Perfluorocarbon emulsion enhances MR-ARFI displacement and temperature in vitro: Evaluating the response with MRI, NMR, and hydrophone

Sonosensitive perfluorocarbon F8TAC18-PFOB emulsion is under development to enhance heating, increase thermal contrast, and reduce treatment times during focused ultrasound tumor ablation of highly perfused tissue. The emulsion previously showed enhanced heating during ex vivo and in vitro studies. Experiments were designed to observe the response in additional scenarios by varying focused ultrasound conditions, emulsion concentrations, and surfactants. Most notably, changes in acoustic absorption were assessed with MR-ARFI. Phantoms were developed to have thermal, elastic, and relaxometry properties similar to those of ex vivo pig tissue. The phantoms were embedded with varying amounts of F8TAC18-PFOB emulsion or lecithin-PFOB emulsion, between about 0.0-0.3% v:w, in 0.05% v:w increments. MR-ARFI measurements were performed using a FLASH-ARFI-MRT sequence to obtain simultaneous displacement and temperature measurements. A Fabry-Perot hydrophone was utilized to observe the acoustic emissions. Susceptibility-weighted imaging and relaxometry mapping were performed to observe concentration-dependent effects. 19F diffusion-ordered spectroscopy NMR was used to measure the diffusion coefficient of perfluorocarbon droplets in a water emulsion. Increased displacement and temperature were observed with higher emulsion concentration. In semi-rigid MR-ARFI phantoms, a linear response was observed with low-duty cycle MR-ARFI sonications and a mono-exponential saturating response was observed with sustained sonications. The emulsifiers did not have a significant effect on acoustic absorption in semi-rigid gels. Stable cavitation might also contribute to enhanced heating.


