Magnetic Resonance Thermometry Targeting for Magnetic Resonance–Guided Histotripsy Treatments

Abstract

Objective:

The potential of transcranial magnetic resonance (MR)–guided histotripsy for brain applications has been described in prior in vivo studies in the swine brain through an excised human skull. The safety and accuracy of transcranial MR-guided histotripsy (tcMRgHt) rely on pre-treatment targeting guidance. In the work described here, we investigated the feasibility and accuracy of using ultrasound-induced low-temperature heating and MR thermometry for histotripsy pre-treatment targeting in ex vivo bovine brain.

Methods:

A 15-element, 750-kHz MRI-compatible ultrasound transducer with modified drivers that can deliver both low-temperature heating and histotripsy acoustic pulses was used to treat seven bovine brain samples. The samples were first heated to an approximately 1.6°C temperature increase at the focus, and MR thermometry was used to localize the target. Once the targeting was confirmed, a histotripsy lesion was generated at the focus and visualized on post-histotripsy MR images.

Discussion:

The accuracy of MR thermometry targeting was evaluated with the mean/standard deviation of the difference between the locus of peak heating identified by MR thermometry and the center of mass of the post-treatment histotripsy lesion, which was 0.59/0.31 mm and 1.31/0.93 mm in the transverse and longitudinal directions, respectively.

Conclusion:

This study determined that MR thermometry could provide reliable pre-treatment targeting for transcranial MR-guided histotripsy treatment.

Introduction

Transcranial magnetic resonance–guided focused ultrasound (tcMRgFUS) uses external ultrasound energy focused through the skull to create local heating for thermal tissue ablation. tcMRgFUS has been used clinically to treat essential tremor [1,2]. However, because of the potential concern for overheating of the skull, tcMRgFUS is limited to treatment of a small volume in the center of the brain [3,4]. By use of microsecond-length, high peak negative focal pressure (P⁻ < − 26 MPa) ultrasound pulses through the skull, transcranial histotripsy has been found to generate controlled and confined cavitation to mechanically destroy targeted brain tissue [5,6]. Histotripsy minimizes heating by using a very low duty cycle (<0.1%) while maintaining effective therapy. This allows transcranial histotripsy to potentially treat a wide range of locations and volumes in the brain [5,7].

Histotripsy typically uses ultrasound imaging for treatment targeting and monitoring, because histotripsy-generated cavitation is easily visualized by ultrasound imaging [8]. However, for transcranial MR-guided histotripsy (tcMRgHt) brain treatment, ultrasound imaging is not practical because of the presence of the skull. MRI guidance is desirable, as MRI is the gold standard for the diagnosis and post-therapy follow-up for many brain pathologies. Recently, our laboratory designed and built an MR-compatible transcranial histotripsy system [9]. This system was used to treat in vivo intact pig brain through an excised human skull [6]. However, in that study, a stereotactic approach was used with fiducial markers for targeting. An MR-based, non-invasive targeting method for transcranial histotripsy would be desirable.

Magnetic resonance–guided focused ultrasound heating therapies also use low-temperature heating and MR thermometry for pre-treatment targeting. Commercial MR-guided HIFU systems such as the Sonalleve V2 by Profound Medical (Mississagua, ON, Canada) for the treatment of osteoid osteoma [10] and uterine fibroids [11] and Exablate Body by Insightec (Tirat Carmel, Israel) for the treatment of uterine fibroids and bone metastasis use this method for pre-treatment targeting [12,13]. tcMRgFUS thermal therapy uses MR thermometry for pre-treatment targeting [14,15] and treatment monitoring [16–19]. For pre-treatment targeting, the focal region is heated by a few degrees Celsius, remaining below the damage threshold, and is visualized in real time using MR thermometry to confirm the location of the treatment region [14]. Low-temperature pre-treatment targeting is used in both the Sonalleve V2 system and the ExAblate System for transcranial treatments [3,20]. The low-temperature heating for the MR thermometry targeting approach works well for tcMRgFUS as both pre-treatment targeting and treatment are performed via ultrasound heating. Another alternative for tcMRgFUS pre-treatment targeting is MR acoustic radiation force imaging (MR-ARFI), which sensitizes contrast to the displacement induced by radiation force from the absorption of the ultrasound beam [21–25]. The displacement from ultrasound radiation force is very small and transient (e.g., 1 μm over <1 s), but recent studies have acquired high-spatiotemporal resolution MR-ARFI images [26,27] in tissue phantoms. Hardware requirements for MR-ARFI are similar to those for MR thermometry.

