Model predictive filtering MR thermometry utilizing ultrasound beam modeling SAR predictions
Henrik Odéen1, Scott Almquist2, Joshua de Bever1, and Dennis L Parker1

1Utah Center for Advanced Imaging Research, Department of Radiology, University of Utah, Salt Lake City, UT, United States, 2School of Computing, University of Utah, Salt Lake City, UT, United States

Synopsis

Thermal model based reconstruction of subsampled MR temperature data for focused ultrasound applications rely on acoustic and thermal parameters that are often analytically determined from a pre-treatment sonication. In this work we combine a thermal model based reconstruction method with ultrasound beam simulations to determine the specific absorption rate in order to avoid potentially damaging the tissue during a pre-treatment sonication. Proof-of-concept experiments are performed in a homogenous gelatin phantom and a gelatin phantom embedded with a plastic skull. The temperature estimations using US modeling show the same accuracy as those using a pre-treatment sonication.

Introduction/Purpose

MRI has become the imaging modality of choice for monitoring focused ultrasound treatments due to its ability to combine good soft tissue contrast with temperature measurements obtained using MR temperature imaging (MRTI). One current challenge in MRTI is being able to combine high spatio-temporal resolution imaging, to ensure accurate temperature measurements, with large field-of-view, to be able to monitor any unwanted US power deposition occurring in the US near- and far-field. To achieve increased acquisition speed, which in turn can be traded for higher resolution and larger field-of-view, various compressed-sensing and model based reconstruction methods of subsampled k-space data have been described.

In this work we combine a previously described thermal-model-based reconstruction method for subsampled MRTI data called model predictive filtering (MPF)1 with US beam modeling using the hybrid angular spectrum (HAS) method2. The thermal and acoustic parameters used in MPF have previously been derived using an analytical model fit3,4 to a pre-treatment, low power sonication. Utilizing US modeling to derive the specific absorption rate (SAR) can avoid the need for this potentially tissue damaging sonication.

Methods

In MPF, temperature maps for time-frame (n+1) are calculated by combining subsampled k-space data with a forward-prediction of the temperature distribution from time-frame (n) using the Pennes Bioheat Transfer equation. After combination and projection into image space, temperatures are calculated from the image phase using the proton resonance frequency shift method.

In this proof-of-concept study, experiments were performed in a homogenous gelatin phantom5 as well as in a gelatin phantom with an embedded homogenous PVC plastic skull, Figure 1. For the skull phantom experiment, US phase aberration correction was simultaneously performed using HAS6. MRTI reconstructions were performed with MPF using HAS with no pretreatment heating and using analytically derived SAR and thermal conductivity from a low power pre-treatment heating. The analytical solution uses tabular values for density and specific heat, whereas MPF with HAS uses tabular values for all tissue properties. MPF reconstructions were compared to a temporally constrained reconstruction (TCR) method7 as well as to fully sampled “truth”.

All MR imaging was performed using a 3D segmented EPI pulse sequence on a 3T MRI scanner (TIM Trio, Siemens, Erlangen, Germany) with imaging parameters summarized in Table 1. k-space was subsampled by varying the sampling density in the kz-slice encode direction to achieve denser sampling over the k-space center, while keeping the spacing between the echoes down the echo train equal for all echoes8. FUS heatings were performed with an MR-compatible phased-array US transducer (256 elements, 1 MHz frequency, 13 cm radius of curvature, 2x2x8 mm focal spot FWHM, Imasonic (Besancon, France) and Image Guided Therapy (Bordeaux, France)). In the homogenous phantom FUS sonications were performed at 40 W for 28.7 s (subsampling and “truth”) and at 14 W for 28.7 s (for analytical property determination), and in the skull phantom at 125 W for 23.9 s (subsampling and “truth”) and at 100 W for 14.3 s (analytical property determination). The high-power sonications were repeated three times.

Results

Figure 2 shows the temperature evolution of the hottest focal spot voxel versus time for the FUS sonications in the two phantoms, as well as the error compared to “truth”. The mean and standard deviation (SD) of the root mean square error for MPF HAS, MPF analytical, and TCR in the homogenous phantom was 0.61±0.02, 0.78±0.02, and 0.54±0.21 °C, respectively, and in the skull phantom 0.45±0.09, 0.40±0.01 and 0.32±0.07 °C, respectively. The spatial temperature distribution for “truth”, and the spatial error of both MPF and the TCR reconstructions, are shown in Figure 3. In Figure 4 three orthogonal views of the estimated SAR patterns are shown. In the skull phantom the large FOV temperature maps detected an approximately 8 °C temperature rise in the near-field on the “brain” surface, and an 8% increase of max temperature rise when using the aberration correction (data not shown).