Introduction
Magnetic resonance-guided focused ultrasound (MRgHIFU) is a noninvasive ablative procedure that allows reduced morbidity and faster recovery for certain medical treatments. Under MRI guidance, the operator can monitor the temperature change in each voxel to ensure that the cancer cells were sufficiently heated to induce apoptosis. This allows accurate determination of positioning, in addition to providing more certainty to a sufficient thermal dose. The temperature increase generated by the focused ultrasound beam reduces the amount of hydrogen bonding in the water molecules, reducing the resonance frequency of these atoms, resulting in a temperature-dependent shift in proton spectra (1)(2)(3), that generates a measurable change in the phase maps of MRI images (4,5). Commercial magnetic resonance thermometry (MRT) sequences allow monitoring of intraoperative temperature changes using this principle.
Alternative experimental MRI sequences are of interest to monitor HIFU procedures for other applications. Magnetic resonance acoustic radiation force impulse (MR-ARFI) imaging can measure the simultaneous increase in temperature and displacement from a focused ultrasound (FUS) pulse (6)(7)(8), with displacement increasing linearly with applied ultrasound power (8,9). The technique is used for applications, including distinguishing healthy tissue or diseased tissue, quantifying tissue stiffness, identifying ablation lesions, focal point tracking without heat damage, adjusting for phase aberrations, and focal spot shaping (10)(11)(12)(13)(14)(15). MR-ARFI has been implemented in pilot studies, such as for noninvasive breast palpation before and after focused ultrasound lesion ablation (10) and for transcranial displacement measurements (12). Simulation of the MR-ARFI displacement can be performed with finite-difference solutions to equations of motion (16)(17)(18) or by convolution techniques (15,19). Not least of all, the MR-ARFI sequence has potential for focal spot tracking during localized, targeted, low-temperature focused ultrasound-mediated sonoporation drug delivery. Perfluorooctyl bromide (PFOB) nanoemulsions have shown effective in vitro and in vivo for non-thermal and non-cavitational drug delivery (20,21). Clinical studies for focused ultrasound-mediated drug delivery in liver tumors with liposomal doxorubicin formulations (ThermoDox, Celsion Corp, Lawrenceville, New Jersey, USA) reported improved tumoral doxorubicin concentrations (22,23).
Focused ultrasound is indicated for ablation in a variety of tumor types. However, the procedure faces certain limitations, particularly for hepatobiliary treatments (24). Notable impediments in liver ablation include interactions with the thoracic cage and respiratory motion (25-28). Additional measures are often required, such as anesthesia, intubation, and rib resection (29,30). The left liver lobe is exposed for a portion of the respiratory cycle allowing ablation to be performed, although the right liver lobe is predominantly located behind the rib cage which complicates the treatment (31). Also, high-risk tumor sites are located near to adjacent organs, particularly near the gallbladder and bowels. Retrospective review has given common side-effects in hepatobiliary ultrasound-guided FUS procedures to be skin burns, pain, and fever at 15%, 5%, and 2%, respectively (29). Though, more adverse events have been reported, including biliary obstruction, fistula formation, osteonecrosis, and diaphragmatic rupture (30). A variety of techniques were developed to overcome some limitations of the procedure including rib shielding devices (32), a large acoustic window super-focusing transducer dedicated to intra-abdominal treatments (33), and techniques to account for respiratory motion (34)(35)(36)(37).
The use of exogenous cavitation nuclei is particularly interesting for focused ultrasound-mediated blood-brain barrier (BBB) opening to enhance drug delivery in brain tumors. Lowintensity MRgFUS in combination with microbubbles can increase BBB permeability, by temporarily and reversibly disrupting these pathways, allowing tumoral accumulation of higher amounts of standard chemotherapeutics and larger molecular weight chemotherapeutics that would otherwise be incapable of paracellular transport (38). Microbubbles typically exhibit a lower threshold for cavitation than droplets, as the droplets must first undergo vaporization, whether in the perfluorocarbon or in the surrounding medium, to promote further gas accumulation (39). Microbubble contrast agents also have a particular advantage in human studies in that these can be used off-label in neuro-oncological therapeutic treatments. This emulsion formula, at reduced droplet size, has shown capable as an in vivo drug delivery vehicle in combination with microbubbles for focused ultrasound-mediated BBB disruption (21).
In previous studies, perfluorocarbon nanoemulsions were developed for theranostic applications, both as a chemotherapy drug carrier and 19 F-MRI contrast agent (40). The formula was later adapted for ablative HIFU applications. Micron-sized F 8 TAC 18 -PFOB droplets were designed for intra-capillary transpulmonary circulation, illustrating repeatability and enhanced heating in a tissue mimicking material (TMM) phantoms with focused ultrasound (41). Subsequent experiments in ex vivo pig kidneys and liver showed an increased energy deposition by a factor of 3, for a 0.24% v:v solution of 1.42-2.16 ± 0.68 μm in the perfused fluid, at flow rates up to 0.15 mL.s -1 , capable of repeatable sonication without rupture or embolism (42). The acoustic absorption coefficient of a TMM phantom perfused with a solution of this perfluorocarbon emulsion was observed to scale monoexponentially with concentration to an asymptotic maximum, at doses expected safe for human use (43). The results gave insight into the size of the probability of temperature increase around an individual droplet, which was denoted as the interaction radius. This observation allowed development of a simulated model to predict an interaction radius of 12.5 μm for droplets with diameters of 1.9-2.3 μm.
In this study, experiments evaluated whether TMM gel phantoms exhibited an increase in the displacement of MR-ARFI due to perfluorocarbon droplets and whether cavitation was present. It was hypothesized that enhanced absorption could also increase displacement. Further measurements were performed with MRI, NMR, hydrophone, and rheometry to characterize the droplets and TMM phantoms. To the authors' knowledge, this is the first study using MR-ARFI to evaluate acoustic absorption changes from sonosensitive droplets. Additionally, a TMM gel phantom formula was further developed that was capable of use with microscale MR-ARFI displacement measurements and MR thermometry with highertemperature sustained sonications. Moreover, new details were gained on the acoustic absorption mechanism of this particular perfluorocarbon emulsion formula.
The F 8 TAC 18 surfactant consists of a water-soluble oligomer of Tris(hydroxymethyl)aminomethane (Tris), which constitutes the polar head of the surfactant, and a perfluorinated tail group (44,45), that allows tuning the droplet size by altering the number of dimethylene groups in the perfluorinated tail group. The F 8 TAC 18 surfactant and emulsions were prepared as described in previous publications (40 -43, 46). In brief, a low energy process was employed to prepare the emulsions using a homogenizer system (Polytrons PT 3100, Kinematica, Luzern, Switzerland). The general routine procedure consisted of dissolving 500 mg of emulsifier in 196 mL of water, after which 4 mL of PFOB was added. The solution was then cooled in an ice bath. The solution was homogenized three times for 15 minutes at 20,000 rpm for the lecithin emulsion. Similarly, the solution was homogenized three times for 15 minutes at 24,000 rpm for the F 8 TAC 18 emulsion. The emulsions were then refrigerated at 4°C until use.
The particle size distributions were determined by laser diffraction and Mie light scattering theory, using a Mastersizer 2000 with a Hydro2000S dispersion unit module (Malvern Instruments, Orsay, France). The dispersant solution consisted of water with a refractive index of 1.333, while the refractive index of PFOB was 1.305. The emulsions were added to the Hydro2000S dispersion unit while stirring at 18,550 rpm. Then the system formulated the volume-weighted mean diameter (D [4,3]). The volume fraction of PFOB in the solution was determined with 19 F nuclear magnetic resonance (NMR), as described in previous studies (46). The D [4,3] was 1.83 μm for the lecithin droplets studies. The D [4,3] ranged between 1.70-1.76 μm for the F 8 TAC 18 droplets, except one sample which was 1.24 μm.

