4741

Improved MR Temperature Imaging at 0.5T with Multi-echo Thermometry

Ziyi Pan¹, Jianxiong Hu², Hai Luo², Simin Liu¹, Sisi Li¹, Ziyue Wu², and Hua Guo¹
¹Center for Biomedical Imaging Research, Department of Biomedical Engineering, School of Medicine, Tsinghua University, Beijing, China, ²Marvel Stone Healthcare Co., Ltd., Wuxi, China

Synopsis

Low field MR-guided thermotherapy can provide some key advantages over the high-field alternative, including reduced cost, decreased susceptibility artifacts, and improved safety of interventional devices. However, both the accuracy and the speed of PRF temperature measurement suffer at the low field due to the reduced SNR, limited receive channels, and declined temperature-induced phase changes, making it unreliable for clinical MRgLITT treatments. In this study, we demonstrate that the multi-echo thermometry together with the view-sharing acceleration can be utilized to achieve high-quality PRF thermometry at 0.5T with satisfactory temperature measurement precision and temporal resolution.

Introduction

MR-guided laser interstitial thermal therapy (MRgLITT) has been applied as a minimally invasive treatment in neurosurgery. Recent research¹ has shown the great potential of low field MR systems (such as 0.55T) on MR-guided interventions and thermotherapies, benefiting from reduced cost, declined interventional device heating and decreased susceptibility artifacts. However, the precision of the proton resonance frequency (PRF) shift-based thermometry² can be deteriorated at low-field due to low SNR^3,4, hindering its clinical reliability for LITT temperature monitoring. Additionally, sampling acceleration is needed in MRgLITT to ensure enough temporal resolution, yet maintaining SNR in the meantime has to be considered.

This study aims to improve the temperature precision at low-field using multi-echo GRE and to accelerate the acquisition using a view-sharing approach^5-9 without sacrificing SNR. The performance of this method was validated on a 0.5T scanner.

Theory and Methods

The influence of B₀ Decrease on PRF Thermometry
PRF thermometry is challenging at 0.5T due to the decrease of the magnetic field B₀.
- The accuracy of temperature measurement drops dramatically due to the B₀ decrease. The uncertainty of temperature measurement at 0.5T is 20 times higher than that at 1.5T according to (1)³：
$$\delta T=\sqrt{2}/(2\pi\cdot\gamma\cdot B_0\cdot\alpha\cdot TE\cdot SNR),\ \ SNR\propto B_0^{7/4}\ \ \ (1)$$
- The sensitivity of temperature measurement is declined at 0.5T because temperature-induced phase difference is inversely proportional to B₀ according to (2)² :
$$\triangle \phi = \alpha\triangle T\cdot\gamma B_0\cdot TE\ \ \ (2)$$
Multi-echo Thermometry
Longer TE can provide better temperature precision and sensitivity according to Eq. (1) and (2) in GRE-based thermometry. Therefore, multi-echo thermometry^10,11 with its ability to use a wide range of TEs without increasing scan time is well suited at low fields. A bipolar GRE sequence¹² was applied to achieve higher SNR efficiency.

- Echo Misregistration Correction: Since the receiver bandwidth is kept low to ensure sufficient SNR in this study, field inhomogeneity and eddy current-induced misregistration between even and odd echoes is corrected¹³ with an estimated B₀ field map from multi-echo data.

- Echo Combination: The measured temperature maps are combined into a single estimate using a tSNR optimal weighted echo combination approach¹⁴, whereby the weights are the production of image magnitude and TE.

View-sharing Acceleration
The view-sharing-based approach, with its ability to achieve fast reconstruction, and to preserve accelerated image SNR without the need for multiple receive channels, is beneficial for the low field. So this study adopted the view-sharing acceleration strategy and investigated its feasibility.

Experiments
All experiments were performed in a 0.5T MRI scanner (Marvelstone, Wuxi, China) with an eight-channel receive coil. Temperature data were acquired by a multi-echo bipolar GRE with the following parameters: flip angle = 30°, TE = 7.5~40.5ms, nTE / delta TE = 7 / 5.5ms, TR = 138ms, matrix = 108×110, FOV = 220×220 mm², slice thickness = 5mm, 3 slices (no gap), sampling bandwidth (BW) = 27.8 kHz, view-sharing acceleration = 3 (variable density), temporal resolution = 5 sec/volume.

