Fast Acoustic Wave Sparsely Activated Localization Microscopy: Ultrasound Super-Resolution Using Plane-Wave Activation of Nanodroplets

Localization-based ultrasound super-resolution imaging using microbubble contrast agents and phase-change nanodroplets has been developed to visualize microvascular structures beyond the diffraction limit. However, the long data acquisition time makes the clinical translation more challenging. In this study, fast acoustic wave sparsely activated localization microscopy (fast-AWSALM) was developed to achieve super-resolved frames with subsecond temporal resolution, by using low-boiling-point octafluoropropane nanodroplets and high frame rate plane waves for activation, destruction, as well as imaging. Fast-AWSALM was demonstrated on an in vitro microvascular phantom to super-resolve structures that could not be resolved by conventional B-mode imaging. The effects of the temperature and mechanical index on fast-AWSALM were investigated. The experimental results show that subwavelength microstructures as small as $190~\mu \text{m}$ were resolvable in 200 ms with plane-wave transmission at a center frequency of 3.5 MHz and a pulse repetition frequency of 5000 Hz. This is about a 3.5-fold reduction in point spread function full-width-half-maximum compared to that measured in the conventional B-mode, and two orders of magnitude faster than the recently reported AWSALM under a nonflow/very slow flow situations and other localization-based methods. Just as in AWSALM, fast-AWSALM does not require flow, as is required by current microbubble-based ultrasound super-resolution techniques. In conclusion, this study shows the promise of fast-AWSALM, a super-resolution ultrasound technique using nanodroplets, which can generate super-resolution images in milliseconds and does not require flow.


I. INTRODUCTION
L OCALIZATION-BASED ultrasound super-resolution imaging techniques are rapidly evolving. Early studies reported that the detection of isolated microbubble signals in the radio frequency (RF) domain [1], [2] or image domain [3], [4] can be used to generate super-resolution imaging in vitro. Two recent studies have shown the capability of using these techniques with microbubbles in vivo to super-resolve the microvessels and also obtain flow information [5], [6]. Subsequently, a number of papers have been published which improve and extend the microbubbleassisted super-resolution imaging technique in different ways, including signal filtering and detection [7], [8], choice of localization method [9], motion correction [10], signal tracking [11], and 3-D imaging [12]- [14].
Current localization-based super-resolution imaging techniques using microbubbles rely on sufficient flow and most require a low bubble concentration to provide spatially isolated localizations. Thus, longer acquisition time is required for sampling microbubbles in small vessels as they are very sparsely distributed at the capillary level and flow slowly in such vessels.
Nanodroplets, as ultrasound contrast agents, have three main advantages over microbubbles [15]. First, the nanosize of nanodroplets potentially allows extravasation into cancerous tissue due to its leaky vasculature and enhanced permeability and retention effects. Second, it has been found that nanodroplets can last longer during in vivo circulation than microbubbles [16]. Third, the nanodroplets can be selectively activated, both spatially and temporally, to provide an ultrasound contrast signal [17]. This potentially provides more flexibility during the ultrasound scanning.
Recent studies have reported ultrasound super-resolution imaging using low-(boiling point = −2°C) [18], [19] and high-(boiling point = 56°C) [20], [21] boiling-point perfluorocarbon nanodroplets. For the super-resolution imaging technique using high-boiling-point nanodroplets, optical activation via pulsed laser illumination was required to perform the super-resolution imaging, thus restricting its penetration depth to regions that can be illuminated optically. Furthermore, as the pulse repetition frequency of laser activation was 10 Hz, longer acquisition time is required in comparison to the fast ultrasound imaging methods with a pulse This work is licensed under a Creative Commons Attribution 3.0 License. For more information, see http://creativecommons.org/licenses/by/3.0/ repetition frequency over thousands of Hertz. The "acoustic wave sparsely activated localization microscopy (AWSALM)," developed by our group, has shown that the use of low-boilingpoint perfluorocarbon nanodroplets (with much lower activation threshold than high-boiling-point nanodroplets) can be acoustically activated and deactivated by diagnostic ultrasound pulses to perform ultrasound super-resolution imaging in deep tissue. Unlike microbubble-based super-resolution techniques, AWSALM does not require flow, nor does it require a low concentration of the contrast agent.
Microbubble-based ultrasound super-resolution technique requires long data acquisition time, which together with the motion during data acquisition makes it challenging for clinical use. AWSALM requires different ultrasound transmissions for droplets activation/deactivation and imaging, limiting the temporal resolution achievable.
In this study, we develop fast-AWSALM and demonstrate that a high concentration of low-boiling-point octafluoropropane (OFP) nanodroplets can be simultaneously imaged, activated, and deactivated by high-frame-rate plane-wave pulses at an intermediate acoustic amplitude to perform ultrasound super-resolution imaging on the time scale of hundreds of milliseconds. One of the fundamental distinctions between bubble-and droplet-based super-resolution imaging techniques is that while using bubbles localizations in consequent frames are not independent and they decorrelate with the velocity of flow, which is slow in the capillaries even if high concentrations of bubbles are present. However, with droplets being activated and deactivated at the same rate with ultrasound pulse repetition frequency, new localizations can be accumulated as quickly as the imaging pulses are transmitted. This fast and controlled activation and deactivation of nanodroplets make it possible to achieve a much faster super-resolution imaging.
In order to condense microbubbles into nanodroplets, the headspace of the vial was pressurized according to the previously described methods [25]. The vial of microbubble solution was immersed in an ice-salt bath (about −10°C) for 3 min followed by pressurization (85-130 lbf/in 2 ) with ambient air into the vial septum for the condensation.

