Characterization of Contrast Agent Microbubbles for Ultrasound Imaging and Therapy Research

The high efficiency with which gas microbubbles can scatter ultrasound compared with the surrounding blood pool or tissues has led to their widespread employment as contrast agents in ultrasound imaging. In recent years, their applications have been extended to include super-resolution imaging and the stimulation of localized bio-effects for therapy. The growing exploitation of contrast agents in ultrasound and in particular these recent developments have amplified the need to characterize and fully understand microbubble behavior. The aim in doing so is to more fully exploit their utility for both diagnostic imaging and potential future therapeutic applications. This paper presents the key characteristics of microbubbles that determine their efficacy in diagnostic and therapeutic applications and the corresponding techniques for their measurement. In each case, we have presented information regarding the methods available and their respective strengths and limitations, with the aim of presenting information relevant to the selection of appropriate characterization methods. First, we examine methods for determining the physical properties of microbubble suspensions and then techniques for acoustic characterization of both suspensions and single microbubbles. The next section covers characterization of microbubbles as therapeutic agents, including as drug carriers for which detailed understanding of their surface characteristics and drug loading capacity is required. Finally, we discuss the attempts that have been made to allow comparison across the methods employed by various groups to characterize and describe their microbubble suspensions and promote wider discussion and comparison of microbubble behavior.

TUFFC-07749-2016 1 Abstract-The high efficiency with which gas microbubbles can scatter ultrasound compared to the surrounding blood pool or tissues has led to their widespread employment as contrast agents in ultrasound imaging. In recent years their applications have been extended to include super-resolution imaging and the stimulation of localized bio-effects for therapy. The growing exploitation of contrast agents in ultrasound, and in particular these recent developments, have amplified the need to characterize and fully understand microbubble behavior. The aim in doing so is to more fully exploit their utility for both diagnostic imaging and potential future therapeutic applications.
This paper presents the key characteristics of microbubbles that determine their efficacy in diagnostic and therapeutic applications and the corresponding techniques for their measurement. In each case we have presented information regarding the methods available and their respective strengths and limitations, with the aim of presenting information relevant to the selection of appropriate characterization methods. First we examine methods for determining the physical properties of microbubble suspensions and then techniques for acoustic characterization of both suspensions and single microbubbles. The next section covers characterization of microbubbles as therapeutic agents, including as drug carriers for which, detailed understanding of their surface characteristics and drug loading capacity is required. Finally we discuss the attempts which have been made to allow comparison across the methods employed by various groups to characterize and describe their microbubble suspensions and promote wider discussion and comparison of microbubble behavior.
Index Terms-Ultrasound contrast agents, Therapeutics, Medical Imaging, Acoustic microscopy and imaging.

I. INTRODUCTION
Gas microbubbles are now in widespread use as contrast agents in ultrasound imaging as a result of their exceptional echogenicity [1,2]. Over the past 2 decades, their ability to produce localized bio-effects for therapeutic applications particularly for the treatment of cancer [3], drug and gene delivery [2], opening of the blood brain barrier [4,5] and to seed ablation of cancerous tissue during high intensity focused ultrasound [6] has also been widely investigated. In addition, their use in super-resolution imaging techniques, which offer unprecedented detail in ultrasound imaging is currently being explored [7,8]. This rapid growth in exploitation has greatly increased the need for better understanding of microbubble behavior and hence for better characterization methods. Ultrasound contrast agents are made up of gas microbubbles, commonly a biologically inert fluorocarbon, stabilized by a lipid, protein or polymer shell. Several commercial agents exist and these are supplied either as a polydisperse suspension of pre-formed microbubbles, or lyophilized powder that is mixed with saline and agitated by the clinician prior to use. The suspensions typically contain microbubbles ranging in size from 1-10 µm in diameter corresponding approximately to the size of a red blood cell (~ 8 µm). This is important if microbubbles are to remain within the vasculature but able to traverse capillaries. It also mitigates the risk of embolism presented by adding gas bubbles to the blood stream. In addition, particularly for the purposes of research, contrast agents can be fabricated in-house by means of sonication or microfluidics, with the specific characteristics of the resulting suspensions varying according to the fabrication technique and materials employed.
The aim of this paper is to present the key characteristics of microbubbles that determine their efficacy in diagnostic and therapeutic applications and the corresponding techniques for their measurement. In each case we have presented information regarding the methods available and their respective strengths and limitations, with the aim of presenting information relevant to the selection of appropriate characterization methods. First we examine methods for determining the physical properties of microbubble suspensions; second, techniques for acoustic characterization, both of microbubble suspensions and single microbubbles; third, we review techniques appropriate to therapeutic

Characterization of Contrast Agent
Microbubbles for Ultrasound Imaging and Therapy Research applications, for which detailed understanding of microbubble surface characteristics and drug loading capacity is required. Finally we discuss the methods for comparing data sets from across the literature.

A. Introduction
The extent to which microbubbles are able to scatter sound or generate bio-effects is a function of their oscillatory behavior when subject to the alternating pressure of an ultrasound field. Upon exposure to ultrasound, the highly compressible gas core of a microbubble causes it to undergo a volumetric contraction and expansion. There is a fortunate correlation between the range of microbubble sizes that is appropriate for clinical use and that which is resonant, i.e. exhibits a maximum amplitude of expansion, at the range of frequencies commonly employed in diagnostic ultrasound imaging (1)(2)(3)(4)(5)(6)(7)(8)(9)(10)(11)(12)(13)(14)(15), as illustrated in Figure 1. Thus the distribution of sizes in a microbubble suspension is an important factor in predicting both their individual fundamental response [9] and in predicting the response of the population [10,11]. Fig. 1. Undamped linear natural frequency of an air-filled microbubble as a function of diameter calculated using Minnaert's equation [215].
Due to the range of microbubble preparations available and their variability, it is essential that researchers can easily generate detailed size and concentration information for the particular suspension they intend to use. There are several methods available for the sizing and counting of microbubbles, including: electro-impedance volumetric zone sensing, typically by means of the commercial device known as a Coulter counter or Multisizer (Beckman Coulter (UK) Ltd., High Wycombe, UK); laser diffraction, commonly undertaken using a Mastersizer (Malvern Instruments, Worcestershire, UK), and optical microscopy and singleparticle optical light scattering. Each method can deliver size distribution and/or concentration measurements for suspensions of microbubbles and has particular merits and limitations. For example, commercial devices offer automation and provide swift analysis, however, the cost of investing in and operating specialist equipment may be limiting. In the case of laser diffraction, accuracy can be impaired by the reliability of the scattering model used as a basis for sizing the microbubbles due to variations in microbubble shape, the fluid they are suspended in and accuracy with which the refractive index of the fluids involved can be estimated. Microscopy, which is more time consuming than the alternatives, offers excellent microbubble sizing accuracy. It also allows the analysis to be easily customized to deliver information about the size ranges of particular interest to ultrasound research. For example, sizing metrics arising from laser diffraction can be skewed by the high sensitivity of the technique to the sensing of nano to microscale bubbles. Since bubbles of these sizes do not contribute significantly to the echogenicity of a microbubble suspension at typical clinically applied frequencies [9], inclusion in the analysis can result in a misleading set of suspension metrics. A more detailed comparison and discussion of the merits of electroimpedance, laser diffraction and optical microscopy can be found in [12]. This discussion has been extended by Satinover et al. [13], who compared single particle light scattering sizing methods such as those employed in the AccuSizer (Particle Sizing Systems, Santa Barbara, CA, USA) with electrozone sensing. The authors use the generalized Lorenz-Mie Scattering theory to explain the resulting differences in measurement recorded, revealing that discrepancies in sizing occur due to the non-linearity of the relationship between the extinction cross section of light scattering from a particle and the diameter curve. They conclude that flow cytometry can provide superior sizing information as compared to light obscuration methods because in the former both the forward and side scattering measurements are used for each particle to calculate diameter.

