Microbubble Composition and Preparation for High-Frequency Contrast-Enhanced Ultrasound Imaging: In Vitro and In Vivo Evaluation

Although high-frequency ultrasound imaging is gaining attention in various applications, hardly any ultrasound contrast agents (UCAs) dedicated to such frequencies (>15 MHz) are available for contrast-enhanced ultrasound (CEUS) imaging. Moreover, the composition of the limited commercially available UCAs for high-frequency CEUS (hfCEUS) is largely unknown, while shell properties have been shown to be an important factor for their performance. The aim of our study was to produce UCAs in-house for hfCEUS. Twelve different UCA formulations A-L were made by either sonication or mechanical agitation. The gas core consisted of C<sub>4</sub>F<sub>10</sub> and the main coating lipid was either 1,2-distearoyl-sn-glycero-3-phosphocholine (DSPC; A-F formulation) or 1,2-dipalmitoyl-sn-glycero-3-phosphocholine (DPPC; G-L formulation). Mechanical agitation resulted in UCAs with smaller microbubbles (number weighted mean diameter ~1 <inline-formula> <tex-math notation="LaTeX">$\mu \text{m}$ </tex-math></inline-formula>) than sonication (number weighted mean diameter ~2 <inline-formula> <tex-math notation="LaTeX">$\mu \text{m}$ </tex-math></inline-formula>). UCA formulations with similar size distributions but different main lipid components showed that the DPPC-based UCA formulations had higher nonlinear responses at both the fundamental and subharmonic frequencies in vitro for hfCEUS using the Vevo2100 high-frequency preclinical scanner (FUJIFILM VisualSonics, Inc.). In addition, UCA formulations F (DSPC-based) and L (DPPC-based) that were made by mechanical agitation performed similar in vitro to the commercially available Target-Ready MicroMarker (FUJIFILM VisualSonics, Inc.). UCA formulation F also performed similar to Target-Ready MicroMarker in vivo in pigs with similar mean contrast intensity within the kidney (<inline-formula> <tex-math notation="LaTeX">$n = 7$ </tex-math></inline-formula>), but formulation L did not. This is likely due to the lower stability of formulation L in vivo. Our study shows that DSPC-based microbubbles produced by mechanical agitation resulted in small microbubbles with high nonlinear responses suitable for hfCEUS imaging.


Microbubble Composition and Preparation for High-Frequency Contrast-Enhanced Ultrasound
Imaging: In Vitro and In Vivo Evaluation U LTRASOUND imaging at high frequencies (≥15 MHz) enables high-resolution imaging at the price of lower penetration depth, making this technique highly suitable for imaging of small animals [1] and superficial organs in large animals (e.g., skin [2]) and humans (e.g., eye [3] and skin tumors [4]). Alternatively, an endoscopic (e.g., transrectal [5]) or intravascular probe [6] could be used if the organ of interest lies deeper within the body.
Contrast-enhanced ultrasound (CEUS) imaging allows assessment of blood flow to improve diagnosis and monitor therapy. For CEUS, intravenously injected ultrasound contrast agents (UCAs) are needed that consist of gas-coated microbubbles dispersed in saline [7]- [9]. Examples of preclinical high-frequency CEUS (hfCEUS) are tumor angiogenesis imaging [10] and cerebral microvascular hemodynamics assessment in rats [11]. Microbubbles are effective UCAs with a strong resonance structure and inherently nonlinear behavior in response to a time-varying pressure field [10], [11]. These nonlinear oscillations can be present at the subharmonic (SH), fundamental, ultraharmonic, and higher harmonic frequency. The nonlinear signals generated by the microbubbles are utilized in imaging techniques to separate the UCA signal from that of the surrounding tissue [12].
Methods based on the detection of higher harmonics are hindered by artifacts when the excitation frequency is high (≥15 MHz), such as nonlinear propagation artifacts [13], [14]. In addition to those artifacts, higher harmonics undergo dramatic attenuation because of their high frequencies, which limits the penetration depth. Nonlinear fundamental contrast imaging is the most common approach This work is licensed under a Creative Commons Attribution 3.0 License. For more information, see http://creativecommons.org/licenses/by/3.0/ for nonlinear hfCEUS imaging [12]. The nonlinear fundamental component suffers less from attenuation, but nonlinear propagation is still a drawback. SH imaging, on the other hand, is free from such artifacts, is less attenuated, and neither generated during propagation in tissue nor scattered by tissue [15]. An SH signal can be achieved with minimum amplitude excitation if the driving frequency is twice the resonance frequency of the microbubbles [16], [17]. To increase the sensitivity of hfCEUS, both the imaging techniques and the UCA design need to be optimized. The performance of UCAs highly depends on the resonance behavior of a microbubble, which is inversely related to its diameter [17], [18]. Therefore, UCAs with smaller microbubbles can improve the sensitivity of hfCEUS imaging. Next to size, microbubble shell properties have been shown to be an important factor for their performance, mainly in the generation of SH [19], [20]. The lipid composition of the microbubble shell in the commercially available UCAs for hfCEUS is known only for Definity [21], but unknown for MicroMarker [22] and Targestar P-HF [23]. For highfrequency ultrasound molecular imaging, functionalizing the microbubbles is required to target them to the biomarker of interest [24]. Although Target-Ready MicroMarker provides streptavidin linkage [25], no UCA is commercially available to provide covalent coupling of the ligand to the microbubble shell, limiting flexibility in choice of ligands. In-house produced UCAs could overcome these disadvantages.