In this study, we investigate the ability of MR thermometry to predict focal locations generated from histotripsy treatments. Even though MR thermometry for tcMRgFUS is well established, there are technical challenges that need to be addressed when it is applied to histotripsy. First, ultrasound heating arises from the absorption of acoustic pulses, whereas histotripsy is a threshold phenomenon and occurs by exceeding a pressure threshold. The peaks of the acoustic field for heating and histotripsy may not coincide and need to be assessed. Second, histotripsy transducers and electrical drivers are designed to generate microsecond-length pulses at very high pressure and low duty cycle (<1%), minimizing heating. In this work, a modified electrical driver was used to produce both low-temperature heating and histotripsy.

To test the feasibility and accuracy of using MR thermometry for pre-treatment targeting of histotripsy, we designed and constructed a 15-element MR-compatible histotripsy array (Fig. 1) with a modified electrical driver capable of both histotripsy and low-temperature heating, and quantified the accuracy of MR thermometry for histotripsy pre-treatment targeting. This system was designed using an f-number and frequency similar to those for a transcranial array designed for human experiments [9]. Because the 15-element array is not powerful enough to produce cavitation through the skull, the testing was done in the ex vivo bovine brain tissue without a skull. Even in the absence of the skull, any discrepancy between the thermometry focal location and histotripsy lesion location can still be evaluated.

Figure 1.

Left: T2-weighted image of the experimental setup inside the magnetic resonance scanner. Right: Cross section image of the array. The transducer was constructed so that the midpoint drawn from the two fiducials estimates the geometric focus of the array.

Methods

Histotripsy transducer and drivers

All experiments were performed using a custom-built, 15-channel, 750-kHz hemispherical array (Fig. 1) with a 50-mm focal radius and an f-number of 1. The array was designed to be a watertight container while keeping a small footprint, to ensure a small field of view (FOV) on MR images. The array focal zone has a full-width half maximum (FWHM) of 3 mm in the longitudinal plane and 1.2 mm in the transverse plane in free field. The same driving electronics were used to produce both histotripsy and heating acoustic waveforms.

Experimental setup

Seven ex vivo bovine brain tissues were harvested from a nearby slaughterhouse and stored under refrigeration in 1% benzalkonium chloride solution prior to use (within 1 wk). Immediately prior to the experiments, the tissues were de-gassed in a vacuum chamber and set in 1.5% agarose gel to ensure that there was no bulk displacement throughout the duration of the experiments (Fig. 1). To produce low-temperature heating for pre-treatment targeting, 6-μs pulses (Fig. 2) at relatively low pressure (peak negative pressure of 4 MPa) and a high duty cycle (30% duty cycle, equivalent of 50-kHz pulse repetition frequency [PRF]) were used. Given the low-pressure and high-focal-gain sound field, no significant non-linear propagation effects were expected. The tissues were heated for 15 s while simultaneous MR thermometry images were acquired.

Figure 2.

Free-field pressure waveform from one transducer element at the array focus.

To test the precision of MR thermometry in estimating the histotripsy focus, we created the smallest histotripsy lesions that were readily visible on MRI. Histotripsy was delivered using 6-μs pulses at high pressure (estimated peak negative pressure of 54 MPa) and a low duty cycle (0.3%, equivalent of 50-Hz PRF). The array focus was electronically steered to a 3- × 3- × 2-mm grid with 100 pulses per focal location and 0.5- × 0.5- × 1-mm spacing between adjacent focal locations.