B. Optical microscopy for microbubble suspension characterization
Optical microscopy offers a readily accessible, low cost and simple means of accurately sizing microbubble suspensions. The main steps in sizing a suspension in this way are illustrated in Figure 2. The suspension is prepared, following either the manufacturer's instructions for commercial products or at the concentration it will be used at for in-house fabricated suspensions, and gently inverted to re-suspend the bubbles. A known volume of the suspension is added to 1 ml of saline and 10 µl of this diluted suspension is added to the counting chamber of a haemocytometer using a 20 µl pipette (P20). The haemocytometer is placed on a microscope stage and bubbles are allowed to rise under buoyancy to achieve a static state. Using a 40x objective the concentration of the microbubble suspension is checked to confirm its suitability for microscopy counting. The recommendation is that there are at least 50 isolated microbubbles in the imaging frame. Should this not be the case, the dilution step should be repeated, increasing or decreasing the volume of microbubble suspension added to the saline as appropriate. A series of individual images sufficient to include at least 5000 individual microbubbles are acquired from the haemocytometer, as representative of the suspension as a whole, and analyzed using MATLAB (The Mathworks Inc., Natick, MA, USA) to produce sizing and concentration data as illustrated in Figure  Fig  . Microbubble sizing using microscopy: (a) a typical bright field optical microscope image of an albumin microbubble suspension; (b) a histogram illustrating the size distribution as percentage number of the same suspension obtained using the MATLAB based statistical analysis program described in [14].

C. Optical scattering methods
Laser diffraction (LD) is an alternative method for sizing and counting microbubbles. In an LD analyzer (e.g. Mastersizer from Malvern Instruments, Worcestershire, UK), diluted particles are exposed to a laser beam and the angular light distribution and intensity scattered from the particles is measured. Particles crossing the laser beam interact with the incident beam through scattering and absorption to generate flux intensities at angles inversely proportional to the size of the individual particles. A theoretical model based on Mie or Fraunhofer scattering theories can be used to determine the particle size by fitting of the measured data. The Mie scattering model is in general more accurate for small particles, particularly those less than tens of micrometers down to tens of nanometers, but requires knowledge of the refractive indices of the particles and the medium. The refractive index of microbubbles can be determined experimentally e.g. on a polarizing microscope using the Becke line method [15]. There are variations of the Mie model, in the Mastersizer control software, and for microbubbles a model that does not ignore particle absorption is recommended. Care should be taken to use appropriate concentrations in order to minimize multiple scattering at high particle densities. A study of the different sizing and counting methods [12] has shown some relatively large errors in the LD results comparing to the optical microscope or the electroimpedance volumetric zone sensing. The errors are likely due to the inaccuracy in the estimation of the complex refractive indices of the microbubbles.

D. Electro-impedance volumetric zone sensing
Electro-impedance volumetric zone sensing "Coulter" counters and similar instruments have been widely applied to counting and sizing particles by measuring changes in the electrical impedance of a sensing zone as individual particles pass through it. In a typical set-up, test particles are diluted in a conductive liquid, homogenized by stirring, and pumped through the sensing zone across which electrodes are placed. The count of changes in impedance measurements corresponds to bubble count, while the amplitude of the measurements, after calibration, corresponds to the size of the particles. A recent study has shown that a Coulter counter can generate highly reproducible results in sizing and counting microbubbles [12]. A typical Coulter counter system requires differently sized apertures to capture broad particle size distributions, as the particle size range that can be determined using a given aperture is of the order 2% to 40 % of the aperture [16], outside of which errors could be significant. While coincident passage of two or more particles through the sensing zone is a potential source of error, this can be overcome by employing a pulse discrimination system that rejects distorted pulses, which is a common feature of modern commercial systems. However this could still potentially affect the counting results.

E. Important points to note
It is important to note that in order to produce data that are relevant for in vivo microbubble applications the potential differences in environmental conditions capable of influencing microbubble stability, size and concentration should be carefully considered. It has been shown that both the temperature and gas content of the suspending fluid can influence microbubble stability and diameter [17][18][19]. It is also widely accepted that microbubbles are particularly vulnerable to handling and rapid changes in pressure which may be encountered during transfer between solutions. Barrack and Stride and Browning et al. both showed that the shear stresses and pressure gradients exerted by drawing suspensions through a syringe and needle can lead to microbubble destruction [20,21]. In our labs we have observed that microbubble stability can be undermined for commercial contrast agents if the septum seal is removed, exposing the suspension to the atmosphere rather than the perfluorobutane or other gas included in the head space of the vial. Each of these factors should be carefully considered both in the preparation and handling of microbubble suspensions and during examination and consideration of the results. The relatively recent development of phase change contrast agents has offered exciting opportunities for extending the use of ultrasound contrast agents to the extravascular space [22]. Existing micrometer-sized bubbles are purely blood pool agents as they are too big to cross the endothelial wall (extravasation). While this is an advantage for vascular imaging, they are incapable of interrogating or delivering therapy within the interstitial space e.g. in solid tumors. Sub micrometer bubbles are difficult to make and are not efficient in providing image contrast. Phase change contrast agents are typically perflurocarbon droplets which can be prepared with nanoscale dimensions to facilitate their extravasation, after which they can be vaporized to for microbubbles using ultrasound (or other sources of energy). The characterization of such nano droplets however, is more challenging due to their size. While a typical optical microscope can be used to visualize some sub micrometer droplets, it will not have sufficient resolution to accurately size them. Dynamic Light Scattering (DLS) is commonly used for sizing and counting such small particles. Detailed characterization of their shape and surface structure can also be achieved, for example through the use of transmission electron microscopy (TEM). The development of higher frequency contrast imaging, requiring the use of smaller diameter microbubbles, presents similar challenges for characterization.

