice were scanned on a Bruker BioSpec 9.4T scanner using a xenon surface coil (1cm diameter) inserted in a 1H volume coil. The animals were anesthetized and mechanically ventilated with 75-vol% hyperpolarized (polarization ~15%, Polarean, Inc., Durham, NC) 129Xe (natural isotopic abundance) and 25-vol% oxygen. Body temperatures were allowed to equilibrate at set bore temperatures for 0.5h before 129Xe and 1H CSI maps were acquired. From these maps, 129Xe, water and methylene resonance frequencies were obtained, while temperature maps were derived by using a temperature coefficient of ‑0.01ppm/°C and -0.209ppm/°C for water and LDX, respectively.

Results

Fig. 1 shows the chemical shift temperature dependence of both unreferenced and methylene-referenced PRF and LDX. The temperature coefficient of PRF (1A) averaged to ‑0.012ppm/°C, in agreement with literatures1, with a maximal variation of 0.002ppm/°C across four samples. The temperature coefficient of LDX (1C) averaged to ‑0.209ppm/°C, with a variation of 0.05ppm/°C across the samples. When referenced to the chemical shift of the methylene peak, the average temperature coefficient of PRF method (1B) became ‑0.008ppm/°C, with a variation of 0.04ppm/°C, while that of LDX (1D) became ‑0.212ppm/°C, with a variation of 0.01ppm/°C. For a given temperature (25°C), the water-methylene resonance frequency difference showed a variation of 0.3ppm among the samples analyzed, whereas the LDX-methylene resonance frequency difference showed a variation of 0.4ppm. Fig. 2 shows the referenced and unreferenced PRF and LDX frequency distribution (2A) across the abdomen of an obese mouse acclimated at two different temperatures as well as the calculated relative temperature change (2B) obtained from these maps by using a temperature coefficient of ‑0.01ppm/°C for PRF and a temperature coefficient of ‑0.209ppm/°C for LDX. The relative temperature distribution at 29°C calculated by using the LDX-methylene resonance frequency difference is also shown in Fig. 2C. Fig. 3 shows the absolute value of the LDX-methylene resonance frequency difference obtained in vitro and in vivo for all animals and rWAT samples analyzed.

Discussion

For a given temperature, the water chemical shift and its temperature coefficient displayed a large variation across the samples, even when the methylene spins were used as reference. On the other end, the LDX temperature coefficient was close to what we previously found in vivo5 and with a small difference among all samples. In addition, the use of methylene spins as an internal reference for the LDX further reduced this difference to 0.01ppm/°C. The superior accuracy of LDX-based thermometry was also evident in the in vivo studies. The temperature distribution obtained using LDX frequency as a temperature probe was quite homogenous, with a variation of less than 1°C across the entire mouse abdomen. On the other hand, PRF-based MR thermometry yielded a temperature variation across the same area of more than 10°C. More interestingly, Fig. 3 shows the methylene-referenced LDX frequency could be used to accurately predict absolute temperature with an accuracy of better than 1°C in all four mice.

Conclusion

By utilizing the strong linear temperature dependence of lipid-dissolved 129Xe chemical shift and referencing it to the chemical shift of neighboring lipid spins, a novel MR thermometry method, more accurate and precise than PRF, is proposed.

Acknowledgements

No acknowledgement found.

References

[1] Venkatesh AK, Kacher DF, Kuroda K, et al. Temperature Measurement Using the 129Xe Chemical Shift. Proc ISMRM, Glasgow, Scotland. 2001 [2] Bolan PJ, Henry P-G, Baker EH, et al. Measurement and correction of respiration-induced B0 variations in breast 1H MRS at 4 Tesla. MRM 2004;52(6):1239-1245 [3] Hofstetter LW, Yeo DTB, Dixon WT, et al. Fat-referenced MR thermometry in the breast and prostate using IDEAL. JMRI 2012;36(3):722-732 [4] Szczepaniak LS, Dobbins RL, Stein DT, et al. Bulk magnetic susceptibility effects on the assessment of intra- and extramyocellular lipids in vivo. MRM 2002;47(3):607-610 [5] Branca RT, He T, Zhang L, et al. Detection of brown adipose tissue and thermogenic activity in mice by hyperpolarized xenon MRI. PNAS 2014;111(50):18001-18006

Figures

Figure 1: Methylene-referenced and unreferenced water (A&B) and LDX (C&D) chemical shift as function of temperature for different adipose tissue samples: rat white fat (rWAT), a mixture of rat white fat and rat muscle tissue (rWAT/muscle), human white fat (hWAT).

Figure 2: (A) Water, Water-methylene, LDX and LDX-methylene frequency maps obtained from the same animal acclimated at two different temperatures. (B) Calculated relative temperature change maps obtained by using the water, water-methylene, LDX and LDX-Methylene resonance frequency (C) Relative temperature distribution map at 29°C calculated from LDX-methylene chemical shift difference.

Figure 3: Absolute LDX-methylene chemical shift difference as a function of temperature obtained in all animals and samples analyzed. The value of the LDX-methylene resonance frequency difference was very similar across all animals and samples analyzed, despite experiments being performed on different days, different animals and at different field strengths.



Proc. Intl. Soc. Mag. Reson. Med. 24 (2016)
0551