- Simulation Experiments: Fully sampled data were first acquired on pork tissues using the sequence mentioned above but without acceleration (temporal resolution=12 sec/volume, heat at 4W for 120s). Two different k-t undersampling patterns (Fig.1A) were tested. A sliding window-based algorithm⁵, which recovered undersampled k-space using previous time frames, was implemented for reconstruction. The reconstructed temperature maps were compared to those reconstructed using GRAPPA.

- In Vitro Tissue Heating Experiments: In vitro MRgLITT experiments (MR-Guided Laser Ablation System, Sinovation Medical, Beijing, China) were carried out in pork tenderloin samples (N=3, heat at 6W for the 60s, wait for 60s, then heat for another 60s) and pig brain tissues (N=2, heat at 4W for 50s) separately. An MR-compatible fiber optical sensor was inserted into the tissue to obtain ground truth. Temperature uncertainty was calculated as the standard deviation (SD) of the temperature difference between MR and fiber-optic measurements.

Results and Discussion

Multi-echo Thermometry
As illustrated in Fig. 2, the image misregistration between odd and even echoes can lead to wrong temperature measurements (e.g. reduced temperatures at the heating center). Therefore, echo misregistration correction is necessary for bipolar multi-echo thermometry, especially when using low readout bandwidth (greater SNR and also greater echo mismatch).

Fig. 3 demonstrates that the precision of temperature measurement can be substantially improved after echo combination when compared with single-echo measurements. Therefore, multi-echo thermometry is highly recommended for low-field PRF measurement.

View-sharing Acceleration
Simulation experiments show finer temperature maps with lower errors in the 3-fold view-sharing acquisition (VarDensity, Fig. 1B). GRAPPA reconstructed images suffer from SNR decrease which can lead to increased temperature uncertainty. View-sharing acceleration (Interleaved) fails at rapid temperature rise (Fig. 1B, first row) probably because the reconstruction uses too much temperature data from previous time frames.

Temperature Measurement Accuracy
Multi-echo thermometry with view-sharing acceleration (3-fold, VarDensity) results in high-quality temperature maps (Fig. 4) and excellent precision with the average temperature uncertainty of less than 0.8°C (Tab. 1) in in vitro heating experiments.

Conclusion

We have shown experimentally that multi-echo thermometry with view-sharing acceleration can achieve high-quality PRF thermometry at 0.5T. The measured temperature uncertainty is within a clinically acceptable threshold (<3°C). Still, it needs further investigation on whether the view-sharing acceleration is acceptable from a clinical perspective, as the temperatures may change more complicatedly during treatment.

Acknowledgements

No acknowledgement found.

References

1. Campbell-Washburn AE, Ramasawmy R, Restivo MC, et al. Opportunities in Interventional and Diagnostic Imaging by Using High-Performance Low-Field-Strength MRI. Radiology 2019;293(2):384-393.

2. Ishihara Y, Calderon A, Watanabe H, et al. A precise and fast temperature mapping using water proton chemical shift. Magn Reson Med 1995;34(6):814-823.

3. Botnar RM, Steiner P, Dubno B, Erhart P, von Schulthess GK, Debatin JF. Temperature quantification using the proton frequency shift technique: In vitro and in vivo validation in an open 0.5 tesla interventional MR scanner during RF ablation. Journal of Magnetic Resonance Imaging 2001;13(3):437-444.

4. Waqas Majeed AJK, Sunil Patil, Henrik Odéen, John Roberts, Florian Maier, Dennis L. Parker, and Himanshu Bhat. Feasibility of Magnetic Resonance Thermometry at 0.55T. Proc Intl Soc Mag Reson Med 29 (ISMRM & SMRT Virtual Conference & Exhibition), 2021. p.

5. Riederer SJ, Tasciyan T, Farzaneh F, Lee JN, Wright RC, Herfkens RJ. MR fluoroscopy: technical feasibility. Magn Reson Med 1988;8(1):1-15.