A. Bulk suspensions
Bulk acoustic measurements provide information about the scattering and attenuation characteristics of a microbubble suspension and are generally undertaken to provide information which will allow acoustic performance in vivo to be anticipated. A known quantity of microbubble suspension is added to a vessel of water, itself positioned within a larger volume of water for ease of acoustic coupling. The target vessel is positioned coincident with the focus of an ultrasound transducer such that the sound can propagate through an acoustic window and through the microbubble suspension within the vessel. The suspension is gently agitated, typically using a magnetic stirrer, to ensure that the interrogation volume is regularly replenished with a fresh supply of suspended microbubbles in between measurements. Acoustic signals are then collected using either a matched transducer positioned at 90° from the excitation source for acoustic scattering, or a hydrophone positioned in line with the excitation source for acoustic attenuation, and acquired to a PC for post-processing. Typical arrangements used to perform such measurements are shown in Figure 4 and a more detailed explanation can be found in [23]. Comparable measurements can also be performed using a clinical imaging system [24]. In each case it is important to ensure that the characteristics of the acoustic set up are removed from the data and where a clinical imaging system is used, that data can be accessed independent of proprietary post-processing, or alternatively that steps are made to mitigate its effect, for example by comparing data acquired under identical conditions. Fig. 4. Schematics showing typical arrangements for measurement of (a) acoustic attenuation and (b) acoustic scattering. Where attenuation is calculated in decibels against a reference measurement taken with no microbubble suspension in the propagation path and the received signal due to scattering is calibrated against the known pressure response of the receive transducer to derive scattered pressure in kPa.

B. Methods of standardizing results for comparison across research groups
Despite much attention having been given to the issues discussed above, ease of comparison of scattering and attenuation data across research groups can be challenging. This is due in part to the fact that no universally agreed system of comparison exists and in part to the individual environmental and set-up conditions employed by different groups, as discussed in section 1. In terms of comparison, a metric which combines both scattering and attenuation measurements (linear and/or non-linear ) has been proposed in the form of an index termed STAR: Scattering to Attenuation Ratio or nSTAR [25]. This index was established to describe the extent to which a contrast agent can enhance the scattered acoustic signal from a tissue perfused with contrast agent, while also presenting information about the degree to which the agent contributes to ultrasound absorption. However, the measure is equally relevant to therapeutic ultrasound in which echogenicity and treatment depth are of similar importance. Several groups have used this approach to make comparisons across microbubble formulations [26,27]; however it is not clear why it has not been more widely adopted.

C. Single microbubble techniques
One drawback of bulk acoustic measurements is that they do not provide any information regarding the variation in acoustic response across a microbubble population. At a given ultrasound frequency and pressure, both the amplitude and nonlinear characteristics of the oscillation of an individual microbubble are known to be highly dependent upon its size and coating properties. In order to address this, several different techniques have been developed using: (i) ultra-high speed video microscopy, which will be addressed in section 4, (ii) acoustic scattering measurements and (iii) laser scattering measurements.

1) Acoustic scattering a) Isolated capillary measurements
One of the simplest methods of measuring individual microbubble responses is to perform scattering measurements similar to those described in section 3.1, but with a highly dilute suspension of microbubbles, such that only a single bubble is present in the focal volume of the transducer at any one time. In principle, the same apparatus as described in section 3.1 could be used, but in practice some means of aligning bubbles with the transducer focus is desirable. One way to achieve this is to confine the microbubbles within an acoustically transparent tube whose diameter is large compared to the microbubble but comparable or smaller than the acoustic wavelength. For example, Casey et al. [28] utilized 200 µm diameter cellulose dialysis tubing for this purpose. A highly dilute suspension of microbubbles is drawn into the tube by capillary action or under mild suction from a syringe pump. The capillary is suspended in a water bath at the focus of a water immersion microscope objective lens. Suitable positions at which isolated bubbles are located can then be identified. Ideally the objective and ultrasound transducer should be mounted on a positioning rig such that their foci coincide. Thus, once a suitable microbubble has been identified a scattering measurement can be made.
There are a number of drawbacks to this technique, however. First, it is relatively labor intensive and the rate at which measurements can be made is quite low. This is undesirable as it limits effective statistical analysis of microbubble population characteristics. Second, confining the microbubbles in the capillary will alter their response as compared to that exhibited by an unconfined microbubble [29]. Thus the measured responses will differ from those that might be expected in large blood vessels or predicted by theoretical models. Third, it is difficult to measure the sound field to which a microbubble will be exposed with the microscope objective in situ. Thus, either there will be some uncertainty in the exposure parameters due to scattering from the objective or the objective must be retracted immediately before the experiment at the risk of agitating the fluid and causing the bubble to move.

b) Flow focusing
An alternative means of confining microbubbles that avoids the need for a solid boundary is to create a hydrodynamically focused stream of microbubbles whose axis can be aligned with the transducer focal volume. This is achieved by using a pair of coaxially aligned nozzles whose flow rates can be controlled individually e.g. through separate syringe pumps. The nozzles are immersed in the liquid filled chamber in which the scattering measurements are to be made. The inner nozzle carries the bubble suspension and the outer nozzle the so-called "sheath" flow that confines the bubbles to a fine stream. By adding a small amount of dye to the bubble suspension the ultrasound transducer can be optically aligned to this stream. Roos [30] and subsequently Roy and Apfel [31] demonstrated the efficacy of this technique for the acoustic characterization of red blood cells. It was subsequently applied to microbubbles by Everbach et al. [32] and Sboros et al. [33]. It is still necessary to use a very dilute suspension of microbubbles to minimize the risk of multiple bubbles passing through the ultrasound focus simultaneously, but a significantly higher rate of scattering measurements can be achieved with this technique. The main drawback of this method is the lack of feedback on microbubble position and hence the precise pressure amplitude with which they are excited. Acoustic pulses are transmitted at a set repetition frequency but there is no means of detecting where in the pressure field a microbubble is located when a scattering measurement is recorded. Given the highly pressure dependent nature of microbubble scattering (unlike that of solid or liquid particles), this introduces considerable uncertainty into the measurements. In addition it is not possible to determine the initial size of each microbubble.

c) Acoustic 'camera'
A more recent approach, published by Renaud et al. [34] utilizes similar apparatus to that described in section 3.1, but with three co-focally aligned ultrasound transducers. The first transducer is used to "drive" the microbubbles in the range 1-10 MHz whilst the second and third transducers emit and receive pulses respectively at a much higher frequency (30 MHz) in order to "image" the microbubbles. The imaging frequency is sufficiently above the microbubble resonance frequency for them to behave as linear scatters and thus the recorded pressure is directly proportional to the bubble volume. The "camera" thus provides data that are more directly comparable with those obtained with ultra-high speed imaging and can be more readily compared with theoretical predictions but with less stringent sampling frequency restrictions. In its current form, the system suffers from the same limitations as the flow focusing systems, namely uncertainty regarding initial bubble radius and acoustic excitation pressure amplitude. Based on this acoustical measurement setup, a combined optical sizing and acoustical characterization technique was later developed [35]. A coaxial flow device was used to direct individual microbubbles to the confocal region of the ultrasound transducers and a high-resolution microscope objective. Bright-field images of microbubbles flowing through the optical field of view were captured by a high-speed imaging camera synchronous with the acoustical recordings.