Fabrication of TMM phantoms
The phantom formulas were made using modifications to previously described MR-ARFI phantoms (47,48). The TMM phantom formulas were adjusted to emulate ex vivo pig tissue properties, by comparing temperature and displacement measurements in ex vivo pig kidney, pig liver, and pig myocardium. An example of MR-ARFI temperature and displacement measurements in an ex vivo pig kidney is shown in Figure 1. The figure shows the displacement and temperature maps generated during MR-ARFI sonications overlaid on the MRI magnitude images in the location of the insonicated area. These values were used for comparison with subsequent iterations of TMM sample formulations, using equivalent sonication conditions. The figure also provided a good depiction of how the focal spot displacement and temperature profiles appear relative to mammalian anatomical features.
The proportions of starch, condensed milk, gelatin, and agarose were adjusted until the phantom provided similar MR-ARFI displacement and temperature response as seen in ex vivo pig organs. Agarose was later determined to make the phantoms too inelastic, and the substance was omitted from the formula. Some formulas, such as those incorporating agarose, tended to give good thermal response, but the elastic properties were too stiff, and displacement could not readily be measured. Alternatively, with other formulas, good elastic properties were obtained, but the phantoms would then melt under higher power sustained sonications. Two emulsifiers were tested: lecithin and F 8 TAC 18 . The final gel formula was 143 grams total, composed of 3% w:w gelatin, 6% w:w starch (maize flour), 5% w:w powdered milk (containing 36% w:w proteins, 0.8% w:w fats), and 0.021% w: w sodium azide as a preservative, between 0.05-0.3% v:w PFOB emulsion, and the remainder of approximately 86% v:v degassed and deionized water. Powdered milk at 10% w:w was used in certain gel samples, for powdered milk containing 18% w:w proteins. Also, benzalkonium chloride (BAL) was used as a preservative in lieu of sodium azide on occasion. Two control gels were made simultaneously, free of any PFOB emulsion. 19 F-NMR was used to verify the concentrations of selected gel sets. Phantoms used both emulsion formulas near 0.05% v:w, 0.1% v:w, 0.2% v:w, and 0.3% v:w. An example of quantitative 19 F-NMR measurements for PFOB concentration is available in the supplementary materials.
A protocol to fabricate the TMM gel sets with this final formulation can be found in the supplemental materials. Briefly, degassed and deionized water was heated to boiling, then the heat was removed. Then added to the hot water was condensed milk, starch, and gelatin. The solution was mixed homogeneously and cooled at room temperature to about 65°C before being mixed with the emulsion solution. The phantoms were cast into two plastic cups, one placed within the other. The bottom of each plastic cup was removed, and a thin piece of food cellophane was placed between the cups, to create a smooth, watertight surface. The warm liquid gel was poured into this container, allowed to cool at room temperature for about one hour, placed in an ice bath to solidify, then stored in the refrigerator at 4°C. An MR image of a TMM gel sample is shown in Figure 2. The size varied slightly on occasion, as some samples were fabricated at different facilities, with different inventory of equipment. However, the important aspect was for the size of the TMM gel phantom to be large enough, relative to the size of the focal spot, to negate effects of ultrasound reflections at the  Illustration of TMM gel phantom placed on top of the MRgHIFU transducer system. This MR image is a transverse view generated from a T1-VIBE sequence. The inset shows an enlarged view of the TMM phantom. Also visible in the image is the transducer and coupling components.
TMM surfaces from influencing the measurements. Ultrasound coupling gel was further placed on top of the TMM phantom to prevent ultrasound waves reflecting at the surface but was not applied until after this image was acquired, just prior to performing the sonications.
Gas entrapment was found to occur when the emulsion was mixed too vigorously and when the gel was cooled too quickly. Measures were taken to reduce gas entrapment. First, degassed and deionized water was used in the fabrication of the gel sets. The degassing was done by circulating deionized water through a dedicated degassing unit, using a vacuum pump and a semipermeable membrane-based gas extraction. The dissolved oxygen content was then measured with a fluorescence probe (Neofox, Ocean Insight, Orlando, FL, USA). The emulsion was also dispersed with a syringe and a needle, with the tip submerged in the liquid gel solution. The syringe was further used to mix the TMM solution with repeated cycles of drawing and dispensing, rather than with manual stirring. Additionally, the gels were cooled at room temperature for a prolonged time, prior to solidifying the gel sets in an ice bath, to allow entrapped air bubbles to naturally collect at the top surface. These methods help reduce the complications of unintentional atmospheric gases being entrapped in the TMM gel matrix.

Focused ultrasound system
The incident high intensity focused ultrasound (HIFU) sonications were generated by an acoustic field that was pulsed according to the duty cycle settings, at an acoustic frequency of 1.031 MHz. The transducer was a pseudorandom array consisting of 256 elements, 131 mm focal length, and 140 mm aperture. The transducer was connected to a beam former and impedance matching unit (Image Guided Therapy, Pessac, France). The 3T MRI system (Prisma Fit, Siemens, Erlangen, Germany) guided the procedure, using the body coil as the transmit antenna and an 11 cm loop coil as the receiver antenna. The HIFU unit was controlled from the Thermoguide graphical user interface (version 1.37.2002, Image Guided Therapy, Pessac, France).