MR thermometry pre-treatment targeting

All MR images were acquired on a GE MR750 3T scanner (GE Healthcare, Waukesha, WI, USA) using a 32-channel head array (Nova Medical, Wilmington MA, USA). MR thermometry was performed using a gradient-recalled echo (GRE) scan with a spiral acquisition. The slices were centered at the focus estimated using the fiducials placed on the array (Fig. 1). MR thermometry scan parameters were as follows: flip angle 60°, TE/TR 30 ms/1000 ms, 13 cm FOV, 10 slices, 2 mm slice thickness, 128 × 128 matrix size and 4 shots. Thus, the temporal resolution was 4 s, and the spatial resolution was 1 mm for the imaging plane and 2 mm in the slice direction. Conjugate phase reconstruction was used to reconstruct the images with reduced off-resonance distortions [28]. Fourier interpolation was applied to resample the images to a 256 × 256 matrix size.

Drift correction was applied to the phase-difference images by calculating the mean phase difference from a region outside of the heating zone and subtracting it from the entire image. The precision of temperature measurement from MR thermometry was estimated by measuring the temperature change in the tissues with no heating applied.

To estimate the targeting location from the heating map on MR-thermometry images, a region of interest (ROI) that contained the heated region was selected and interpolated to a 4 × finer grid. Then a 2-D Gaussian surface was fitted to the temperature values. The peak of the fitted Gaussian was estimated as the heating focus (Fig. 3). To make the Gaussian fit robust to noise, moving averages of three heating map images were taken before estimating the fit. To reduce the effect of thermal diffusion altering the location of the estimated target, the fit was calculated from the earliest image during heating where a temperature change was easily observable. This was done by measuring the signal-to-noise ratio (SNR) of the heated region and using the earliest time point within 90% of the maximum SNR to fit the Gaussian.

Figure 3.

Time series display of magnetic resonance thermometry for one slice at every other time point, indicating change in temperature (in °C). Heating starts at 28 s and ends around 40 s. With longer heating times, diffusion effects spread out the heating zone.

Estimation of array geometric focus

A set of fiducials were placed on the histotripsy array housing, such that the midpoint estimated from the two fiducials approximates the geometric focus of the array (Fig. 1). Both MR and ultrasound factors will contribute to errors in the estimated array focus. As the histotripsy lesion was produced at the geometric focus, the difference between the focal location measured by the fiducial markers and the location of the lesion generated by histotripsy on the MR image was measured to assess the focal position shift caused by non-linear ultrasound propagation with the high-pressure histotripsy pulses and acoustic aberration. As the MR thermometry–estimated focus is assumed to overlap with the fiducial geometric focus, the difference between the MR thermometry targeting location and the geometric focus location estimated from the fiducial markers on MR images was measured to evaluate the inaccuracy caused mainly by MR-related factors (e.g., non-uniform gradient distribution across the field of view, noise in Gaussian fit).

Targeting accuracy evaluation

To measure the accuracy of MR thermometry pre-treatment targeting, the target location estimated from MR thermometry was compared with the central location of the histotripsy lesion measured from post-treatment MR scans. Pre- and post-treatment T2-weighted, T2*-weighted and diffusion-weighted imaging (DWI) scans were used to visualize the histotripsy lesion. The imaging parameters for T2*-weighted images were as follows: spoiled GRE with TE/TR of 10 ms/300 ms, FOV 13 cm, slice thickness 2 mm, matrix size 192 × 192 and flip angle 42. T2-weighted images were acquired using Fast Spin Echo with a TE/TR of 66 ms/3 s, FOV 13 cm, slice thickness 2 mm, matrix size 192 × 192 and echo train length 19. DWI (b value = 0, 1000 s/mm²) were acquired using a spiral acquisition with a TE/TR of 60 ms/2 s, FOV 13 cm, slice thickness 2 mm, matrix size 128 × 128, 4 shots and 40 averages. Spiral acquisition parameters and the reconstruction pipeline were kept the same for the DWI and MR thermometry scans to ensure that no bias caused by image acquisition and reconstruction processes is seen in the estimated focus, although eddy currents caused by diffusion gradients could lead to additional distortions that are not accounted for in this method. As we cannot ensure the lesion to be fully contained within a voxel (partial-volume effect), it was expected that the lesion would be visualized in a different number of voxels for each tissue. To estimate the lesion location from post-treatment MR images, all the voxels with a hypo-intense region were identified on DWI images. A center of mass was evaluated from these voxels and designated as the lesion center. To measure the accuracy of targeting, the focal location estimated from fiducials and MR thermometry targeting location were compared with the lesion center estimated from post-treatment MR by calculating the difference in the location in longitudinal and transverse axes.