2) Optical scattering
An alternative means of measuring microbubble volume is to exploit the fact that an oscillating microbubble will scatter light as well as sound. The principle is similar to that of the acoustic camera, except that the two high frequency transducers are replaced by a light source, typically a laser, and an optical detector, such as a photomultiplier tube (PMT) or photodiode. Mie scattering theory provides a direct means of determining microbubble volume from the measured optical signal.
Optical scattering measurements offer a few advantages over purely acoustic methods, albeit at the cost of additional system complexity. At low powers, a laser beam will produce negligible heating or radiation force on a microbubble. Thus, unlike a high frequency ultrasound beam, it can be continuously transmitted throughout an experiment. This means first, that the laser can be used as means of determining the exact position of a microbubble and triggering the acoustic excitation; and second, as means of directly measuring microbubble volume before, during and after that excitation. Guan and Matula and subsequently Tu et al. [36] demonstrated this technique using a modified flow cytometer for the characterization of the commercial contrast agent SonoVue®. One limitation of their experiment however was that the bubbles were confined by a rigid surface during the measurements and there was considerable uncertainty regarding the characterization of the sound field. More recently, Rademeyer et al. [37] utilized a similar technique but confined the bubbles using a hydrodynamically focused flow, thereby removing the effects of a rigid boundary and enabling accurate characterization of the acoustic field. They were able to record the volumetric responses of 12000 bubbles in 30 minutes including their initial and final sizes. All of the techniques discussed in this section are limited by the maximum signal to noise ratio that can be achieved in a specific set up and hence in terms of the minimum bubble size and/or amplitude of oscillation that can be detected. A further limitation is the inability to distinguish between spherical and non-spherical oscillations. This requires the use of direct imaging methods discussed in the next section.

AGENTS
Optical characterization methods rely on the acquisition of nanosecond exposure optical images of the dynamic response of microbubbles to ultrasound excitation [38,39]. In contrast to the acoustic characterization techniques described in Section 3, optical approaches are able to resolve single microbubbles, as the size of a microbubble is larger than the wavelength of light. Optical measurements based on highspeed imaging techniques can directly visualize the vibrations of individual microbubbles in response to an ultrasound field, as well as their interactions. The physical properties can be inferred by varying the applied ultrasound parameters, and acquiring the radius versus time (R-t) response. Moreover, various interesting phenomenon (e.g., microbubble destruction and lipid shedding) specific to the applied acoustic field can be observed.

A. Ultra-high speed imaging facilities
To resolve the transient dynamics of microbubbles activated by ultrasound at Megahertz frequencies, the sampling rate should exceed the central driving frequency by at least a factor of 8-10 [38]. Therefore a frame rate of at least several millions of frames per second (fps) is required. In pioneering onedimensional optical studies of microbubbles, streak imaging and strobe imaging techniques were used [40,41]. Subsequently, 2D ultra-high speed photography was developed that meets both the temporal and the spatial requirements for two dimensional visualization of microbubble dynamics [42].

1) Streak imaging
Streak imaging can continuously record an event over time through a slit. For capturing microbubble dynamics, it continuously acquires a single line through the center of a bubble [40]. These one dimensional images can be recorded with a nanosecond or picosecond time interval between frames. Alternatively, strobe imaging records microbubble dynamics by composing images obtained repetitively using a time-delayed, single-shot method [41,43]. However this is not a real-time imaging method, and is restricted to repeatable events.
2) Ultra-high speed photography One approach to ultra-high speed imaging of microbubble contrast agents is to use a shutter-controlled camera such as the Imacon 200 (DRS Hadland, Cupertino, CA, USA). In this approach a beam splitter is used to divide an image over several recording devices, achieving a maximum recording speed of 200 million fps [44]. However, the number of maximum total recording frames available is very limited (e.g., 16 frames for Imacon). The Brandaris 128 ultra-high speed imaging camera was introduced in 2003 by Chen et al., [39,45] and specifically designed with microbubble research in mind. The primary image acquired is projected onto the rotating mirror, which is mounted on a helium-driven turbine. The mirror can redirect the light beam and sweep across the lens bank and refocus the image onto 128 CCD sensors. This optical design was also employed for the development of another high-speed rotating mirror system (UPMC Cam) [46], with further improvements in the sensitivity, spatial resolution for bright field and fluorescence imaging of microbubble dynamics. The Brandaris 128 camera exceeded earlier generations of high-speed imaging cameras in its extremely high frame rate (up to 25×10 6 fps) and relatively large number of recording frames (128). The camera can be coupled with various optical setups (e.g., a microscope lens) to suit the experimental conditions. In a typical experiment, the contrast agent microbubbles are injected into a capillary tube or a chamber which are both optically and acoustically transparent, and mounted in a water tank. The water tank is positioned below a microscope connected to the Brandaris 128 camera, and is configured with an ultrasound transducer pointing towards the optical field of view. For each individual bubble, the camera is able to record 6 movies of 128 frames, or 12 movies of 64 frames in a single run, which can be accomplished within a few seconds [38]. Figure 5 shows the setup of the Brandaris 128 camera.

B. Optical Characterization techniques
Based on the ultra-high speed imaging camera system, many novel optical techniques have been developed to study the acoustic behavior of ultrasound contrast agent microbubbles. Typically, the radial dynamic response is recorded at relatively low acoustic pressures (tens of kPa), and the radius-time curves extracted to further explore the physical properties of the agent. In the same low pressure regime, various interesting physical phenomena have been observed, e.g., thresholding behavior [47], "compression-only" response [48] and surface mode vibrations [49]. At higher pressures (hundreds of kPa), the violent vibrations of microbubbles tend to disrupt the shell and are accompanied by shedding of lipid shell materials [35,[50][51][52][53][54]. Similar optical studies have also been made of other varieties of contrast agents, e.g., nanodroplets and echogenic liposomes (ELIP) [55,56].