MRI sequences
The gels were assessed with MR-ARFI and MRT to observe acoustic absorption. Simultaneous displacement and temperature were measured with MR-ARFI using a FLASH-ARFI-MRT sequence. Additionally, standard MRT was performed to ensure repeatability, to measure the acoustic absorption coefficient, to verify the focal point was in-plane, and to verify that the droplets were exhibiting more heating at higher concentrations. Care was taken to ensure that the positions of the MRT sonications and the MR-ARFI sonications were in distinct locations, as previous heating was indicated in the gels to affect the displacement measurements. The temperature and displacement map data were analyzed with a custom MATLAB script (The MathWorks, Inc., Natick, MA, USA).
To generate the FLASH-ARFI-MRT sequence, a standard fast low angle shot (FLASH) sequence (49) was modified to incorporate vertical motion encoding gradients (MEGs) in the focused ultrasound direction (7,50,51). The sequence featured symmetric bipolar MEGs centered near the external trigger, with equal amplitudes and opposite phase. The MEG duration was reduced to maintain echo time (TE) less than 15 ms for good signal-to-noise ratio (SNR). These tissue displacement measurements are relatively slow, with maximum displacement requiring a few milliseconds (8). Hence, the need to displace the external trigger from the center of the MEG gradients. A variable delay was incorporated to adjust the timing between the MEG pulse and an infrared fiber-optic signal to trigger the focused ultrasound with each repetition time (TR). The motion encoding gradients, which form the basis of the MR-ARFI imaging, were eventually incorporated into the standard Siemens MR thermometry EP_SEG_THERM sequence. This allowed for an EP_SEG_THERM_ARFI sequence, with the commercial sequence adjustments available, although the sequence was not used to generate the data in this report.
The phase contrast was equated to displacement and temperature from Equations 1 and 2 (7). The phase change in the displacement phase map can be increased linearly with a larger MEG gradient amplitude or longer MEG duration. Similarly, a larger TE will result in a larger phase contrast in the temperature phase images.
Here, u i is pixel displacement (μm), T i is pixel temperature (°C), g is gyromagnetic ratio of hydrogen (267.52e6 rad.s -1 .T -1 ), d is the MEG duration per lobe (ms), M is MEG gradient amplitude (mT.m -1 ), f + i is the positive polarity pixel value during HIFU sonication, f + 0 is the baseline positive polarity pixel value without HIFU sonication, f − i is the negative polarity pixel value during HIFU sonication, f − 0 is the baseline negative polarity pixel value without HIFU sonication, B 0 is the main magnetic field magnitude of 3 T, c is the proton resonance frequency shift (PRFS) temperature coefficient of -0.0094 ppm.°C -1 .
For standard gradient recall echo (GRE) thermometry sequences, the optimal SNR in temperature phase imaging occurs with TE at the effective spin-spin relaxation time (T * 2 ). (4) Since the FLASH-ARFI-MRT sequence is diffusion-weighted, the SNR is dependent on the MEG b-value. The minimum standard deviation of the average displacement measurement for the bipolar FLASH-ARFI-MRT sequence occurs approximately when one MEG lobe duration (d) is equal to ½T * 2 . The b-value was determined from Equation 3 Where b is the b-value, g is gyromagnetic ratio of hydrogen, M is MEG gradient amplitude, d is the MEG duration per lobe.
The phantom structure and droplet viability were verified by a T 1 -VIBE sequence and by good repeatability with MR temperature measurements of standard HIFU sonication. These measures indicated that the droplets had not aggregated or undergone Ostwald ripening, and the TMM phantoms were fabricated adequately and were free of entrapped gas bubbles. Previous studies showed that repeated sonications on this droplet formula in TMM gel phantoms resulted in a slight decrease in size and almost identical temperature gains (41).
In the scanner, the gel position was determined with a "localizer". A 3D T 1 -VIBE sequence was obtained for higher resolution and accuracy, for synchronizing the MRI coordinate position, verifying no prefocal abnormalities, and to place the specimen directly at the focal length of the HIFU transducer. The Standard MRT (EP_SEG_THERM, Siemens, Erlangen, Germany) was used for high-duty cycle and highertemperature sonications, more similar to ablative conditions. The sequence used a multislice temperature sensitive RF-spoiled segmented GRE-EPI sequence composed of three interleaved slices to monitor temperature elevation: [ A relaxometry mapping protocol (Myomaps, Siemens, Erlangen, Germany) was created for spin-lattice relaxation time (T 1 ) mapping, effective spin-lattice relaxation time (T * 1 ) mapping, and spin-spin relaxation time (T 2 ) mapping. The effective spin-spin relaxation time (T * 2 ) should be measured in future studies to optimize the SNR of the FLASH-ARFI-MRT sequence. Susceptibility-weighted imaging (SWI) was added to the protocol. A 16-slice coronal SWI sequence used the following parameters: [TE, 20 ms; TR, 27 ms; FA, 15°; BW, 120 Hz; SL, 1 mm; NA, 1; FOV, 120 mm x 120 mm; AM, 128x128]. The protocol was performed on select lecithin-PFOB and F 8 TAC 18 -PFOB gel sets. The body coil was used as the transmit coil and the 11 cm loop coil as the receiver. Relaxometry and SWI map values were compared across selected gel sets, at all concentrations. The values were post-processed using an average region of interest in the gel centers, across multiple slices.

Phantom and droplet absorption
From the MRT data, the change from baseline in the acoustic absorption coefficient at each droplet concentration interval was assessed. These values were measured relative to baseline, represented by the average of the control measurements in each individual gel set. A mono-exponential fit line was fit to the MRT data points using Equation 4.
Where a is the acoustic absorption coefficient (m -1 ), A is the maximum change in the acoustic absorption coefficient (m -1 ), C 0 is the critical concentration (% v:v), [C] is the percent volume fraction of the emulsion (% v:v).
The displacement data was compared between experiments, also relative to baseline. Denoted as the spatial integral displacement, all pixels with a measurable change in displacement values at the focal region were summed. The MR-ARFI spatial integral displacement data were fit with linear regression, and the fit line formulas along with R 2 values were determined. The interaction radii for these sets of Where R int is the interaction radius (μm), R 0 is the physical droplet radius (μm), C 0 is the critical concentration (v:v). The prime symbol distinguishes the values from separate experiments. The mean interparticle distance at the critical concentration was determined by Equation 6 (52).
Where a(C 0 ) is the mean interparticle distance at the critical concentration (μm).
The value Y is related to the skin depth and particle radius of the shear wave, given as . The variable R o denotes the droplet radius (m), a tv is the thermoviscous attenuation (m -1 ), w is the angular frequency (2pf, rad.s -1 ), f is the HIFU frequency (Hz), v m is the kinematic viscosity of the medium (m 2 .s -1 ), [C] is the droplet concentration (v:v), r e is the emulsion density (kg.m -3 ), and r m is the medium density (kg.m -3 ).