Based on a preliminary experiment we conducted before [26], the aim of our study was to produce lipid-coated UCAs in-house with high nonlinear response for hfCEUS. Based on microbubble size, a resonance frequency can be predicted [18], but the actual response still depends on the shell microstructure, as we previously showed for the main lipids used in commercially available UCAs [20], [21], [27]: 1,2-distearoyl-sn-glycero-3-phosphocholine (DSPC) or 1,2-dipalmitoyl-sn-glycero-3-phosphocholine (DPPC). We therefore studied 12 different UCA types with the aim to produce microbubbles with a size similar to Definity [21] and MicroMarker [22]. UCAs were either produced by sonication, i.e., the most common UCA production method [28], or mechanical agitation, i.e., the method by which Definity is prepared [21]. In addition, we compared the UCA compositions we [29] and others [30], [31] have used throughout the years, and quantitatively evaluated the performance of the in-house produced UCAs to Target-Ready MicroMarker in vitro and in vivo with hfCEUS using the most commonly used high-frequency ultrasound imaging platform (Vevo2100, FUJIFILM VisualSonics, Inc., Toronto, ON, Canada).
Coating formulations A-D and G-J were identical to our previously produced UCAs made by sonication for 10 s (number weighted mean ∼4 µm) [29]. The UCAs made by vial shaking either contained no glycerol and no propylene glycol (formulation D and J) as reported in [33] and [34] or contained 5% v/v glycerol (818709, Merck Millipore, Merck, Amsterdam, the Netherlands) and 5% v/v propylene glycol (82280, Sigma-Aldrich) (formulation E and K) as reported in [31], or 10% v/v glycerol and 20% v/v propylene glycol (formulation F and L) as reported in [30]. Our UCA formulations E, F, K, and L contained only two different lipids, which is typical for in-house produced UCAs made by vial shaking [30], [35]- [37]. We chose DSPE-PEG2000 in addition to the main lipid DSPC or DPPC because this lipid can be utilized for targeting, whereas the PEG40-stearate cannot [24]. The molar ratio of DSPC or DPPC to (DSPE-PEG2000 + DSPE-PEG2000-biotin) was kept identical to UCA formulations A-D and G-J.
UCAs produced by sonication (formulations A-C and G-I) were put in 5-mL glass serum bottles (223738; Wheaton, Millville, NJ, USA), topped with C 4 F 10 gas, and closed with a rubber stopper (Z166065; Sigma-Aldrich) and aluminum cap (224193-01; Wheaton). Before the experiments, UCAs made by sonication were washed three times by centrifugation at 400 g for 1 min (Heraeus Biofuge, Thermo Scientific, Etten-Leur, the Netherlands) to remove the excess lipids. For the vial shaking method, 1 mL of the solution (lipid concentration 0.4 mg/mL for formulations D-F and J-L) was pipetted in a 2-mL glass screw top vial (5182-0714; Agilent Technologies, Santa Clara, CA, USA), topped with C 4 F 10 gas, and closed with a screw cap (5182-0717; Agilent Technologies).
Target-Ready MicroMarker UCA was prepared from the ready kit (FUJIFILM VisualSonics, Inc.) according to the manufacturer's instructions: the lyophilisate cake was reconstituted with 1 mL of 0.9% sodium chloride solution. The microbubbles in this UCA consist of a phospholipid shell encapsulating a C 4 F 10 /N 2 gas core [22]. All UCAs were produced or reconstituted not more than a few hours before the experiments. Size distributions of the UCAs were measured using a Coulter Counter (Multisizer 3, Beckman Coulter, Mijdrecht, the Netherlands). A 20-µm aperture tube was used, allowing quantification of particle diameters between 0.4 and 12 µm using a linear spacing between the 256 channels. Measurements were repeated three times for each UCA to obtain the mean microbubble diameter, size distribution, and concentration. Polydispersity of the UCAs was calculated by assessing the SPAN, which illustrates the width of the distribution, using (d90%-d10%)/d50% where d10, d50, and d90 are the microbubble number weighted diameters below which 10%, 50%, and 90% of the cumulative amount of number weighted microbubbles is found. Data are presented with standard deviations (SDs).