1) Low acoustic pressure regime
The volume pulsation of a microbubble in response to an external oscillating pressure field is a function of multiple factors, e.g., the pressure amplitude, the frequency and the physical nature of the bubble itself. As above, most clinical ultrasound contrast agent (UCA) bubbles are encapsulated in order to counter surface tension and stabilize the gas core. The presence of this outer shell can dramatically influence the microbubble acoustic behavior. Theoretical modelling combined with optical measurement techniques have been developed to investigate the bubble dynamics at the low pressure regime, and physical properties of the bubble shell were obtained. The first model was proposed by de Jong et al. [57], which incorporated experimentally determined elasticity and viscosity parameters into the Rayleigh-Plesset equation [58]. A more recent model was developed by Marmottant et al. [59] which considers a nonlinear effective surface tension at the bubble wall. Van de Meer et al. [60] employed a technique to characterize the shell properties, by sweeping through the driving frequency at a very low pressure (<50 kPa) in a single run of the Brandaris camera. The radius/diameter versus time curves were extracted from the videos and analyzed in the frequency domain, and resonance curves specific to single microbubbles were reconstructed. By comparing the experimental results with the simulated response for amounts of microbubbles, the elasticity and viscosity of encapsulating shell can be derived. This so-called "microbubble spectroscopy" was applied in series of studies and showed a damping effect and a shift of the resonance frequency due to the presence of the shell [61][62][63][64]. Figure 6 presents the setup of Brandaris 128 imaging camera and an example of a recording in a "microbubble spectroscopy" experiment. Examples of ultra-high speed image frames recorded for the same microbubble insonified at 1.1 MHz are shown by figure x+1 (corresponding to the third panel in Figure 7).  2) High acoustic pressure regime Ultrasound contrast agent microbubbles have also shown great potential for therapeutic use, e.g., thrombolysis [65], gene/drug delivery [66] and reversible opening of the blood brain barrier (BBB) [67]. Relatively high acoustic pressures (on the scale of several hundred kilo Pascal to mega Pascals) and long ultrasound bursts (up to several thousands of cycles) are typically used. In this pressure regime, contrast agent microbubbles can undergo dramatic changes in physical properties and vibration dynamics. Therefore to characterize the microbubble response optically is of high clinical relevance.

a) Observations of microbubble destruction
Previous optical studies have revealed the deflation of microbubbles under relatively high acoustic pressures [51,[68][69][70], or when exposed to prolonged burst ultrasound excitations. Microbubble deflation has been ascribed to a few different physical phenomena. One is the relatively slow gas convective diffusion driven by the ultrasound exposure, which can happen within hundreds of milliseconds up to a few minutes [71]. Another process is the inertial collapse and rapid destruction of microbubbles into smaller bubbles or fragments, which is then released into the surrounding medium. The latter process can happen transiently (within a few ultrasound cycles) during violent bubble vibrations [72]. Studies based on ultra-high speed imaging technique have shown changes in microbubble dynamic responses relative to the immediate size reduction of the microbubble [51,70,73].
To investigate the microbubble dynamics in response to long ultrasound pulses, a new imaging mode on the Brandaris 128 camera has been developed to extend the total time span of acquisitions [74]. The 128 CCD sensors were grouped into separate segments, with the inter-segment time set to be as low as multiples of the rotation period of the turbine. Successive imaging segments were distributed over the time window of the ultrasound application. In a recent in vivo study, the segmented imaging mode was utilized to visualize the microbubble response in a chicken embryo model under a prolonged ultrasound burst lasting for over 1 ms [75].

b) Observations of lipid shedding behavior
There has been wide interest in gaining knowledge regarding lipid shedding behavior under the exposure to ultrasound [35,54,71,76]. The loss of lipid shell material has been found to be closely related to the accompanying microbubble dissolution [51,77].. Utilizing a high-speed imaging camera Photron (operated at 50-100 kfps), lipid shedding dynamics from fluorescently labelled microbubbles under ultrasound excitations can be clearly visualized [52]. A conventional high speed camera can also be combined with the Brandaris 128 camera to monitor the lipid shedding behavior relative to the microbubble vibration dynamics [35]. In an experiment in which both the top view and the side view were captured, two separate physical processes were observed: the detachment of lipid particles from the microbubble shell, and the subsequent transport of shed materials by the convective streaming flow field. The detachment of lipids was observed to occur within a few ultrasound cycles above a certain pressure threshold, while the transport trajectory was found to follow the streaming flow field and is perpendicular to the nearby cell membrane. Figure 8 shows the shedding of lipid shell materials, and the corresponding microbubble vibration dynamics and/or destruction recorded by the Brandaris 128 camera. Fig. 8. (a-c). Examples of lipid shell material shedding shown by high-speed fluorescence imaging, recorded after the ultrasound application. Ultra-high speed images recorded by the Brandaris128 camera indicate the vibration dynamics and the destruction of microbubbles. A single burst of 100 cycles at above 100 kPa (128 kPa for a and c, 170 kPa for b) was applied. Scale bars represent 10 µm.

3) Limitations and future prospects
The frame rate in combination with the recording (exposure) time is in general a limiting factor for high speed imaging facilities. For example, for the ultra-high speed imaging camera Brandaris 128, the total exposure time of each recording is restricted to 12 µs at a frame rate of 10 Mfps. Although the segmented mode was developed to counter this disadvantage, the inter-segment gap inhibits a full acquisition of the bubble dynamics during the entire ultrasound exposure. Moreover, the optical setup based on a high-speed imaging camera was relatively difficult and expensive to build, comparing with other characterization techniques such as the acoustical approach. Additionally, the low throughput of the high-speed imaging measurement is another disadvantage. Measuring single microbubbles requires diluting, selecting and manipulating the bubble in a capillary or flow channel, and two people are usually required to perform a high-speed imaging experiment.
The optical system is limited by its numerical aperture and the wavelength of the light. Optical sizing of ultrasound contrast agents with a radius approaching the wavelength of light is difficult, as it can produce a complex bubble image strongly influenced by diffraction and scattering of the incident light. For example, the optical resolution of the Brandaris 128 camera was 0.4 µm [45]. This can result in a bias of the size range of the contrast agents selected for study, especially for smaller contrast agent particles such as ELIP and Nanodroplets [78]. However, this limitation can be compensated by fluorescence imaging, since no transmitted light is required, and bubble-background contrast can be greatly enhanced [38,52]. Furthermore, during an optical measurement, microbubbles will reside at the top of a capillary or a chamber. Under the ultrasound excitation, the bubble does not just undergo radial excursions, but also translational motion with respect to the optical focal plane [79]. This can produce out-of-focus errors depending on the amplitude of oscillation. When the amplitude is kept below one micrometer, the error will be on the order of 5-10% [52].