Results

MR thermometry pre-treatment targeting

The tissues were heated to a ~1.6°C increase (Fig. 3) in temperature, as measured by MR thermometry, using the parameters described under Methods. The baseline standard deviation of temperature change without heating was ~0.2°C. In Figure 4 is an example of the MR thermometry image measuring the heating generated from the modified histotripsy transducer and driver. An ROI is also marked indicating the area that was selected to fit a 2-D Gaussian over the heating image. An example targeting location estimated by MR thermometry for this tissue is also provided.

Figure 4.

Magnetic resonance thermometry heating image overlaid on b = 0 s/mm² diffusion-weighted image. Left: Raw heating image with a region of interest that is used to fit the Gaussian. Right: Fitted Gaussian along with the estimated magnetic resonance thermometry focus.

Post-treatment MRI of histotripsy

In Figure 5 are MR images of a tissue before and after histotripsy treatment. Histotripsy lesions were visualized in all tissues. Lesions were not visible on T2*-weighted scans for any tissue. T2-weighted scans were able to visualize the lesion in most tissues, although the contrast of the lesion was low. Therefore, the post-treatment T2-weighted scan was not used for targeting accuracy evaluation. DW images depicted the lesion as a hypo-intense spot [29] that was easily visible in all tissue samples and was used for determining the histotripsy focal location. Figure 6 provides an example of how a histotripsy lesion was estimated on post-treatment DWI scans.

Figure 5.

T2*-weighted, T2-weighted and diffusion-weighted (b = 1000 s/mm²) images before and after histotripsy treatments. The lesion contrast is best visualized on the diffusion-weighted image (DWI) because of homogenization of cellular matter in that region, causing higher diffusion in the lesion compared with the surrounding tissues.

Figure 6.

Visualization of the histotripsy lesion along with estimated focus from magnetic resonance thermometry and fiducials.

Targeting accuracy evaluation

The targeting accuracy was evaluated by measuring the mean and standard deviation of the absolute difference in the MR thermometry targeting location and the center of mass of the histotripsy lesion measured using the post-treatment DWI scan (Table 1). The absolute mean/standard deviation difference of the estimated lesion location using MR thermometry was 0.59/0.31 mm in the transverse and 1.31/0.93 mm in the longitudinal planes, respectively.

Table 1.

Error in focus estimation

Tissue	Difference between fiducial and lesion locations (mm)		Difference between MR thermometry and lesion locations (mm)		Difference between MR thermometry and fiducial locations (mm)
	Transverse	Longitudinal	Transverse	Longitudinal	Transverse	Longitudinal
1	0	1.3	0.8	1.8	0.7	0.5
2	−0.4	1.7	−0.3	2.3	0.2	0.6
3	−0.2	0.9	−0.5	1.1	−0.3	0.1
4	0	0.4	0.6	−2.3	0.6	−2.6
5	0.3	0.4	0.1	0.1	−0.1	0.2
6	−0.5	0.8	0.9	0.1	1.4	−0.6
7	0.2	1.2	0.9	1.5	0.6	0.4
Mean absolute error	0.23	0.96	0.59	1.31	0.56	0.71
SD absolute error	0.19	0.48	0.31	0.93	0.44	0.85

The minus sign is toward the array in the longitudinal axis and to the left in the transverse axis.

MR, magnetic resonance; SD, standard deviation.

To assess the focal position shift caused by non-linear ultrasound propagation histotripsy pulses and acoustic aberration, the difference between the geometric focal location measured by the fiducial markers and the actual location of the histotripsy lesion on the post-treatment MR scan was measured (Table 1). The absolute mean/standard deviation difference between the geometric location measured with the fiducials and the central location of the histotripsy lesion measured using the post MR scan was 0.23/0.19 mm in the transverse and 0.96/0.48 mm in the longitudinal planes, respectively.