B. Characterization of OFP Microbubbles and Nanodroplets
The size and concentration of OFP nanodroplets was measured by NanoSight NS300 (Malvern Instruments Ltd., Malvern, U.K.) via nanoparticle tracking analysis. The nanodroplet solution was diluted 1000-fold in degassed and deionized water. Three samples were prepared and three measurements were performed for each sample. Nine measurements in total were performed to obtain the mean values of diameter and concentration of the nanodroplet solution. The precursor OFP microbubbles were sized and counted using the approach detailed in [26]. A diluted microbubble solution was introduced into the counting chamber of the hemocytometer. Images were acquired from a microscope system (Nikon Instruments Inc., Melville, NY, USA) and counting and sizing were performed in a fully automatic and bespoke MATLAB algorithm. Three measurements in total were performed to obtain the mean values of the diameter of the OFP microbubble solution.

C. Experimental Setup
A crossed-tube phantom was fixed and immersed in a water tank. A schematic showing the experimental setup is illustrated in Fig. 1. The crossed-tube phantom was made of two 200-µm-cellulose tubes (Hemophan, Membrana). The tube is a thin-walled cellulose capillary tube of internal diameter of 200 ± 15 µm with a wall thickness of 8 ± 1 µm in the dry state (specifications provided by the manufacturer). An L11-4v transducer equipped with ultrasound research platform (Verasonics Vantage 128, Kirkland, WA, USA) was held 20 mm above the center of the crossed-tubes. OFP droplets were tested as a function of temperature (18°C, 24°C, 30°C, and 37°C) and tested as a function of mechanical index (MI) (0.55, 0.66, and 0.76) at 24°C to determine the optimal experimental conditions. Equal concentrations of 1:9 water-diluted nanodroplet and microbubble solutions (∼ 3.3 × 10 8 agents/mL) were prepared, respectively, in two beakers. The solution was magnetically stirred in the beaker before being transferred into the syringes. The experiments were repeated at different temperatures and MIs.

D. Characterization of Imaging System
The ultrasound imaging system was characterized by measuring the diffraction limited resolution through system point spread function (PSF) and localization precision. A 50-µmdiameter tube was fixed horizontally in a water tank to imitate a point scatterer [5]. This was performed at depths of between 1.5 and 2.5 cm from the transducer surface to test the changes in PSF over the imaging depth. The resolution of the ultrasound imaging system was measured as the mean value of full-width half-maximum (FWHM) over 100 frames in the lateral and axial directions. The localization precision was measured to be the standard deviation of the localized center of mass over 100 frames. A single average PSF was used for the super-resolution processing at all depths.