A. Modelling and mechanical approaches
The microbubble coating plays a key role in determining both its stability and acoustic response. Accurately measuring the properties of microbubble coatings however presents some significant challenges. The majority of existing theoretical models characterize the influence of the microbubble coating in terms of an effective elasticity and effective viscosity [57,59,80]. One way to determine these properties is to utilize one of the methods described in sections III or IV to obtain the radial response of the microbubble as a function of time and then fit the data to the theoretical model. Whilst it can be shown that for two coating parameters a unique solution does exist, there will inevitably be considerable uncertainty in the values obtained [81]. Alternative methods include micropipette aspiration [82] and atomic force [83] microscopy (AFM). In the former, a small section of the microbubble coating is drawn into a glass pipette under a known pressure and the resulting deformation measured via optical microscopy. This technique however only provides a measure of coating elasticity under quasi static conditions and the relevance of this to microbubble response at ultrasonic frequencies is unknown. In AFM, a cantilever probe is either drawn across the microbubble surface or actively vibrated against it in order to produce deformation and hence determine elasticity. Again, however the rate of deformation is slow compared to that of acoustically driven microbubble oscillations and the mode of deformation is different (bending rather than dilation).

B. Fluorescence microscopy
In applying all of the above techniques it is implicitly assumed that the microbubble coating can be treated as a continuous, single phase material. Several studies have now demonstrated however that partitioning and domain formation occurs even in single component microbubble coatings [82,84,85] and that these domains govern the surface microstructure and mechanical properties. It is more appropriate therefore to consider the microbubble coating as layer of adsorbed surfactant molecules whose concentration can vary both with time (particularly if the microbubble is oscillating) and position on the bubble surface. The "elasticity" and "viscosity" of the coating are functions of this variation in molecular concentration and thus will vary with individual microbubble composition, time, temperature and other environmental conditions. There are a number of models that have been developed on this basis [54,59]. In order to characterize these surface properties in more detail a variety of different quantitative fluorescence imaging techniques can be applied. In each case a suitable fluorescent molecule needs to be incorporated into the membrane in order to probe a specific characteristic.

1) Fluorescence lifetime imaging (FLIM)
FLIM can be used to investigate microbubble surface properties in combination with a specific type of fluorescent probe known as a molecular rotor. The rate of decay of the fluorescent emissions from the rotor is sensitive to its local environment. Thus by incorporating the rotor into the microbubble coating it can be used to map variations in molecular arrangement across the microbubble surface, as shown in Figure 9. Examples of molecular rotors that can be used to interrogate phospholipid membranes include Borondipyrromethene (BODIPY) C10 and C12. These molecules preferentially locate in the hydrophobic "tail" region of the membrane and so indicate the degree of molecular interaction in this region. These interactions have been shown in previous studies to correspond to the apparent viscosity of the membrane [86]. In order to perform FLIM measurements on microbubbles the molecular rotor needs to be combined with the microbubble coating during fabrication. This is typically done during the preparation of the lipid film [87] that is subsequently hydrated and processed to form bubbles. A fluorescence microscope with a FLIM detector is required to perform the required measurements and the data then need to be fitted to a suitable decay model to determine the corresponding lifetimes. The Forster-Hoffman equation provides the relationship between fluorescence lifetime and viscosity. Hosny et al. reported average measurements of microbubble phospholipid coating viscosity of the same order of magnitude as values derived from high speed camera measurements but also showed that viscosity varied considerably both between microbubbles and over the surface of individual bubbles. One advantage of FLIM is that the timescales of intramolecular rotation (the equivalent "rate of rotation" of the rotor) correspond to those experienced by microbubbles oscillating at ultrasonic frequencies. An important point to note, however, is whilst the measured "viscosity" and the viscosity term appearing in theoretical microbubble models will be closely related, they may not be equal and further work is required in this area to formalize the definitions.

2) Fluorescence recovery after photobleaching (FRAP)
Kooiman et al. [88] used an alternative technique to study membrane viscosity. In their study biotinylated phospholipid bubbles were prepared and conjugated to fluorescent streptavidin Oregon Green 488. High resolution three dimensional images ( Figure 10) were obtained using a specialized confocal laser-scanning microscope containing a matched pair of aligned opposing 100x objective lenses [89]. A high intensity laser was used to bleach the fluorophores within a specified region on the bubble and further images were then acquired to measure the rate at which fluorescence emissions recovered from this region. This time is determined by the rate at which molecules are able to diffuse over the microbubble surface, which is itself proportional to membrane viscosity. This study also reported viscosity values consistent with those obtained from high speed imaging although, again, the definition of viscosity requires further attention. Fig. 10. Targeted DPPC microbubbles (red fluorescent) bound to streptavidin-coated surface (green fluorescent; covalently linked) reproduced from [216].

3) Spectral imaging
Fluorescent probes that locate preferentially in the hydrophilic regions of surfactant membranes can also be used, e.g. Laurdan and its derivatives. The emission spectrum of Laurdan shifts as a function of the dipolar water relaxation, and thus the level of hydration within its surrounding microenvironment [90], which in turn is indicative of molecular concentration or "packing". This is a potentially extremely useful measure, as both "elasticity" and "viscosity" can be directly related to it, although the specific functional relationships for microbubble coating materials have yet to be defined. Fluorescent probes of this kind can be incorporated into microbubble coatings during preparation as above but in this case measurement needs to be made of the spectral content of the fluorescent emissions rather than the lifetime. This requires a microscope equipped with a detector array having the required bandwidth for a given probe. Molecular packing is defined in terms of a "Generalized polarization" (GP, ranging from -1 to 1): Where I 440 and I 490 correspond to the fluorescence intensity at 440 nm and 490 nm emission wavelengths, respectively. A high GP value corresponds to a small amount of water in the vicinity of the probe, and hence to a high degree of molecular packing.

C. Electron microscopy
Direct visualization of the microbubble surface at the nanoscale requires the superior spatial resolution of electron microscopy (EM). Unfortunately the vacuum environment inside the EM system makes it extremely challenging to image microbubbles intact. Microbubbles have nevertheless been successfully examined using a variety of both scanning and transmission electron microscopy (SEM, TEM) techniques. For example, polymer coated microbubbles have been imaged by [91,92]. Vlaskou et al. [93] and Hithcock et al. [94] studied acoustically active liposomes by embedding them in gelatin blocks that were then sectioned and stained with uranyl acetate and/or lead citrate. Freeze fracture TEM has also been employed by Kim et al. (Kim et al. 2003) and Borden et al. 2006 to study the surface of phospholipid coated microbubbles. Owen et al. recently proposed a low cost method for examining microbubbles intact under TEM, using the uranyl acetate stain itself to stabilize the bubbles [43]. This enabled visualization of nanoparticles embedded within the microbubble coating.