Statistical analysis of acoustic absorption
The resulting data from MR thermometry and MR-ARFI integral displacement measurements were assessed statistically with a bootstrap method to determine if the surfactant groups were significantly different. The F 8 TAC 18 and lecithin data sets were merged to a single set. Half the total data points were selected at random and collected in a group. A separate group was made with the remaining data points. Then, the two groups were separately fit with the same regression algorithm used previously, to obtain the critical concentration. For the bootstrap method, the difference between the critical concentrations was obtained at each iteration. This was repeated one thousand times to arrive at a histogram. The p-value was determined by finding the proportion of instances where the differences in the critical concentration were greater than those found in the original regression fitting.

Hydrophone system
Fabry-Perot fiber-optic hydrophone (FOHS, Precision Acoustics, Dorchester, UK) measurements were performed on a control gel and a 0.3% v:w F 8 TAC 18 -PFOB gel. A catheter was used to guide a hydrophone needle to the gel center, shown in Figure 3. A custom fiber optic cable was provided by the manufacturer, which was longer than the standard size. The cable was long enough to reach the center of the MRI bore, while connected to the hydrophone which was located outside the MRI Faraday cage, within the control room. Damaging the sensor would have been costly and caused many delays in data acquisition, so an alternative configuration was chosen. The hydrophone was designed to be used parallel to the beam path, though a perpendicular configuration was adopted for these experiments to prevent damage. In previous studies, the directional sensitivity of the sensor varied with frequency and the decibel recordings increased by 10% (about 1 dB) to account for the altered directionality (57). The voltage waveform was equated to pressure, with a hydrophone fiber optic sensitivity value of 126.3 mV.MPa -1 , corresponding to a HIFU frequency of 1.031 MHz, with uncertainties of 14% in this frequency range (57).
The actual hydrophone needle diameter was less than the MRI voxel resolution. The catheter size in Figure 3A appeared larger due to a susceptibility artifact. The focal spot was positioned incrementally, until the maximum temperature was obtained. The HIFU focal spot position was adjusted with electronic beam steering in the Thermoguide interface, as shown in Figure 3. First, the x-position was adjusted until the location for the maximum temperature increase was observed, and the coordinate was fixed. Next, the y-position was adjusted at this fixed x-coordinate until the maximum temperature increase was observed with equal sonication conditions. This location was then used for subsequent acoustic spectra recordings.
The hydrophone temperature recordings were performed before acoustic spectra recordings, to verify the samples were stable at increasing acoustic powers. The control gel was sonicated with a 99.9% duty cycle without simultaneous MRT, at 1.031 MHz, for 10 seconds with acoustic powers of 19 W, 32 W, 48 W, and 67 W. The 0.3% v:w perfluorocarbon gel was measured only at 67 W, with 5 second sonications to prevent overheating. The hydrophone voltage spectra were normalized to the main peak with logarithmic division. The relative increase in the harmonics between gels was compared to identify changes indicative of cavitation.
NMR system and sequences 19 F diffusion-ordered spectroscopy (DOSY) NMR was used to experimentally measure the droplet diffusion coefficient (58, 59). Topspin 4.0.5 (Bruker BioSpin Gmbh, Billerica, MA, USA) was used to obtain and process the spectra. The diffusion coefficient for a 2.3% v:v stock solution of 1.24 μm diameter F 8 TAC 18 -PFOB droplets was measured using a 19 F-DOSY pulse sequence on an NMR system (300 MHz, Avance III HD NanoBay, Bruker, Billerica, MA, USA). The sequence was longitudinal eddy current delay (LED) with bipolar gradient pulses and two spoiler gradients (59,60). The 19  The theoretical Stokes-Einstein diffusion coefficient was estimated using Equation 8.
Where the variable D denotes the diffusion coefficient of an individual PFOB droplet (m 2 .s -1 ), k b represents Boltzmann's constant (J.K -1 ), T is the absolute temperature (K), h denotes the dynamic viscosity of the medium (Pa), and R o is the particle radius (m).
Then, the time-averaged diffusion distance during HIFU sonications were determined using Equation 9.
Here, t is time (s), r is the three-dimensional displacement distance (m), and D is the diffusion coefficient (m 2 .s -1 ).
Quantitative 19 F-NMR was performed to verify the perfluorocarbon concentration in selected gel sets. The sequence was a 1D sequence with inverse gated decoupling and a 90°RF pulse (zgig), which generates a 1 H-decoupled fluorine spectrum without the nuclear Overhouser effect (NOE

Ultrasonography and perfusion system
The flow phantom used for ultrasonography was described in previous articles (42,43). The system was perfused by a pulsatile perfusion machine that was originally designed to evaluate kidney function from non-heart beating organ donors, prior to transplant, with 31 P-MRI spectroscopy (61). The perfusion system was composed of three primary components: the perfusion module, a drive module, and an "umbilical" system. The perfusion module was a container positioned within the MRI bore, in which was placed the kidney, or TMM phantom in this case. The drive module was composed of the hardware unit with a control interface, which was located in the MRI control room, and used compressed oxygen as the driver. An "umbilical" system connected the compressed oxygen to the perfusion module and the drive module, by passing through the ports in the Faraday cage.
To obtain the ultrasonography images, a specially designed 192-element MR-compatible abdominal ultrasound probe (IBMT, Fraunhofer, Sulzbach/Saar, Germany) was used with a clinical ultrasound scanner (Acuson Antares, Siemens Healthcare, Mountain View, CA; USA) to image cavitation effects during HIFU sonications. The transducer was designed to avoid interference from the MRI scanner during concurrent MRI and ultrasonography, with operating frequencies of 3-7.5 MHz and HIFU frequency at 1.031 MHz (37,62,63). The design reduced interference from the HIFU transducer and MRI system, allowing pulsed color Doppler and B-mode imaging of the perfused phantom.