The difference between the MR thermometry targeting location and the focal location estimated from the fiducial markers was measured to evaluate the inaccuracy caused mainly by MR-related factors (Table 1). The absolute mean/standard deviation difference between the MR thermometry targeting location and the geometric focus location estimated from the fiducial markers was 0.56/0.44 mm in the transverse and 0.71/0.85 mm in the longitudinal planes, respectively.

Discussion

This work illustrates that MR thermometry–based targeting can successfully predict the location of histotripsy lesions with sub-millimeter precision. We investigated the accuracy of using MR thermometry for histotripsy targeting by comparing the peak location of a Gaussian fitted to the MR thermometry map and the center of mass in post-treatment DWI of histotripsy lesions.

This approach has several limitations. First, using different MRI pulse sequences for MR thermometry and DWI means that issues such as susceptibility and motion could cause varying image distortions for different pulse sequences. This could cause shifts in estimated focus that could be accounted for by performing image registration or by carefully matching acquisition and reconstruction methods. A calibration phantom can be used to evaluate distortions caused by the different pulse sequences. Second, using a Gaussian fit to determine the target location from the thermometry map can be confounded by heat absorption, diffusion, flow and the ROI selected to estimate the Gaussian. This error can be reduced by using the earliest heating image with high SNR before the heating profile broadens. We found that the fiducial marker approach can be used to estimate the ultrasound transducer geometric focus precisely, but its role may be limited for transcranial ultrasound therapy where significant aberrations from the skull might cause more pronounced focal shifts. Additionally, larger imaging FOV with more B0 non-uniformity and gradient non-linearity across the imaging plane could reduce MR fiducial accuracy. However, there remain other issues with the use of ultrasound heating for histotripsy targeting. First, heating is affected by thermal diffusion and flow, thus spreading the heating region with time and biasing the targeting location estimate. Second, ultrasound propagation for low-temperature heating is in the linear regime, and thus, the effect of aberrations on the focal shift may be different in the case of non-linear propagation in histotripsy, although this difference may be small, as transcranial histotripsy transducers have a very low f-number and ultrasound beams do not overlap significantly until they get close the focal regions. Third, for tcMRgFUS experiments, it has been observed that getting focal heating is more difficult closer to the skull. Thus, using MR thermometry for pre-treatment targeting in shallow regions might be difficult.

In this study, we found that in the case of limited aberration, MR thermometry can be effective for pre-treatment targeting for histotripsy. In future studies, we plan to modify the driving electronics to the 360-element, 700-kHz hemispherical transcranial array designed for human use, which will be able to perform aberration correction through the skull and electrical focal steering. We will also explore the effects of skull aberration on the accuracy of this method. The precision of using MR thermometry targeting for histotripsy with electrical steering also needs to be investigated. The larger transcranial array allows for steering and more pressure output than the array used in this study. Another way of performing pre-treatment targeting is by using MR-ARFI to measure acoustic pressure directly, which may provide a better targeting accuracy than that of MR-thermometry, as it takes the non-linearity into account and will be investigated in the future. The modifications to our drivers for performing both heating and histotripsy also allow them to perform radiation force sonications. Another benefit of using MR-ARFI is that it can be used for aberration correction prior to histotripsy sonications [22].

Conclusions

The safety and accuracy of tcMRgHt rely on pre-treatment targeting guidance. In this study, by using driving electronics capable of performing low-temperature heating and histotripsy, we determined the feasibility of using MR thermometry to provide sub-millimeter accuracy for pre-treatment histotripsy targeting. In future studies, we will implement the modified driving electronics to the human-scale, 360-element hemispherical transcranial histotripsy array and explore the effect of skull aberration as well as electrical focal steering on the performance and accuracy of MR thermometry for pre-treatment histotripsy targeting.

Data availability statement

To access the data used in this study, please e-mail the authors at dinankg@umich.edu.