6. Jones RA, Haraldseth O, Muller TB, Rinck PA, Oksendal AN. K-space substitution: a novel dynamic imaging technique. Magn Reson Med 1993;29(6):830-834.

7. van Vaals JJ, Brummer ME, Dixon WT, et al. "Keyhole" method for accelerating imaging of contrast agent uptake. J Magn Reson Imaging 1993;3(4):671-675.

8. Doyle M, Walsh EG, Blackwell GG, Pohost GM. Block regional interpolation scheme for k-space (BRISK): a rapid cardiac imaging technique. Magn Reson Med 1995;33(2):163-170.

9. Han Y, Mun C. Evaluation of the keyhole technique applied to the proton resonance frequency method for magnetic resonance temperature imaging. J Magn Reson Imaging 2011;34(5):1231-1239. 10. Marx M, Pauly KB. Improved MRI Thermometry with Multiple-Echo Spirals. Magn Reson Med 2016;76(3):747-756.

11. Svedin BT, Payne A, Bolster BD, Parker DL. Multiecho pseudo-golden angle stack of stars thermometry with high spatial and temporal resolution using k-space weighted image contrast. Magn Reson Med 2018;79(3):1407-1419.

12. Wieben O LJ, Mansson S, Hennig J. Multi-echo balanced SSFP imaging for iterative Dixon reconstruction. Proceedings of the 13th Annual Meeting of ISMRM Miami Beach, 2005. p.

13. Lu W, Yu H, Shimakawa A, Alley M, Reeder SB, Hargreaves BA. Water-fat separation with bipolar multiecho sequences. Magn Reson Med 2008;60(1):198-209.

14. Madore B, Panych LP, Mei CS, Yuan J, Chu RX. Multipathway Sequences for MR Thermometry. Magn Reson Med 2011;66(3):658-668.

Figures

FIG. 1. (A) Two different undersampling patterns for 3-fold view-sharing acceleration. For variable density undersampling, data at the k-space center is fully sampled (like keyhole) while the rest is sparsely undersampled. (B) 3 representative sets of temperature maps of the simulation experiments calculated from fully sampled, GRAPPA reconstructed and view-sharing accelerated data. Red arrowheads indicate rough edges of the heating zone due to reduced SNR. Yellow arrowheads indicate false low temperatures caused by undersampling.

FIG. 2. (A) Heating curves of Voxel 1 at the heating center (solid lines) and Voxel 2 at the heating edge (dotted lines) with (in red) and without (in black) echo misregistration correction, respectively. (B) From the first column to the last column are the zoomed-in GRE magnitude images of TE₅, TE6, their interlay (TE₅ in green and TE₆ in magenta), and corresponding multi-echo combined temperature maps. Red lines (solid in the first column and dashed in the second) indicate the position of the optic fiber in TE₅, while blue lines are marked for TE₆.

FIG. 3. (A) Representative temperature maps measured from each echo individually and from the tSNR optimal weighted echo combination. Calculated SD_fiber between the MR and fiber-optic measured temperatures is indicated in the upper left corner of each subgraph. (B) Corresponding PRF heating curves of TE₁=7.5ms (first row), TE₇=40.5ms (second row), and the multi-echo combination (ME Combined, last row), respectively (shown in black line). Note that only temperatures at the cooling stages are measured by the optic fiber (shown in red dots) due to band interference.

FIG. 4. Representative zoomed-in temperature images obtained in MRgLITT treatments in pork tenderloin (first row) and pig brain (second row) tissues, separately. The temperature maps overlayed on the T2 weighted images demonstrate good quality.

Tab. 1. Calculated temperature accuracy across subjects in in-vitro tissue heating experiments. MR vs. Optic fiber (Error Range) and (± SD) are the range and the SD of the temperature difference between MR and fiber-optic measurements, respectively. Phase noise (± SD) is the SD of the temperature measurements in a nonheated area.

Proc. Intl. Soc. Mag. Reson. Med. 30 (2022)

4741

DOI: https://doi.org/10.58530/2022/4741