Rheometry analysis
Certain gel sets were characterized with rheometry measurements to estimate the elastic modulus, shear modulus, Note, the estimated location of the focal spot relative to the sensor is within the distance of one voxel, providing a spatial precision of one-half voxel or 0.5 mm. The relative positions are enlarged to enhance visibility of the sonication pattern. and viscosity. This was performed primarily in an attempt to relate the outlier displacement to the relative elasticity between the samples. Measurements were made on a HAAKE MARS 40 rheometer (Thermo Fisher Scientific, Waltham, MA, USA), with rotational frequencies between about 1-50 Hz. The recorded values included the complex shear modulus (G*), shear storage modulus (G'), shear loss modulus (G " ), and the frequency (Hz). From this data, assuming a Poisson ratio of 0.5, the complex elastic modulus (E*) was estimated as 3G*. The dynamic viscosity was determined from Equation 10 (64). The viscosity value at 1 Hz frequency was used to approximate the zero-shear viscosity limit.
Results and discussion MRI displacement and temperature measurements In Figure 4A, the sustained sonications, used with the MR-ARFI TMM phantoms, were fit with a mono-exponential function given in Equation 4. One data set was omitted from the figure due to droplet size differences. The regression line fit variables of the F 8 TAC 18 -PFOB measurements (n=74) and the lecithin-PFOB measurements (n=25) are given in Table 1, together with the estimation of the thermoviscous attenuation and interaction radius. Each MRT data point represents independent cycles of prolonged heating for 33 seconds followed by cooling to room temperature. From each cycle, a single measure of acoustic absorption was obtained. The data consisted of four sets of F 8 TAC 18 -PFOB measurements and two sets of lecithin-PFOB measurements, with about five gels per set, and about four to five measurements per gel (maximum of seven).
In Figure 4B, the MR-ARFI spatial integral displacement data are shown. Outliers were observed in some scans. The outliers were possibly due to problems in constructing the samples for the individual experiment or technical difficulties with the MRgHIFU system. Simple linear regression was performed. The fit constants and R 2 values are shown in Table 2.
The phantoms were effective for low-duty cycle MR-ARFI measurements between 40-45°C, which were similar to mild hyperthermia applications, and sustained high-duty cycle MRT measurements, which were more similar to ablative sonication conditions. Even with low-energy and low-duty cycle sonications from MR-ARFI, the sonosensitive droplets were shown to increase ultrasound absorption. Figure 4 shows an expected increasing trend in MRT temperature and MR-ARFI displacement as the droplet concentration increased. The spatial integral of the displacement field gave less noisy data, when compared to values of maximum displacement and temperature measurements. The values of maximum displacement, maximum temperature, and integral temperature were also obtained. However, these values did not tend to return to equilibrium before subsequent sonications and were omitted for brevity.

Bootstrap statistics
Lecithin-PFOB droplets were previously tested extensively as intravenous radiological contrast agents, with results illustrating safety and efficacy in human subjects (65,66). This F 8 TAC 18 -PFOB emulsion has not been assessed in human studies, but many aspects of the behaviour in humans can be anticipated by comparing to lecithin-PFOB clinical studies. A comparative measure was wanted using an alternative surfactant formula to identify any effects that could result from the surfactant alone. The lecithin-PFOB emulsion was chosen as a comparison for this reason. Similar cavitation effects from the two formulas, using droplets of equal sizes, should give similar measures of acoustic absorption. The comparison was performed to verify this experimentally. The results were assessed statistically. A bootstrap method was applied to the MR thermometry critical concentrations in Figure 4A and the slope of the MR-ARFI integral displacement data in Figure 4B. The results indicated that for both the MR thermometry data (p=0.38) and the MR-ARFI integral displacement data (p=0.34) in Figure 4, there was insufficient evidence to reject the null hypothesis. That is, there was no significant difference in the observations between the two surfactant groups.

Hydrophone temperature and acoustic spectra measurements
The hydrophone temperature recordings in Figures 5A, B showed elevated temperatures in the perfluorocarbon sample compared to the control. The 0.3% v:w sample was heated at only half the duration of the control sample, though still showed an increased temperature profile. There was also a higher rate of change of temperature with time at 67 W, which may be due to nonlinear effects increasing the acoustic absorption coefficient. Techniques such as cavitation threshold measurements and passive acoustic mapping might be useful to quantify cavitation and nonlinear effects under a variety of scenarios, such as higher powers and alternate sonication conditions (39).
The hydrophone voltage recordings of the acoustic spectra are shown in Figures 5C, D. The 0.3% v:w F 8 TAC 18 -PFOB TMM gels showed lower amplitude sub-harmonics and increased ultra-harmonics, which might have contributed to the enhanced thermal absorption. Table 3 gives the harmonic values of the hydrophone voltage spectra normalized to the main peak. For the 0.3% PFOB sample, the scans were performed at 67 W acoustic power, had a 1.6 MPa peak negative pressure, a 2.7 MPa peak-to-peak pressure, a mechanical index of 1.6, and a duty cycle of 4.7%. For the control gel, the acoustic power was 67 W, a 0.5 MPa peak negative pressure, a peak-to-peak pressure of 1.0 MPa, a mechanical index of 0.5, and a duty cycle of 4.7%. The spectra were noticeably different at higher HIFU amplitudes and in the perfluorocarbon sample. Normalized relative to the main peak, the subharmonics decreased by approximately a factor of 4 and the ultraharmonics increased by approximately a factor of 3, for the 0.3% v:w F 8 TAC 18 -PFOB sample compared to the control. Stable cavitation generally results in increased sub-harmonics and ultra-harmonics, while inertial cavitation increases broadband emissions, often resulting from collapsing microbubbles (67). As the acoustic absorption coefficient scales with frequency, higher frequency harmonics from cavitation and nonlinear wave propagation can result in increased acoustic absorption coefficients, resulting in increased heating (67).
It is expected to be unfavorable for these droplets to undergo a phase change at these temperatures, acoustic pressures, and sonication conditions. Studies for low pressure chemotherapeutic drug delivery also illustrated a lack of PFOB vaporization under similar experimental conditions and droplet formula (20,68). Droplets made with PFOB have generally been stable at larger pressures and have not been readily used for techniques such as acoustic droplet vaporization. In some instances, the droplets in this study appeared to produce cavitation at lower mechanical index than the FDA diagnostic ultrasonography limit of 1.9. At similar