E. Data Acquisition and Processing
A customized continuous "Activation/Imaging" pulse sequence was developed and implemented on the ultrasound research platform with an L11-4 linear array probe. A sequence of 1-cycle, 3.5-MHz single-angle plane-wave pulses was transmitted at a MI of 0.76 in order to simultaneously activate, deactivate, and image the OFP nanodroplet population. All 128 elements were used to transmit and receive signals. The corresponding spatial and temporal peak negative pressure was 1.42 MPa. The pressure field of the plane wave was calibrated using a 0.2-mm needle hydrophone (Precision Acoustics, Dorchester, U.K.). No transmit apodization was applied on the acquisition sequence. An optical image of the crossed-tube phantom was taken by a charge coupled devices (CCD) camera.
Thousand images were acquired during a 200-ms period at a frame rate of 5000 frames/s for each experiment. The received RF signals were beamformed using delay-andsum (DAS) [27]. Singular value decomposition (SVD) was used to remove background noncontrast agent signal [8]. The SVD thresholds were automatically determined by the point with the largest second derivative of energy with respect to singular value order curves.
All the ultrasound image frames were envelope detected after SVD processing. Super-localization processing was performed on each ultrasound image after setting an image pixel value threshold to reject the noise and detect potential activated or deactivated droplet signals. Each observed PSF was compared with a calibration PSF according to their area ( A), intensity (I ), and shape/eccentricity (E). These parameters were used to discard potential nonmicrobubble echoes and noises. All the observed PSFs with the corresponding three attributes were summarized into three matrices. All the values were normalized in each matrix. A filtering threshold was set in order to discard the larger ( A > 0.46), brighter (I > 0.21), and misshaped (E < 0.66) PSF. The locations of isolated signals were calculated by the "centroid" method [5]. The centroid of each localized signal was computed by calculating the intensity-weighted center of mass. All the localizations from all the imaging frames were assembled into the final super-resolution image.

A. Characterization of OFP Microbubbles and Nanodroplets
The normalized size distribution of the OFP microbubbles was measured using an optical microscopy approach detailed in [26] and the result was illustrated in Fig. 2(a). The mean microbubble diameter was 1.12 ± 0.36 µm and the mean   Fig. 2(b). The mean diameter of the OFP nanodroplets was 153.4 ± 6.9 nm. The concentration of nanodroplets was (3.3 ± 0.2) × 10 9 droplets/mL. Fig. 3(a) shows an example image of the 50-µm wire cross section at a depth of about 2 cm, which is the same depth used for imaging the center of the crossed-tube phantom. The wire acts as a stationary point scatterer and was used to estimate the actual diffraction limited resolution over the depth ranged from 1.5 to 2.5 cm. All the parameters and conditions were  kept the same as used for imaging the crossed-tube phantom. The mean axial and lateral resolution was measured as the FWHM of 664 and 657 µm, respectively, at a depth of about 2 cm. The localization precision was 20 µm, which was the value used for the later localization processing.

C. Localization of Signals
The average number of localization events is 15 per frame. As can be seen from Fig. 4, 14918 localization events were obtained in total and used to generate the super-resolution image at a temperature of 24°C and an MI of 0.76. Fig. 5 shows, over a series of high-pressure plane-wave imaging pulses, the sparse activation of nanodroplets and the destruction of microbubbles signals over time. Fig. 6 shows the superlocalization images obtained using the OFP nanodroplets at four different temperatures (18°C, 24°C, 30°C, and 37°C). Fig. 7 shows the number of localization events obtained at different temperatures. The   Fig. 8 shows the super-localization images obtained using the OFP nanodroplets at three MIs (MI = 0.55, 0.66, and 0.76). Fig. 9 shows the number of localization events at different MIs. The mean and standard deviation values were obtained by three repeated experiments. Fig. 10 represents the summation of 1000 B-mode, SVD-filtered, and super-localization frames acquired without contrast agents, and with microbubble and nanodroplets, respectively, in the nonflow cross-tube phantom. As can be seen in Fig. 10(d), for the control experiment with water, after SVD filtering, no bubble signals were detected. For the microbubble experiment, after starting acquiring data, most of the bubbles were destroyed by the high MI pulses during the first few frames and insufficient localization events were recorded to generate a proper super-resolution image. For the nanodroplets, they can be continuously activated and deactivated by high-frame-rate plane-wave pulses to  form a super-resolution image from 1000 frames acquired over 200 ms.