B. In Vitro hfCEUS Imaging and Quantification
A schematic of our experimental setup is depicted in Fig. 1. The 10-mm diameter thin shell cylindrical tube, made of polypropylene film backing and coated with a water-based acrylic adhesive having a total thickness of 52 µm (Tesa 4024 PV 2), was mounted in a water tank with its center at the focus (18 mm) of the imaging probe. The tube was filled with 15 mL air-saturated PBS. UCA dilutions were pipetted in the tube and mixed gently using a magnetic stirrer in order to have a homogeneous suspension. For the first B-mode scan, the UCA concentration was 5 × 10 4 microbubbles/mL. Next, the effect of the UCA concentration on SH imaging was tested for selected UCAs using two concentrations: 8 × 10 6 and 4 × 10 5 microbubbles/mL, hereafter referred to as high and low concentration, respectively. All measurements were conducted within 10 min after pipetting the UCA suspensions into the tube. For each new UCA or dilution, the tube was washed with distilled air-saturated water, filled with air-saturated PBS, and placed in the same location in the water tank.
We used a high-frequency preclinical ultrasound scanner operated at 15 or 30 MHz, with two linear array transducers [MS200 probe (15 MHz) and MS250 probe (30 MHz), FUJI-FILM VisualSonics, Inc.]. The wide beamwidth setting was chosen in order to have a low more uniform transmit pressure over depth in the tube [12]. To study the scattering properties of all 12 UCA formulations, the MS200 probe was used at 15-MHz transmit frequency at 1% transmit power and one cycle pulse duration in B-mode. On the selected UCA formulations (C, I, F, L, and Target-Ready MicroMarker), SH imaging was performed with the MS250 probe. This probe has a center frequency of 22.5 MHz and a −6-dB two-way bandwidth of 70% (15)(16)(17)(18)(19)(20)(21)(22)(23)(24)(25)(26)(27)(28)(29)(30) [12], therefore suitable for SH imaging if transmitting at 30 MHz. The selection of the SH imaging parameters such as transmit frequency (30 MHz), power (10% corresponding to a peak-to-peak pressure of ∼200 kPa), pulse sequence (pulse inversion [12]), and length (20 cycles for better separation of the SH component in the frequency domain) were based on a previous study in which these parameters were optimized for SH imaging [15]. The focus of the MS250 probe was set at 18 mm.
For all experiments, radiofrequency (RF) signals were reconstructed from I/Q samples and further postprocessed offline using MATLAB (The MathWorks, Natick, MA, USA). Three regions of interest (ROIs) were selected within the tube containing the UCA: ROI 1 was centered at 15 mm; ROI 2 at 18 mm (focal point); and ROI 3 at 21 mm. Two hundred RF lines were averaged in the frequency domain for analysis of UCA responses at the fundamental (15 or 30 MHz) and SH (15 MHz) frequencies within each ROI.
RF data were digitally bandpass filtered around SH frequencies in the frequency domain, with a fifth-order Butterworth filter. The −6-dB frequency cutoff for the SH filter was chosen from 13 to 17 MHz. One should note that, in our analyses, the signals at the fundamental frequency (30 MHz) correspond to the nonlinear behavior of the UCAs at that frequency and not the linear backscattered signal. This is because the analyses were performed on the RF signals reconstructed from the pulse inversion sequence, which removes the linear components of the backscattered signal at the fundamental frequency [38].

C. In Vivo hfCEUS Imaging and Quantification
The animal protocol was approved by the animal ethics committee of the Erasmus MC (EMC3379 142-14-01) and conducted in strict accordance to the National Guidelines for Animal Care and Handling. After overnight fasting with free access to water, the female pigs (crossbred Landrace × Yorkshire, 3-4 months of age) of approximately 30 kg (n = 4) were premedicated with an intramuscular injection of tiletamine (5 mg/kg), zolazepam (5 mg/kg) (Zoletil, both Virbac Laboratories, Carros, France), and xylazine (2.25 mg/kg) (Sedazine 2%, AST Farma BV Oudewater, the Netherlands). Anesthesia was maintained with a combination of intravenous infusion of midazolam (1.5 mg/kg/h, Actavis, New Jersey, USA), ketamine (5 mg/kg/h, Alfasan, Woerden, the Netherlands), sufentanil (4 µg/kg/h, Sufenta Fort, Janssen Pharmaceuticals Ltd., USA), and rocuronium bromide (4 mg/kg/h, Fresenius Kabi, Germany) through an ear vein cannula. The animals were ventilated through an endotracheal tube (7.0 Fr), placed in the trachea via midline cervical tracheostomy in a volume controlled mode (Servo 300, Siemens-Elema, Solna, Sweden) with a fraction of inspired oxygen of 0.40, a frequency to achieve normocapnia, and a positive end-expiratory pressure of 5 cm H 2 O. Surgery via the right flank was performed to expose the right kidney. hfCEUS imaging was performed by manually injecting a 1 mL UCA bolus in the jugular vein followed by a 10 mL 0.9% sodium chloride flush. The performance of selected in-house produced UCA formulations (F and L) was compared with that of Target-Ready MicroMarker. The order in which they were injected was random. The nonlinear hfCEUS measurements were recorded using the Vevo2100 equipped with an MS250 transducer [18-MHz transmit frequency, 10 frames/s, 10% power, ∼400 kPa (MI < 0.1)]. The Vevo2100 uses the amplitude modulation pulse sequence in the nonlinear contrast imaging mode, which mainly detects the nonlinear fundamental component of the UCA [12]. Immediately after injection of the UCA into the catheter, the ventilation of the animal was paused to minimize movement due to breathing. After 25-30 s, the measurement had been completed and the ventilation was turned on again. Cine loops of side-by-side B-mode and nonlinear contrast mode images were stored as lossless DICOM images for further offline analysis using MATLAB. First, correction for tissue motion in the imaging plane was applied as described previously [39], [40]. Briefly, the motion pattern of tissue in the field of view was extracted from the B-mode images and applied to the contrast mode images to correct for the motion in the field of view. Three ROIs were chosen for every DICOM recording at a depth of 0.5-5, 5-9, and 9-13.5 mm. For each ROI, all pixel intensities were summed and normalized to the area of the ROI. The intensity in each frame was obtained to construct a time-intensity curve (TIC), the frames with the maximum intensity were detected, and the mean intensity and SD of this frame and the five frames before and after this frame were calculated. In addition, alternative ROIs were chosen to compare contrast enhancement in the artery with that in the microcirculation where no clear vascularity was visible. The ratio between the intensities in these ROIs was used to quantify the ability to discriminate the blood vessels from the peripheral enhancement.