SWI, relaxometry, and ultrasonography
SWI was constant for all gel concentrations, with no definitive trend correlating with the emulsion concentration, except in the sample that possibly experienced cavitation. 1 H-MRI relaxometry mapping of the water molecules in the TMM samples showed the T 1 , T * 1 , and T 2 were not affected by the surfactant and no measurable distinction in relaxometry due to possible variations in dissolved atmospheric gases. The relaxometry mapping values are given in Figure 6.
Upon SWI and relaxometry imaging, the perfluorocarbon sample used during hydrophone measurements appeared to have undergone a phase change event during heating. The SWI and relaxometry maps indicated gas accumulation as shown in Figures 7A, B. A lesion was visible on these relaxometry maps, and a large hypointense signal was visible on SWI imaging, indicating likely gas accumulation at this location. Note that the SWI region is larger than the actual size of the gas volume, which is a result of a susceptibility artifact in the image. This artifact corresponded with a lesion in the same location, giving hyperintensity with T 1 , T * 1 , and T 2 mapping. The perfluorocarbon gel was measured one week later with the same sequence parameters and the hypointense SWI region had disappeared, indicating the gas accumulation had diffused back into the gel matrix.
These findings corresponded to previously observed boiling core effects in ex vivo pig kidneys perfused with 0.24% v:v of this emulsion (42). Further observations of these effects were seen with MR-compatible ultrasonography, which illustrated residual hyperintensity after circular-patterned HIFU sonications, as shown in Figures 8A, B. For the video, please see the supplementary materials. In this case, these effects might be explained by sonosensitive droplets acting as cavitation nuclei, to reduce the pressure threshold needed to generate bubble clouds, as seen in boiling histotripsy (71). That is, with a distinct cavitation mechanism occurring at the droplet periphery, rather than within the liquid perfluorocarbon (72).
Local perfluorocarbon concentration can be measured with 19F-MRI, though the acquisition times can be long and dedicated hardware is additionally required. The emulsion might further benefit from incorporating Fe3+ chelate ions Average relaxometry mapping values across multiple coronal slices as a function of emulsion concentration. No noticeable trend was observed with increasing emulsion concentration. The T 1 and T * 1 mapping in the final F 8 TAC 18 concentration was removed as the sample appeared to have possibly experienced cavitation during hydrophone recordings. Susceptibility-weighted imaging and T 1 relaxometry mapping. A residual bubble cloud appears to develop after HIFU sonications and likely contributes to the heat enhancement. (A) 0.3% v:w F 8 TAC 18 -PFOB gel phantom about one hour after HIFU heating during hydrophone recordings. The susceptibility effect was not seen in the control samples and disappeared within one week from the F 8 TAC 18 -PFOB sample. These measurements were performed after HIFU heating during hydrophone spectra recordings. (B) T 1 hyperintensity was visible with relaxometry mapping. and a chelating agent into the solution, to reduce the 19F T1 relaxation time, increase sensitivity, and reduce acquisition times. (73, 74) However, the corresponding T2 shortening might lead to inaccurate temperature measurements with 1H-MRT. 1 H-MRI oxygen mapping techniques to measure the change in R 1 , often employing inversion recovery sequences, have been reported in clinical studies for observing the change in local oxygen content after administering respiratory oxygen (75)(76)(77). It was tested if these effects, which could be an indirect measure of local concentration, could be observed under routine test conditions. The relaxometry measurements are shown in Figure 6. Relaxometry and SWI mapping were independent of emulsifier and showed no measurable distinction in values at higher droplet concentration, in absence of cavitation. The mean relaxometry in the TMM samples were measured at 1045-1085 ms, 720-790 ms, and 65-67 ms for T 1 , T * 1 , and T 2 , respectively. Relaxometry values of MR-ARFI TMM phantoms were close to previously reported similar TMM phantoms and to human cardiac tissue (48, 78).

Rheometry measurements
The rheometry data is provided in Figure 9. The TMM phantoms tended to show a reduction in elastic modulus, shear modulus, and viscosity with increasing PFOB concentration. Dynamic viscosity showed a strong decrease with higher rotational frequency. The viscosity at 1 Hz of 450 Pa.s at 0.3% v:w was used to estimate the zero-shear viscosity, for determining the time-averaged diffusion distance of the droplets in the TMM sample. The TMM control phantom was measured to have an elastic modulus near 9.5 kPa. The elastic modulus was relatively stable across varying rotational frequencies, and was reduced with increasing perfluorocarbon concentration. Elastic modulus for porcine myocardium were reported at about 60 kPa (79), about 8 kPa in breast tissue (80), and 10-190 kPa for liver and kidney (80). Similar gel phantoms reported elastic modulus measurements between 8-34 kPa (47).

