Misung Han1, Wenwen Jiang2, Roland Krug1, Peder Larson1,2, and Viola Rieke1
1Radiology and Biomedical Imaging, University of California, San Francisco, San Francisco, CA, United States, 2Joint Graduate Program in Bioengineering, University of California, San Francisco/Berkeley, San Francisco, CA, United States
Synopsis
High-intensity
focused ultrasound (HIFU) is a promising, noninvasive technique to ablate bone
tumors and palliate painful bone metastases. During HIFU treatment, temperature
mapping is desirable for proper heat deposition to targeted bone regions. Even
though conventional PRF-based thermometry cannot be applied for cortical bone
due to its short T2 relaxation time, it was demonstrated using 3D UTE imaging can
be used to measure T1 changes due to heating. In this work, we accelerated 3D UTE
imaging by combining parallel imaging and compressed sensing and compared calculated
T1 changes due to heating with those from fully sampled data.Purpose
High-intensity
focused ultrasound (HIFU) is a promising, noninvasive technique to ablate bone
tumors and palliate painful bone metastases.
1,2 During HIFU
treatment, temperature mapping is desirable for proper heat deposition to targeted
bone regions. However, conventional PRF-based thermometry cannot be applied for
cortical bone due to its short T
2 relaxation time. Recently, a linear increase
in cortical bone T
1 with temperature was demonstrated using 3D UTE
imaging combined with a variable flip angle method.
3
However, this previously used UTE T
1 mapping method with 3D radial
acquisitions requires a long scan time. In this work, we evaluated acceleration
of this technique by combining parallel imaging and compressed sensing to
depict T
1 changes due to heating with ultrasound.
Methods
An ex vivo study
was performed with a diaphysis segment of bovine femur on a Discovery MR 750w
3T scanner (GE Healthcare, Waukesha, WI) using an eight-channel phased-array wrist coil (Invivo,
Gainesville, FL). A 7.7 MHz catheter-cooled interstitial ultrasound applicator with
two-sectored cylindrical transducers4 was inserted into the fatty
yellow bone marrow. Heating of cortical bone was conducted by delivering high
intensity ultrasound energy with a 180° directional heating pattern towards the
bone. Before heating began, 3D UTE imaging combining non-selective excitation
and 3D radial acquisitions (no undersampling) was conducted using a 11 ms TR, 175
μs TE (center of the RF to the start of acquisition), 1 x 1 x 1.5 mm3
spatial resolution, 9 x 9 x 9.6 cm3 FOV, and RF spoiling. Data
acquisition started with the rising slope of the readout gradient (ramp
sampling), and 256 samples were acquired for each spoke with a sampling
interval of 4 μs. Imaging with two flip angles of 8° and 44° (each with 4 min
scan time) was conducted to calculate T1 using a variable flip angle scheme.5
Temperature was
assumed to reach steady state at 12 min after heating began, and UTE imaging
with the two flip angles was performed again.
The acquired data was
retrospectively undersampled by a factor of 8. Reconstruction was performed
using an autocalibrated parallel imaging method, ESPIRiT, which can incorporate
compressed sensing as well.6 The central, fully-sampled, k-space data was
used for coil calibration. The l1
norm of the wavelet coefficients was added as a sparsity regularization7
during iterative reconstruction of the images (l1-ESPIRiT). T1 maps were calculated using reconstructed
images from fully-sampled data by 3D gridding and those from undersampled data by 3D gridding and l1-ESPiRIT, before and after heating, respectively. The l1-ESPiRIT and gridding reconstructions were performed with Berkeley Advanced Reconstruction Toolbox.8
Results
Figure 1a-c shows reconstructed UTE images applying the 44° flip angle before
heating from full k-space data and from undersampled data. We can see
significant reduction of streak artifacts and noise by
l1-ESPIRiT reconstruction on undersampled data. Figure 2 demonstrate Maps of T
1 changes between before and after heating, measured using the three different sets of reconstructed UTE
images. T
1
increase can be seen in the heated region of cortical bone.
L1-ESPIRiT reconstruction of
the undersampled data provided T1 values similar to those from
fully-sampled data.
Discussion
Acceleration
techniques such as parallel imaging and compressed sensing are well suited for 3D radial acquisitions due to their autocalibration capability from a
central over-sampled k-space data and incoherent aliasing artifact patterns in
the image domain. Combination with compressed sensing can allow for a higher
acceleration factor than just using parallel imaging itself due to its
denoising effects. In this work, we accelerated UTE imaging by a
factor of 8 by using
l1-ESPIRiT
reconstruction, which provided T
1 quantification similar to that from fully-sampled data. Further
acceleration might be achievable by incorporating similarity between two flip
angle images or those over the temporal domain. These acceleration methods might
allow for applying 3D T
1 mapping in cortical bone to monitor temperature during bone HIFU treatment.
Conclusion
We
have shown that
l1-ESPIRiT can accelerate 3D UTE imaging with a high
acceleration factor without significant artifacts, and allows for measuring T
1 changes due to heating similar to those from fully-sampled data.
Acknowledgements
Supported by NIH R00HL097030 and GE Healthcare.References
1. Catane
R, Beck A, Inbar Y, et al. MR-guided focused ultrasound surgery (MRgFUS) for
the palliation of pain in patients with bone metastases- preliminary clinical
experience. Ann Oncol 2007;18:163–167.
2. Li
C, Zhang W, Fan W, et al. Noninvasive treatment of malignant bone tumors using
high-intensity focused ultrasound. Cancer 2010;116:3934–3942.
3. Han M, Rieke V,
Scott SJ, et al. Quantifying temperature-dependent T1 changes in
cortical bone using ultrashort echo-time MRI. Magn Reson Med 2015 (early view).
4. Diederich
CJ, Nau WH, Stauffer PR. Ultrasound applicators for interstitial thermal
coagulation. IEEE Trans Ultrason Ferroelectr Freq Control 1999;46:1218–1228.
5. Fram EK, Herfkens RJ, Johnson GA, et al. Rapid calculation of T1
using variable flip angle gradient refocused imaging. Magn Reson Imaging 1987;5:201–208.
6. Uecker M, Lai P,
Murphy, et al. ESPIRiT—an eigenvalue approach to autocalibrating parallel MRI:
Where SENSE meets GRAPPA. Magn Reson Med 2014;71: 990–1001.
7. Lustig M, Donoho
D, Pauly JM. Sparse MRI: The application of compressed sensing for rapid MR
imaging. Magn Reson Med 2007;58:1182–1195.
8. BART: 10.5281/zenodo.31907.