D. Statistics
The ratios for the arteries and microvasculature of Target-Ready MicroMarker and UCA formulation F were tested for significance using paired-samples student's t-tests, after first ensuring that the data were normally distributed using Shapiro-Wilk normality tests. A p-value <0.05 was regarded as indicating significance.

A. Ultrasound Contrast Agent Preparation
After production, all UCA formulations appeared white as shown in supporting Fig. 1 the volume weighted size distribution is shown in supporting Fig. 2. None of the A-L type UCA formulations had a number weighted mean diameter, volume weighted mean diameter, or size distribution identical to that of Target-Ready MicroMarker. UCA formulations A-C and G-I made by sonication [ Fig. 2(a) and (c)] contained higher amounts of large microbubbles than Target-Ready MicroMarker. All the UCAs produced by sonication had two distinct peaks in their size distribution (0.4-0.5 and 2-3 µm). Longer sonication times did not alter the mean number weighted diameter much, as shown in Table II. The volume weighted diameter decreased upon longer sonication times, while the concentration of microbubbles increased. UCA formulations D-F and J-L made by vial shaking resulted in smaller microbubbles [Figs. 2(b) and 2(d); Table II] than when microbubbles were made by sonication. DSPC or DPPC as main coating resulted in microbubbles of similar mean number weighted diameters for both UCA production methods (Table II). The highest microbubble concentration was found in UCA formulation E. Microbubbles with the smallest mean number weighted diameter were found in UCA formulation D and J, while this was UCA formulation L for microbubbles with the smallest mean volume weighted diameter, all of which were produced by vial shaking. The highest mean number weighted diameters were found in formulations B and H, and the highest volume weighted diameter was found in UCA formulation J. The UCA formulation with the smallest SPAN was formulation D, while formulation F had the largest SPAN.

B. In Vitro hfCEUS
As shown in Fig. 3, different ultrasound scattering intensities of UCA formulations A-L were observed at 15 MHz in vitro. The error bars in Fig. 3 represent the variation in B-mode signal intensity within the ROI (the entire cross section of the cylindrical tube containing the UCA). For both the DSPC-based and DPPC-based UCA formulations, Intensities obtained from the PBS control were subtracted from the intensities obtained from the UCA formulations. A-F type UCAs had DSPC as the main coating lipid, while this was DPPC for G-L type UCAs. microbubbles with larger mean diameters resulted in slightly higher intensities. In addition, UCAs produced by sonication that had similar mean number weighted diameters also showed similar signal intensity in the B-mode scan (A, B, C and G, H, I in Fig. 3). UCA formulation F had the highest signal of the DSPC-based UCAs made by vial shaking, while this was formulation K for the DPPC-based UCAs. However, a large SD was observed for formulation K, due to a nonuniform signal throughout the tube. Two UCA formulations produced by sonication and two produced by vial shaking were selected for further studies. Based on the results so far, UCA formulations C and I were selected out of the sonication produced UCAs as they contained the highest concentration of microbubbles. For the vial shaking produced UCAs, formulations F and L were selected because they gave the highest ultrasound signal in B-mode with the smallest SD.
Images of UCA formulations C, F, I, and L, and Target-Ready MicroMarker at high and low concentrations, filtered around their SH frequency (15 MHz), and the corresponding frequency spectra for the three ROIs are presented in Fig. 4. At high UCA concentration, attenuation was dominant for formulations C, I, and L. This attenuation effect is also reflected in the corresponding spectra of these UCAs, where the amplitude of the spectra at the SH frequency drops about 10 dB for the deeper ROIs with respect to the highest SH amplitude. Target-Ready MicroMarker and UCA formulation F at high concentration had the highest and most homogeneous SH response throughout the three ROIs. At low UCA concentration, the attenuation effect was less pronounced. The SH amplitude of Target-Ready MicroMarker dropped about 11 dB when UCA concentration was reduced by a factor of 20. The SH amplitude of UCA formulation I at such low concentration was 10 dB higher than that for Target-Ready MicroMarker and was homogeneous throughout the UCA area.