Discussion and Conclusions

The results show that it is possible to utilize US modeling to predict the SAR pattern in both homogenous phantoms and when focusing through a plastic skull, and in turn use the modeled SAR pattern in MPF to achieve temperature maps with accuracy similar to MPF using analytically derived SAR and TCR. The analytical SAR estimate was higher and wider than the HAS estimate, but the accuracy in the MPF temperature maps were similar, highlighting the robustness of MPF in general. By utilizing US modeling in conjunction with MPF the need for a potentially tissue damaging pre-treatment sonication can be avoided.

Acknowledgements

This work was supported by The Focused Ultrasound Surgery Foundation, Siemens Healthcare, The Ben B. and Iris M. Margolis Foundation, and NIH grants R01s EB013433 and CA134599

References

1. Todd N, Payne A, Parker DL. Model predictive filtering for improved temporal resolution in MRI temperature imaging. Magn. Reson. Med. 2010;63:1269–79.

2. Vyas U, Christensen DA. Ultrasound beam propagation using the hybrid angular spectrum method. In: IEEE Engineering in Medicine and Biology Society. ; 2008. pp. 2526–2529.

3. Dillon CR, Vyas U, Payne A, Christensen DA, Roemer RB. An analytical solution for improved HIFU SAR estimation. Phys. Med. Biol. 2012;57:4527–44.

4. Dillon CR, Payne A, Christensen DA, Roemer RB. The accuracy and precision of two non-invasive, magnetic resonance-guided focused ultrasound-based thermal diffusivity estimation methods. Int. J. Hyperth. 2014;30:362–71.

5. Farrer AI, Odéen H, de Bever J, Coats B, Parker DL, Payne A, Christensen D A. Characterization and evaluation of tissue-mimicking gelatin phantoms for use with MRgFUS. J. Ther. Ultrasound 2015;3:9.

6. Almquist S, DeBever J, Merrill R, Parker D, Christensen D. A Full-Wave Phase Aberration Correction Method for Transcranial High-Intensity Focused Ultrasound Brain Therapies. IEEE Eng Med Biol Soc. ; 2014. pp. 310–313.

7. Todd N, Adluru G, Payne A, DiBella EVR, Parker D. Temporally constrained reconstruction applied to MRI temperature data. Magn. Reson. Med. 2009;62:406–19.

8. Odéen H, Todd N, Diakite M, Minalga E, Payne A, Parker DL. Sampling strategies for subsampled segmented EPI PRF thermometry in MR guided high intensity focused ultrasound. Med. Phys. 2014;41:092301.

Figures

Figure 1. a) Homogeneous gelatin phantom setup and b) skull phantom setup. The US transducer was coupled with a bath of de-ionized and de-gassed water, and in-house built single-channel loop RF coils were used for signal detection.

Table 1. MR scan parameters. TR – Repetition time, TE – Echo time, FOV – Field-of-view, BW – Bandwidth (in readout direction), ETL – Echo train length, FA – Flip angle, tacq – acquisition time, R – k-space reduction/subsampling factor (R = 1 means fully sampled).

Figure 2. a) and b) show mean and SD of temperature rise of the hottest voxel for the homogenous and the skull phantom, respectively, for all reconstruction methods investigated. c) and d) shows temperature error compared to fully sampled “truth” for the homogenous and the skull phantom, respectively.

Figure 3. Three orthogonal views of the focal spot for a) the homogenous phantom and b) the skull phantom. The first row shows “truth” at the time of maximum temperature rise, and rows 2-4 shows the error compared to “truth” of MPF HAS, MPF analytical, and TCR.

Figure 4. Three orthogonal views of SAR from a)-b) homogenous phantom using HAS and analytical solution, and c)-d) skull phantom using HAS and analytical solution, respectively. The analytical solution estimated 25% and 38% higher maximum SAR than HAS, but also 18% and 29% lower thermal conductivity, for the homogenous and skull phantom, respectively.



Proc. Intl. Soc. Mag. Reson. Med. 24 (2016)
2111