F-DOSY NMR diffusion measurements
The Stokes-Einstein diffusion coefficient for 1.24 μm diameter particles in pure water at 298 K was determined to be 0.44 μm 2 .s -1 (81, 82). The 19 F-DOSY measurement found a diffusion coefficient of 1.191 μm 2 .s -1 , which was about 2.7 times larger than predicted by theory. This difference could be attributed to the emulsion solution having a lower dynamic viscosity than pure water. This corresponds with reduced viscosity at higher PFOB concentrations, as seen with rheometry measurements. An estimation was made for the time-averaged diffusion distance during HIFU sonications, accounting for size differences and medium viscosity, in both a previous parallel-strand TMM laminar flow model and these static ARFI samples. Comparison between time-averaged diffusion distance, the interaction radius, and the mean interparticle distance showed similar values, as shown in Table 4. In future studies, 19 F-DOSY could allow changes in size distribution before and after HIFU, or assess samples from long-term storage (83).

Interaction radius and diffusion analysis
It appeared that the short pulse duration of a few milliseconds for the MR-ARFI sequence generated a more linear temperature elevation with an increase in perfluorocarbon concentration. The longer HIFU pulse duration used in sustained sonication mode (tens of seconds) generated a nonlinear response that was fit with a first order asymptotic exponential function (43). This could be explained by the short pulse duration, that created a small interaction radius surrounding each droplet, which did not interact with neighboring droplets. However, for sustained sonications, the long pulses created an extended interaction radius surrounding each droplet, which increased beyond half the mean distance between particles and the heating effects saturated (43). Diffusion effects were not expected in MR-ARFI TMM due to the highly viscous gel structure. In free liquid solution, the diffusion distance appeared on the same order of magnitude as previous measurements of the interaction radius, as shown in Table 4. This suggested that the saturating effects might be influenced by diffusion in a perfused TMM flow phantom and perfused tissue.
Also, the interaction radius around an individual droplet was anticipated to be about 12.5 μm for a droplet that was 1.15 μm in radius, which would indicate increased acoustic absorption that results from a combination of thermoviscous heating and stable cavitation (43). The effects of stable cavitation were further supported by hydrophone spectra measurements observed in this current study. In a recent study, a novel cavitation mechanism was observed at negative pressures near 20 MPa, at the interface of unemulsified PFOB droplets in water, which could also influence the enhanced heating (72). That is, heterogeneous cavitation occurs at the external hydrophilic/hydrophobic surface of the droplets, Rheometry analysis of a F8TAC18-PFOB TMM gel set. forming cavitation bubbles consisting of water steam, and dissolved exogenous gases like oxygen and nitrogen. These cavitation bubbles can then further act as cavitation sites upon successive cycles from the incident HIFU pressure wave, by drawing more dissolved gases from the surrounding medium.
Other studies showed nucleation occurring in the droplet interior when studying alternative perfluorocarbon emulsions for acoustic droplet vaporization, but the effects from this particular droplet formula appear to exhibit a distinct mechanism. Since the surrounding medium in this case is a TMM gel, with a high viscosity, the cavitation effects might be reduced because the diffusion rate of the individual droplet and dissolved exogenous gases is much lower in the highly viscous medium. This would correspond with previously observed effects of further enhancement of these droplets in free liquid solution, compared to the droplets embedded in TMM gels.

Conclusions
In this study, a novel application using MR-ARFI for monitoring acoustic absorption from colloid-enhanced MRgHIFU therapy was assessed in vitro. Embedding TMM phantoms with perfluorocarbon emulsion adjuvants resulted in an increased focal spot displacement and a simultaneous temperature increase, with higher emulsion concentration. This increase in acoustic absorption and heating with a higher droplet concentration was also observed with sustained sonications, here and in previous studies (41)(42)(43). The change in acoustic absorption coefficient and integral displacement with droplet concentration was independent of surfactant, when comparing droplets of the same size. The results illustrated the great potential for the use of perfluorocarbon droplets to enhance magnetic resonance-guided thermal therapies. The techniques in this study could be applied to evaluate thermal enhancement with MRgFUS using a variety of colloids, including microbubbles, nanoparticles, and liposomes. A unique tissue mimicking gel phantom formula was developed that was compatible for use with both MR-ARFI and sustained MRgHIFU, with similar characteristics as ex vivo pig organs. Moreover, much insight was gained into the mechanism of enhanced acoustic absorption for this sonosensitive PFOB droplet formula. Further understanding could potentially help to improve ablative MRgHIFU procedures or even low power MRgFUS procedures like focused ultrasound-mediated drug delivery.

Data availability statement
The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

Author contributions
All authors listed have made a substantial, direct, and intellectual contribution to the work, and approved it for publication.

Funding
This project has received funding from the European Union's Horizon 2020 research and innovation programme under the Skłodowska-Curie grant agreement No 813766. Open access funding was provided by the University of Geneva.

Acknowledgments
The Center for Biomedical Imaging at University of Geneva is acknowledged for magnetic resonance facility access. Thanks to Andrew Hurrell and Paul Morris at Precision Acoustics for their help troubleshooting the hydrophone system. The authors are grateful to Marion Pupier for helpful discussions and making the NMR measurements. Thanks to Alexandre Porcello for guiding the rheometry measurements. Thanks to Professor Francois Lazeyras for helping to image the ex vivo pig kidney in Figure 1. We are grateful to the Siemens Healthineers, MR division for providing the source code of the proprietary FLASH and EP_SEG_THERM sequences.