In all the corresponding spectra of the UCA signals in both high and low concentration, the amplitude of the nonlinear fundamental signal was maximal at the focus of the transducer, where the acoustic energy was at its maximum (ROI 2). At high concentrations, Target-Ready MicroMarker showed In vitro SH images of Target-Ready MicroMarker and formulation C, I, F, and L at high (8.0 × 10 6 microbubbles/mL) and low (4.0×10 5 microbubbles/mL) concentrations (top) and the corresponding spectra of each ROI at three depths for each image (bottom). the highest nonlinear response at the fundamental frequency. However, at low concentration, all our in-house produced UCAs had higher nonlinear responses at the fundamental frequency than Target-Ready MicroMarker. The attenuation effect was similar for the SH response and the nonlinear fundamental response. UCA formulations C, I, and L showed higher attenuation than Target-Ready MicroMarker and UCA formulation F.

C. In Vivo hfCEUS
Of the four formulations that were studied in vitro for their SH and nonlinear fundamental response, the two best performing UCA formulations were selected for in vivo hfCEUS studies: F and L. This decision was based on the fact that both the acoustic signal and microbubble size distribution resembled Target-Ready MicroMarker the closest. Fig. 5 shows the result of the comparison between Target-Ready MicroMarker and UCA formulations F and L in the same animal (n = 1).
For each recording, three ROIs were drawn: in the focal region (blue), in the middle of the field of view (red), and the bottom of the field of view (yellow). The mean ± SD intensity in each ROI is shown in Fig. 5 (bottom) for all three UCAs. hfCEUS images revealed slightly lower overall intensities for UCA formulation F than for Target-Ready MicroMarker. Formulation L on the other hand revealed only two larger vessels in ROI3, while hardly any intensity increase was observed outside these vessels or in ROI1 and ROI2. Because contrast enhancement of UCA formulation L was negligible, only Target-Ready MicroMarker and UCA formulation F were further evaluated. The TICs corresponding to the example in Fig. 5 for Target-Ready MicroMarker (gray) and UCA formulation F (black) show very similar behavior between both UCAs in all three ROIs (Fig. 6). The experiments comparing Target-Ready MicroMarker and formulation F (n = 7 in total) confirmed the observation that the total contrast enhancement of Target-Ready Micro-Marker was higher, although the difference was lower in ROI3 (Fig. 7).  This implies lower attenuation for UCA formulation F than for Target-Ready MicroMarker. Since UCA formulation F seemed to better visualize the larger vessels and Target-Ready MicroMarker resulted in more enhancement in the microvasculature (see Fig. 5), we quantified the ratio between enhancement in the arteries and the microvasculature for all injections. The symbols that are used in Fig. 7 correspond to those in Fig. 8, and quantification of the example in Fig. 5 (∇) shows better discrimination of the artery from the microvasculature for UCA formulation F. Although this example showed a clear difference, overall the differences Fig. 8. Comparison of the ratios of contrast enhancement in the arteries and the microvasculature between Target-Ready MicroMarker (gray) and UCA formulation F (white). The symbols, identical to the symbols in Fig. 7, indicate the data obtained from the same paired injection, whereas ∇ indicates the example shown in Fig. 5 (measurement 1, animal a). Measurement 2 was done in animal b and measurement 3-7 in animal c. The lines connect the paired injections (injected in random order). Differences between the two UCAs were not significant. between Target-Ready MicroMarker and UCA formulation F were not significant ( p > 0.05).
We also verified the reproducibility of UCA formulation F by repetitive injection in the same animal (n = 3) and found similar intensities in ROI1 and ROI3 between the first and second injection (Fig. 9). In ROI2, the second injection was 17% lower than the first injection. The third injection was 39% lower than the first injection in ROI1, 31% lower in ROI2, and only 11% lower in ROI3. We verified that the baseline values before the start of each measurement were comparable. Reproducibility for Target-Ready MicroMarker was not studied.

IV. DISCUSSION
In this paper, we showed that one of our in-house produced UCAs for hfCEUS resembled the performance of the commercially available Target-Ready MicroMarker both in vitro and in vivo. The best performing UCA was produced by 45 s vial shaking and consisted of 92.4% DSPC and 7.6% DSPE-PEG2000, in an aqueous solution of 70% PBS, 10% glycerol, and 20% propylene glycol with a C 4 F 10 gas core.