A. Characterization of microbubble chemistry
Drug loading of microbubbles can be achieved in multiple ways, as illustrated in Figure 11, and accurate measurement of the amount loaded is vital in order to determine the availability for delivery to a target area. This section will review the main methods for determining the drug loading capacity of a microbubble and quantity of drug actually loaded. Fig. 11. Schematic of different potential loading techniques for lipid microbubbles. All techniques are applicable to protein and polymer microbubbles. Direct incorporation relies on the hydrophobic (or lipophilic), hydrophilic or amphiphilic interactions between the drug and shell. Electrostatic binding relies on charge differences between the drug or drug carrier and shell molecules. A method not described in the text is loading the drug into an oil layer trapped beneath the shell. Approaches relying on functionalisation of the shell and/or drug carrier include carbodiimide chemistry on exposed carboxylic acid (-COOH) or amine (-NH2) groups, tagging with avidin or biotin to form an avidin-biotin bond and using pyridyl disulphide or maleimide to link to exposed sulfhydryl (-SH) groups. Abbreviations: EDC: 1-Ethyl-3-(3-dimethylaminopropyl) carbodiimide, NHS: N-hydroxysuccinimide, PEG: polyethylene glycol.

B. Quantification of loading capacity 1) Direct incorporation
The simplest method for loading is direct incorporation of the drug into the shell of the microbubble. Lipid and protein microbubbles have an amphiphilic structure, which allows loading of amphiphilic, hydrophilic [95] or hydrophobic drugs [96][97][98]. Albumin in particular has excellent inherent capability as a drug carrier [99][100][101]. This method is simple to employ and drugs are rapidly released upon ultrasonic activation; but the drug may influence shell stability, microbubble behavior and release may occur in areas where microbubbles are trapped, for example, the lungs [102], spleen [103,104] or liver [105,106]. The drug loading capacity of microbubbles could be estimated using known parameters of drug loading area, hydrophobicity and drug solubility, however quantification of the loaded drug, by methods described later, are simpler and more accurate.

2) Electrostatic incorporation
Direct incorporation of hydrophilic drugs or drug carriers onto the surface of microbubbles can also be achieved by electrostatic interaction [95,[107][108][109]. Designing a microbubble with the opposite charge of the drug or drug carrier -which may itself be modified to achieve the desired charge -will create an electrostatic force which binds the two components together. In microbubble chemistry, hydroxyl, phosphate and carboxylic acid groups will confer a negative charge. Positive charges are commonly provided by protonated nitrogen groups, e.g. amine, trimethyl amine, etc. The presence of these groups can be confirmed by many standard chemical tests, however the charge conferred is of primary interest. Zeta potential measurements, whilst not specifically the charge present at the surface of the molecule, provides a useful means for determining the charge on a molecule. However, there are some challenges associated with measuring microbubbles in commercial machines as discussed below. Due to their semirigid, cross-linked nature, shell remnants of destroyed protein microbubbles should yield a relatively good estimation of zeta-potential but this is not necessarily the case with all materials. Consideration must also be given to both the buffering solution and effect in plasma, as the charge of a molecule will change with pH and ionic strength. This may also be of use for site-specific release in tumors where the high metabolic rate in tumor cells and excretion of H + ions result in acidification of the tumor environment [111][112][113].

a) Zeta potential measurement
Zeta potential of particles can be obtained through laser Doppler electrophoresis (LDE), which measures MB electrophoretic mobility in an electric field by light scattering and calculating its zeta potential. However, the main limitation of this method is that the buoyancy force acting on the bubbles significantly affects their measurement as commercial systems typically have a vertical. An alternative approach reported in [110] uses a simple microfluidic device for obtaining the zeta potential of three types of commercial microbubbles. The microbubbles are suspended in a horizontal micro-channel under an electrical field and optically tracked as illustrated in Figure 12. The velocity of the bubbles is then used to calculate their zeta potentials. By comparing LDE and the homemade device, a 10 fold improvement in measurement variation was achieved. It should be noted that coating of the channel is needed to avoid MBs sticking to the wall as well as reducing electro-osmotic flow on the walls. Fig. 12. Experimental set up for measuring the zeta potential of microbubbles as used in [110].

3) Conjugation of drug carriers
In many cases, permanent attachment to a microbubble shell may be necessary. In these instances, the drug is typically loaded into a secondary drug carrier, i.e. polymeric nanoparticles or liposomes, and attached to the microbubble via electrostatic or chemical conjugation. This improves colocalization of the drug with the microbubble at the site of activation and can help reduce systemic toxicity by drug encapsulation. Additionally, conjugation of drug carriers to the shell can potentially increase the loading capacity per microbubble in comparison to incorporation into the shell. Kheirolomoom et al. approximately calculated up to 10,000 liposomes of 100 nm were loaded via avidin-biotin linkages on to single microbubbles with mean diameters of 1.7 µm [114]. Compared to microbubbles only, the attachment of these liposomes can give a nearly 35 fold increase in the potential surface area alone [115]. Many conjugation techniques are available and many protocols exist for their use, in both the literature and from commercial suppliers. Direct conjugation of a carrier to a shell is possible via functional ligands, however linker molecules are typically employed to separate the drug carrier from the shell. A popular linker is polyethylene glycol (PEG), due to its low toxicity, low cost, the wide availability of different forms and conjugates, and the ease of manipulation. PEG chains with two differing functional groups -a heterobifunctional ligandcan allow for very precise loading of desired carriers with low interference. For protein microbubbles, the PEG linker should be conjugated onto the protein shell post-production. Preproduction conjugation may interfere with functional groups required for stable shell synthesis, although, with controlled reactions, incomplete usage of functional groups could be attempted. For lipid microbubbles, conjugation can be performed pre-production to reduce the number of handling steps of fragile microbubbles. Alternatively, PEG-lipid conjugates with functional groups are already widely available and can be incorporated directly during microbubble production. Unreacted PEG molecules can be quantified by refractive index detection after separation from microbubbles by washing and size exclusion chromatography (SEC) [116]. Additionally, quantification of the relative and absolute amounts of specific lipids incorporated into the shell could be performed using gas chromatography [117,118], FTIR [119] or HPLC [120,121] on washed microbubbles. HPLC combined with an evaporative light scattering detector (ELSD) has been used to determine the amounts of lipids in commercial and research microbubble formulations [120], but has yet to find widespread use in the field potentially due to the low amounts of lipids used and complexity of establishing the technique [122]. Alternatively, characterization of the available functional groups on microbubbles can be attempted.
A few examples of the most common functional groups will be covered here:

4) Avidin-biotin
Avidin-biotin is one of the strongest non-covalent bonds commonly available to researchers. It is simple to use and many PEG linkers or PEG-lipid conjugates with either biotin or avidin ligands are available. It is highly tolerant to a wide range of buffering conditions [123,124] and is common in the preparation of novel microbubble delivery or imaging agents [114,[125][126][127]. However, the bond is practically irreversible, requires additional washing and handling steps and avidin is potentially immunogenic; although recent research has shown that the last point may not be of as great a concern as previously thought [128][129][130]. Qualitative determination of incorporation into a microbubble shell can be achieved by widely available fluorescent biotin or avidin. Quantitatively, colorimetric and fluorimetric tests using 2-(4'-hydroxyazobenzene) benzoic acid (HABA) are available to rapidly determine biotin levels in a sample [131,132]. This has been performed for biotinylated albumin to determine the binding efficiency of the conjugation of biotin to albumin subsequently used to create microbubbles [133] and on antibodies to be conjugated to microbubbles via avidinbiotin [134]. However, considering the low proportion of biotinylated lipids used in formulations, typically ranging from 5 to 10% molar ratio, microbubbles have not been directly investigated in this manner yet as even fluorimetric HABA assays may struggle to determine biotin quantities.