G. Resolution Measurement
As can be seen from Fig. 11, the corresponding B-mode, SVD-filtered, and super-resolution images were compared with the optical image. Two regions of interest (ROIs) were selected to measure the FWHM of the B-mode, SVD-filtered, and super-resolution images. In the super-resolution image, it can be seen from the line profile shown in Fig. 11(e) that the two tubes are separated by 210 µm ("peak-to-peak" distance), whereas the two tubes cannot be resolved in B-mode or SVDfiltered images. From Fig. 11 IV. DISCUSSION This paper presents an in vitro demonstration of superresolution imaging achieved with the timescale of 200 ms using high-frame-rate plane-wave activation of the OFP nanodroplets, called fast-AWSALM. It takes advantage of the high temporal resolution of high-frame-rate plane-wave imaging and the low activation threshold of the OFP nanodroplets to perform super-resolution imaging orders of magnitude faster than previously demonstrated. Another advantage of having a faster imaging technique as proposed in this work is the ability to reduce the impact of tissue motion. It was demonstrated that tissue motion can be detrimental to ultrasound super-resolution imaging, and motion correction can improve the resolution and fidelity of the super-resolved images [10].
It should be noted that the resolution achieved in this study, 190 µm, is not a limitation of the fast-AWSALM method but a limitation of the phantom available to us (200-µm tube). While the experimental conditions used in this study for comparing droplet-and bubble-based super-resolution approaches are the same, the conditions are not designed for bubblebased approaches. We have previously shown in a similar phantom [3] but with flow and with lower MI that reasonable super-resolution images of the cross-tube phantom can be achieved using microbubbles although it took much longer than the droplet-based approach.

A. Comparison With Existing Localization-Based Super-Resolution Imaging Techniques
The acquisition time of previously demonstrated localization-based super-resolution imaging using microbubbles ranges from 5 s to a few minutes [5], [6], [8], [14].
The AWSALM technique reported previously used separate pulses for imaging and for activation/deactivation of nanodroplets, as the decafluorobutane (DFB) nanodroplets require a relatively high ultrasound amplitude to be activated [18]. Such a high ultrasound amplitude is only achievable at one location at a time through focused transmission. Therefore, it takes time to scan the activating pulses through the whole imaging plane. The previous AWSALM study took 30 s to obtain a super-resolution imaging of the crossed-tube phantom at a frame rate of 100 Hz. However, there should not be a penalty in terms of resolution since the imaging method is the same between AWSALM and fast-AWSALM (i.e., plane wave imaging), and the only difference is that the activation pulse is now also an imaging pulse in fast-AWSALM.
A different formulation of lower boiling-point OFP (boiling point = −37°C) nanodroplets was used in this study compared with the DFB (boiling point = −2°C) nanodroplets used in the previous study. Thus, the OFP nanodroplets could be activated by lower MI plane-wave pulses compared with DFB nanodroplets. The focus-wave pulses are not required in this study as the acoustic pressure provided by plane-wave imaging pulses is sufficiently high to activate the OFP nanodroplets. Therefore, the activation of the droplets can be achieved within the imaging plane by a single plane wave, and such activation can then be imaged by the very same activation plane wave. As such plane waves can be transmitted at thousands pulses per second, a significant increase in imaging frame in superresolution can be achieved. In this case, an imaging acquisition rate of 5000 frames/s frame rate was used, enabling superresolved image data acquisition in 200 ms, which is orders of magnitude faster than the previous localization-based superresolution techniques [8].

B. Sparse Activation and Destruction
The principle of fast-AWSALM is that the first plane-wave imaging pulse will sparsely activate a subgroup of the OFP droplets as the acoustic pressure is selected to only activate the largest most easily activated droplets. These activated droplets will generate acoustic signals until the next imaging pulse. The following imaging pulses will serve three purposes: first to generate images, second to destroy most of the existing microbubbles that were activated by the previous pulse, and third to activate a new subgroup of droplets. This continuous imaging, activation, and destruction can provide different localizations of multiple subgroups of activated droplets within the vessels.
The sparse and stochastic activation of the OFP droplets and the destruction of the OFP microbubbles may be due to a number of factors. A key factor is that the droplet population consists of droplets with a statistical distribution of different sizes and shell properties. The size distribution of nanodroplet can be seen from Fig. 2(b): the nanodroplets are polydisperse and their diameters are from about 50 to 350 nm. The droplet size, gas content, and shell properties are important factors determining the acoustic pressure threshold required for activating the droplets. The activation threshold of nanodroplets has been studied by many researchers and there are several theories that can explain these phenomena under different circumstances [28], [29]. The larger droplets can be acoustically vaporized more easily at a relatively low pressure as the activation threshold is inversely proportional to the diameter of the droplet. Therefore, only a sparse subgroup of relatively large droplets randomly spatially distributed in the tube is sensitive to the low activation pressure. The subsequent imaging pulse will cause a large percentage of activated nanodroplets to be destroyed, including through shell rupture, diffusion, and dissolution [30]. The activation of nanodroplets by the next pulse may be affected by a number of factors. First, these nanodroplets may be affected by any remaining nearby microbubbles generated by the previous pulse-the wave emitted from a generated bubble could be directional and have a random direction, and therefore, a nanodroplet experiencing this wave in addition to the second pulse is likely to be activated more easily. Second, they may be affected by any increase in local thermal energy or acoustic power due to the imaging pulses, as the nanodroplets have a lower activation threshold in total when temperature or pressure increases [31]. Moreover, the nonlinear distorted wave from nonlinear propagation and from generated bubbles may also have superharmonics with a wavelength close to the diameter of the droplet; therefore, contributing to the initiation of droplet activation by the next pulse [29]. Further studies are required to determine which of these effects dominates and if other mechanisms are involved. Fig. 12 shows that when the high-frame-rate plane-wave imaging is on, the contrast signal from microbubbles within the cross-tube decreases during the first few milliseconds due to the high acoustic pressure (MI = 0.76). The normalized contrast signal from the water experiment did not change with time. For the contrast signal of nanodroplets within the crossed-tube, it gradually increased and reached plateau after about 100 ms. One possible reason for this increase may be due to the accumulation of activated droplets that are not completely disrupted by the plane waves. Another  possible reason may be that only a sparse subgroup of the OFP nanodroplets was activated by the first few plane-wave pulses. The subsequent plane-wave pulses may simultaneously deactivate the existing activated droplets and activate the unactivated droplets. It has been found that any remaining nearby activated nanodroplets generated by the previous activation may lower the activation threshold for the next activation [32]. Therefore, the nanodroplets experiencing the later plane-wave pulses will be likely to be activated more easily, thus more contrast signals can be seen at the later stage.