A. Ultrasound Contrast Agent Preparation
In our study, microbubble diameters were larger for sonication produced UCAs than when microbubbles were produced by vial shaking, which is in line with what Sirsi et al. [33] found for their in-house produced UCAs with a coating of DSPC and PEG-40 stearate (9:1 molar ratio) and C 4 F 10 gas core. In contrast, Moran et al. [34] reported similar number weighted mean diameters of ∼0.5 µm for their in-house produced UCAs by sonication and vial shaking. Their nitrogen-filled microbubbles had a coating of dipalmitoyl phosphatidylethanolamine (DPPE), phosphatidylcholine, dipalmitoyl phosphatidyl-DL-glycerol, and cholesterol. The contrasting findings in microbubble diameters between the two production methods, i.e., sonication and vial shaking, suggest that the microbubble composition and gas core may also play a role in the size of the produced microbubbles. On the other hand, different probe-sonication devices were used in these studies (Sonicator Ultrasonic Processor XL2020 in our study; Branson Ultrasonics Model 250 A in [33] and [41]; and Soniprep 150 in [34]), which could have given different ultrasonic power outputs thereby influencing the microbubble size distribution [28].
Kooiman et al. [29] reported that 10-s sonication for the same UCA formulations as A-C (DSPC-based) and G-I (DPPC-based) resulted in microbubbles with a mean number weighted diameter of 4.2 and 3.9 µm, respectively. Although sonication for 60 s resulted in smaller microbubbles (∼2 µm in mean number weighted diameter; see Table II), the number weighted mean diameter was similar after 60-, 90-, or 120-s sonication. This was also observed by Moran et al. [34] who reported that the number weighted mean diameter of ∼0.5 µm did not change when the sonication time was varied among 30, 60, 90, or 300 s for their lipid-coated microbubbles (coating composition see above). However, the number weighted size distribution of our UCAs produced by sonication had two peaks (0.4-0.5 and 23 µm; Fig. 2). A longer sonication time seemed to produce lower amounts of large microbubbles (>2 µm) and more microbubbles in between the peaks (0.5-2 µm). This result is confirmed with the changes in the volume weighted mean diameters, which decreased with increasing sonication time. In addition, microbubble concentrations increased for longer sonication times. It is likely that more microbubbles can be produced during longer sonication times because excess unincorporated lipids are always present when microbubbles are made by sonication [42]. The increase in the concentration and the changes in the size distribution of the UCA suggest that longer sonication times can break up bigger microbubbles into smaller ones.
Our in-house produced UCA formulations D and J were made by vial shaking in the absence of glycerol and propylene glycol, just as Sirsi et al. [33] and Moran et al. [34] did. On the other hand, glycerol and propylene glycol are often added to increase the fluid viscosity when making UCAs inhouse by vial shaking [30], [35]- [37] and are also present in the clinically approved Definity [21]. When the concentrations of glycerol and propylene glycol were increased from 0%/0% to 5%/5% and 10%/20% v/v, we found varying results on microbubble size and concentration. For the DPPC-based UCAs (formulations J, K, and L), increasing the concentration of glycerol and propylene glycol increased the microbubble concentration and the number weighted mean diameter. For the DSPC-based UCAs (formulations D, E, and F), the number weighted mean diameter followed the same trend. The concentration of the microbubbles, on the other hand, first increased with 5%/5% glycerol and propylene glycol and decreased for the highest concentration. In addition, UCA formulations F and L were more stable after 2 h than those without glycerol and propylene glycol. This suggests that glycerol and propylene glycol can play a role not only in increasing the concentration of the microbubbles, but also on the stability of the produced microbubbles. This increased stability of UCA formulations F and L was also observed in the increased intensity in the in vitro B-mode images of these UCAs.

B. In Vitro hfCEUS
In an in vitro setup, mimicking a practical imaging condition, we showed that both the shell microstructure of microbubbles as well as their size distribution have a considerable impact on their nonlinear behavior both at the SH and fundamental frequencies. All our homemade UCAs showed high nonlinear behavior when excited at 30 MHz, which was comparable with the commercially available Target-Ready MicroMarker. Such high nonlinear response at both the SH and fundamental frequencies can be attributed to the small sizes of all studied microbubbles. The larger microbubbles in formulations C and I (2-3 µm) accounted for more attenuation at high concentrations. This hypothesis is confirmed by the lower attenuation of the smaller microbubbles in UCAs F and L, while the nonlinear responses remained high. The attenuation effect was more pronounced for the SH component than the nonlinear responses at the fundamental frequency. This is perhaps due to the threshold behavior of SH oscillations [14]. The microbubbles at the lower part of the focal zone may not have been excited with sufficiently high pressures to undergo SH oscillation. However, when the concentration was 20× lower, the SH response of these larger microbubbles became stronger and more homogeneous throughout the sample.
Comparison of UCA formulations C and I with similar size distributions but different main lipid component showed that the DPPC-based UCA formulation I showed higher nonlinear responses at both the fundamental and SH frequencies. At low concentration, the response of UCA formulation I at the fundamental and SH frequency was 6 and 3 dB higher, respectively, than for the DSPC-based UCA formulation C. This higher SH response has also been reported in [20] utilizing optical single microbubble spectroscopy in the range of 1-4 MHz.
The small size of Target-Ready MicroMarker microbubbles seems to play an important role in its nonlinear behavior, particularly for the SH response at high-frequency excitation. The majority of the microbubbles have a diameter below 1.5 µm, corresponding to a resonance frequency around and higher than 15 MHz [43], [44]. It is conventionally thought that SH generation is achieved most readily (i.e., at lowest pressure threshold) when microbubbles are excited at twice their resonance frequency [16]. This means that for the majority of the Target-Ready MicroMarker microbubbles, the 30-MHz excitation used in this paper is around twice their resonance frequency. Indeed, high amplitude SH response of Target-Ready MicroMarker UCA has previously been reported by Helfield et al. [19], although they studied individual microbubbles in an acoustical setup. Another reason for the high SH behavior of the UCAs we observed may be the rectangular shape of the excitation used in the Vevo2100 scanner. It has been reported that rectangular-shaped excitations generate a self-demodulation signal, which enhances the SH response of microbubbles [14], [15]. Our in vitro results show that Target-Ready MicroMarker is also producing a strong nonlinear response at the fundamental frequency. This characteristic is already being utilized in the Vevo2100 scanner as the default nonlinear contrast mode with amplitude modulation pulse sequence to improve the contrast to tissue ratio in hfCEUS.
It has been shown before that smaller microbubbles provide higher contrast for hfCEUS imaging. Goertz et al. [45] reported that the decantation technique to isolate smaller Definity microbubbles can be employed to produce an attenuation pattern that appears more favorable for hfCEUS. Also Moran et al. [34] have shown that decreasing the mean diameter of the lipid-based sonication UCAs significantly increased the mean backscattering power at 40 MHz. Peyman et al. [46] showed that the majority of the signal measured using the VisualSonics Vevo770 (40 MHz) in the mixed microbubble/nanobubble population was attributable to the nanobubbles, with a much lower proportion of the signal coming from microbubbles. Our results are in agreement with these studies.
We showed that at 30-MHz transmit frequency and at low UCA concentration, DPPC-based UCAs behaved more nonlinearly than those based on DSPC. At higher concentration, smaller microbubbles (Target-Ready MicroMarker and UCA formulation F) produced the most homogeneous SH responses. For UCA formulations C and I, Kooiman et al. [29] observed differences in lipid distribution. The DPPC lipid has a lower elastic compressibility modulus than DSPC [47], [48], which may be related to SH behavior as our DPPC-based UCAs showed more SH behavior than our DSPC-based UCAs, as reported before for single microbubbles [20]. Unfortunately, the coating composition of Target-Ready MicroMarker is unknown, so we do not know how our homemade UCAs resemble the composition of Target-Ready MicroMarker.

C. In Vivo hfCEUS
We studied UCA formulations F and L in vivo and found that the overall performance of UCA formulation F was similar to that of Target-Ready MicroMarker. UCA formulation L, on the other hand, resulted in hardly any contrast enhancement despite the fact that both in-house produced UCAs had similar number-weighted mean diameters and the concentration of microbubbles was even higher for UCA formulation L. The most important difference between the two UCAs is their composition and the associated differences between DSPC as main component (formulation F) and DPPC (formulation L) in terms of microstructure [29, [49]- [51] and acoustic behavior in vitro [20]. Using an Acuson Sequoia 512 in nonlinear imaging mode at 7 MHz (15L8 transducer), the in vitro halflife of DPPC-based UCAs (DPPC:DSPE-PEG5000, 9:1, inhouse produced by vial shaking) was 10× lower than for those based on DSPC, whereas the half-life was better in vivo, but still 5× lower than for DSPC-based UCAs [52]. Although these and our in-house produced DPPC-based UCAs did not perform well in vivo, the commercially available and clinically approved UCA Definity that is also DPPC-based [21] has been shown to provide good contrast enhancement in pigs [53] and humans [54], [55]. It has to be noted that the other components of Definity are DPPA and DPPE-mPEG5000 [21], which can alter the acoustic properties and stability.
Other research groups have also characterized their own inhouse produced UCAs for hfCEUS, but either characterized them acoustically in vitro [19], [20], [46], [56]- [58] and/or imaged them in vivo [33], [46], instead of imaging them in both situations. Although acoustic characterization of single microbubbles provides insights in microbubble behavior, imaging of microbubbles in vitro in a controlled and simplified setup is essential before translation toward complex in vivo applications. Neither in vivo studies compared their in-house produced UCAs to a commercially available UCA as a reference. Peyman et al. [46] imaged their microbubbles (DPPC:DSPE-PEG2000, 4:1, microfluidic production) in the mouse aorta at 40 MHz in fundamental mode using the Vevo770, which resulted in hardly any contrast enhancement. Simultaneously, with their microbubbles, they also produced nanobubbles (∼200-nm diameter), which provided 3× higher contrast peak intensities. Sirsi et al. [33] isolated microbubbles (DSPC:PEG40-stearate, 9:1 molar ratio, both sonication and vial shaking) of distinct size populations (1-2, 2-4, and 6-8 µm) and determined their acoustic impact directly in vivo in the mouse kidney using 40-MHz fundamental mode imaging using the Vevo770. Surprisingly, they report that the smaller microbubbles (1-2 µm) resulted in higher attenuation and less echogenicity than the larger microbubbles (6-8 µm). One reason for these contradictory results could be the differences in microbubble shell composition in these studies. Also, in the study conducted by Sirsi et al. [33], the targets were the small capillaries in the mouse kidney, while Peyman et al. [46] quantified the UCA contrast in the mouse aorta that is a much larger vessel. Both the concentration of the microbubbles in the ROI and the differences in ambient pressures in kidney capillaries and the aorta could have played a role in these contradictory findings.