D. Effect of Temperature
The experiments were performed at four different temperatures ranging from 18°C to 37°C. As can be seen from Fig. 7, the number of localization events decreases when the temperature is 18°C or 37°C. The optimum temperature is likely to be around 24°C according to the suggested results. A relatively higher temperature would lead to a lower acti-vation threshold of nanodroplets, thus increasing the number of localization events. However, a higher temperature will also lead to more spontaneous vaporization of nanodroplets prior to acoustic activation. Therefore, there is a compromise between the ease of activation and the stability of nanodroplets.
The experiments were performed at 24°C in order to achieve relatively optimal performance. The OFP nanodroplets can still be activated at 37°C, as shown in Fig. 9, although fewer localization events can be obtained. In an in vivo situation of 37°C, a significant portion of the OFP nanodroplets are likely to spontaneously vaporize. However, it is still feasible to use the OFP nanodroplets at 37°C and perform in vivo ultrasound imaging since this has been demonstrated by a previous in vivo study [33]. Spontaneously vaporized OFP nanodroplets would not be an issue in fast-AWSALM as they either get deactivated by the pulses or contribute to the signal if they are not destroyed.

E. Effect of Mechanical Index
The experiments were performed using plane-wave pulses at three different MIs (MI = 0.55, 0.66, and 0.76). The results suggested that, at the highest MI, the largest number of total localization events can be obtained. The input voltage was set at 45 V in order to achieve MI of 0.76 at an imaging frequency of 3.5 MHz according to the hydrophone calibration. The transducer used needs to be operated below 50 V in order to avoid damage. Therefore, an MI of 0.76 was chosen for the experiment as the maximum which could be achieved practically. The amplitude can be reduced further through the optimization of the droplet chemistry and ultrasound pulse sequence. The depth of interest in this study is about 2-3 cm; therefore, how the technique work in deeper tissues will need to be studied further.

F. Potential Applications
This study takes advantage of the high temporal resolution of high-frame-rate imaging and the low activation threshold of the OFP nanodroplets to perform faster super-resolution imaging than previously demonstrated. The homemade OFP nanodroplets solution has the same composition as the commercial Definity©microbubble contrast agents [34]. This means that the condensation of Definity©microbubble contrast agents to nanodroplet solution can be used with this technique for preclinical in vivo experiments in the future and could facilitate the clinical translation of super-resolution ultrasound imaging.

V. CONCLUSION
In summary, fast-AWSALM was developed to generate super-resolution ultrasound images in hundreds of milliseconds, orders of magnitude faster than some existing ultrasound localization microscopy techniques. We demonstrate that the high-frame-rate plane-wave pulses can be used to continuously activate and destroy the OFP nanodroplets while imaging on the timescale of milliseconds to super-resolve microstructures that cannot be resolved in the conventional B-mode image, without the requirement of flow. The much smaller size of the nanodroplets compared to microbubble contrast agents means the technique can potentially super-resolve structures beyond the reach of the current microbubble-based ultrasound super-resolution approaches.