The TICs of UCA formulation F and Target-Ready MicroMarker were very similar, but slightly higher peaks and longer contrast persistence were found for Target-Ready MicroMarker. The prolonged contrast enhancement is likely due to the higher acoustic stability of Target-Ready Micro-Marker. These microbubbles not only contain a lipid shell, but are also covered by streptavidin, which has been shown to increase the stiffness and therefore reduce acoustic dissolution in vitro [24]. The higher contrast peak intensity for Target-Ready MicroMarker was shown to be present for all injections, but the difference with UCA formulation F was smaller in the deeper cortex. This may result from the higher concentration of formulation F microbubbles that were injected and appeared in the large vessels in the deeper cortex. The differences in concentration between Target-Ready MicroMarker and UCA formulation F influenced the interpretation of our results. We concluded that the performance between both UCAs was comparable, but it has to be taken into account that an injection of 1 mL UCA formulation F contained ∼7× more microbubbles than 1 mL of Target-Ready MicroMarker. However, both concentrations are the same order of magnitude as the commercially available ones, since a typical human dose of Definity contains 4.2 × 10 9 microbubbles [21] and a typical dose of SonoVue/Lumason varies between 3.0 × 10 8 and 1.1 × 10 9 microbubbles [27]. In addition, the concentration of microbubbles between batches of Target-Ready MicroMarker that were used in our experiments varied between 2.6 × 10 8 and 1.3 × 10 9 microbubbles/mL (n = 5). These concentrations were assessed by us and are in line with what others have reported, namely, 8.4 × 10 8 [25], 9.2 × 10 8 [59], and 1.9 × 10 9 [60]. The batch to batch variability in microbubble concentration for our UCA formulation F ranged from 2.1 × 10 9 to 5.5 × 10 9 microbubbles/mL (n = 5 batches) and was therefore narrower than for Target-Ready Micro-Marker. The variation in the mean microbubble diameter (both number and volume weighted) was similar for Target-Ready MicroMarker and our in-house produced UCA formulation F. The number weighted mean diameter for Target-Ready MicroMarker was 1.8 ± 0.4 and the volume-weighted mean diameter was 4.6 ± 0.8, while this was 1.2 ± 0.1 and 5.6 ± 1.0 for our in-house produced F type UCA (both n = 5 batches). The variability in microbubble size and concentration can be caused by vial handling, even within manufacturer's recommended procedures, as has been suggested by Goertz et al. [45].
Repeated bolus injections of UCA formulation F in the same animal resulted in significantly less enhancement for the third injection. Others have reported differences between consecutive injections in cats [61] and mice [62], [63]. In the kidney and spleen of healthy cats, the second bolus injection of SonoVue resulted in higher peak intensities [61]. Dizeux et al. [63] reported that the maximum intensity (peak enhancement) was constant for four consecutive injections in the healthy renal murine cortex, but increased from the second to the fourth injection in renal tumor tissue. In contrast, Rix et al. [62] reported constant peak enhancement in murine liver tumors and decreasing peak enhancement with consecutive injections in the healthy liver. We can only conclude that variability in consecutive injections is a known issue for CEUS and differs between organisms and organs.

D. Limitations
The main limitation of using in-house produced UCAs is that it can be challenging to produce them under good manufacturing practice conditions [64]. For example, batch to batch reproducibility may be an issue, as is sterility. However, for terminal preclinical experiments as in our study, the UCAs do not have to be sterile. For longitudinal preclinical experiments, there are options for sterilization of lipid formulations [65], [66]. In our in vivo experiments, we used nonlinear fundamental imaging at 18 MHz with amplitude modulation instead of SH imaging with pulse inversion at 30 MHz as used in vitro. Although SH imaging is free from nonlinear propagation artifacts, nonlinear fundamental imaging with amplitude modulation provides higher contrast to tissue ratio with Vevo2100 [15].

V. CONCLUSION
Our results suggest that our UCA formulation F performs equally well as Target-Ready MicroMarker in hfCEUS imaging. This paper shows that small UCAs having high nonlinear responses for hfCEUS can be produced by mechanical agitation, a shell composition of 92.4% DSPC, 6.4% DSPE-PEG2000, and 1.2% DSPE-PEG2000-biotin encapsulating a C 4 F 10 gas core, in a PBS-based liquid with 10% glycerol and 20% propylene glycol.