5) Carbodiimide chemistry
In carbodiimide chemistry, a carboxylic acid group is reacted with 1-Ethyl-3-(3-dimethylaminopropyl) carbodiimide (EDC), typically in the presence of N-hydroxysuccinimide (NHS) or N-hydroxysulfosuccinimide (sulfo-NHS) to improve yield, forming a ligand capable of forming a covalent bond with amine groups. Carbodiimide chemistry is popular as a protein linker ligand, as amine and carboxylic acid groups are common in protein residues. It is also popular for adding other functional groups to a molecule, e.g. sulfhydryls, avidin, biotin, etc. Typically, linkers use NHS due to its increased stability, however the reaction can be performed on carboxylic acid or amine groups already present on the molecules without further manipulation. A lipid or protein microbubble shell with available carboxylic acid groups can be modified by EDC/NHS to bind amine functionalized linkers or drug carriers [135], or vice versa [136]. Lipid-PEG-NHS conjugates are available for direct incorporation during lipid microbubble manufacture. The carbon-nitrogen bond formed is "irreversible" and this may be a disadvantage for microbubble release. Additionally, the reaction can be challenging to perform and optimize as reagents are very sensitive to pH, temperature and time.
Quantifying EDC or NHS availability in the microbubble shell can be a challenge as stability is dependent upon pH and both will cleave over time. Measuring the concentration of unreacted groups may be easier. EDC, and its hydrolyzed decay products, can be detected by mass spectroscopy [137] and colorimetric methods [138,139]. NHS and sulfo-NHS have strong absorbance at 260 nm under basic conditions [140], and their unreacted conjugates can be isolated and detected by fluorescence assay [141] or HPLC-UV using a reversed phase column. Alternatively, during a reaction, released NHS and sulfo-NHS can be quantified by HPLC-UV using a hydrophilic interaction chromatography (HILIC) column [142]. Indirectly, a before and after assay of available amine, e.g. 2,4,6-trinitrobenzene sulfonic acid (TNBS) assay [143,144], and carboxylic acid groups e.g. 5-bromomethyl fluorescein assay [145,146], can determine site usage. Finally, an amine displaying a fluorescent compound can be targeted to the activated EDC/NHS groups to determine availability on the shell.

6) Sulfhydryl linkages
On proteins, sulfhydryl groups are less common than amine groups, increasing reaction selectivity and potentially reducing interference of protein microbubble formation. There are also many methods for adding sulfhydryls to target drug carriers via other functional groups, for example Traut's reagent [147] or N-succinimidyl S-acetylthioacetate (SATA) variants which target amine groups. The availability of these sulfhydryl groups can be quantified by several methods [148][149][150], and commercial assay kits are available [151,152]. Several functional groups are capable of binding to sulfhydryl groups, however, for microbubbles, the authors could only find literature for the use of maleimide [125,[153][154][155] and pyridyl disulphide [156][157][158]. Maleimide reagents are more stable than EDC/NHS and form permanent thioether linkages. The presence of maleimide groups can be detected using commercially available fluorimetric assay kits [159] or spectrophotometric approaches [160]. The maleimide group remains in the compound postreaction as a small linker. Pyridyl disulphide forms a direct disulphide bond with sulfhydryl groups with no linker length and release of the pyridine-2-thione, which can be monitored by absorbance at 340 nm. This release can also be used to quantify the availability of functional groups as the pyridyl disulphide bond cleaves under reducing conditions, for example, in the presence of dithiothreitol (DTT), glutathione or Tris(2carboxyethyl)phosphine hydrochloride (TCEP-HCl). Although pyridine-2-thione has a relatively low molar extinction coefficient which may hinder monitoring of the reaction at low concentrations [161], Yoon YI et al. were able to utilize UV/VIS detection at 270 nm and 340 nm to monitor the conjugation of pyridyl disulphide microbubbles to thiolated liposomes [158]. The bond formed with the drug carrier is similarly not stable, which may allow drug release in the reducing environment of tumors [162].

C. Drug loading quantification
Microbubbles have been developed beyond their original diagnostic purpose into effective, targeted, drug delivery agents. As with all drug therapies, knowledge of the dose delivered is fundamental to ensuring a consistent therapeutic effect without adverse events or severe toxicity. Quantification of drug loading onto microbubbles is therefore vital and the following section will cover 4 broad drug types currently in use with microbubbles. It should be highlighted, however, that the characterization technique required will be entirely dependent upon the drug, availability of equipment, cost and convenience. Furthermore, few drugs considered for targeted microbubble and ultrasound delivery are novel. Literature will likely be available describing standard methods of quantification for the drug and should form the principle basis for any choice in protocol. Ideally, quantification should be carried out after loading, either as a subtraction of unbound drug from total used, or as the amount of drug still in the sample. It should be noted though, that techniques to quantify the total loaded drug may not represent actual release when used; some drug may become irreversibly bound to drug carrier or microbubble and not be available. Loading can be measured on either intact or destroyed microbubbles, however the buoyant and light scattering characteristics of intact microbubbles can interfere with some measurement techniques. Negative pressure, sonication and heat are effective methods for microbubble destruction. In all cases, effective washing will be required to remove contaminants. This can be challenging with microbubbles due to their sensitivity to pressure, shear stress and handling. Centrifugation is the most common method for washing microbubbles, but will inevitably cause some concentration loss and change in size distribution [163]. For sufficiently stable microbubble complexes, buoyancy over time may allow purification [164][165][166].
If the drug is difficult to detect, or difficult to separate from the influence of the carrier or microbubble, quantification of the drug carrier could be attempted instead. Fluorophores, either attached to the carrier or carried with the drug, allow easy quantification of loading onto microbubbles by spectrophotometry. Additionally, fluorescent drug carriers can act as convenient markers, for example of delivered "dose", delivery distances from blood vessels, distribution throughout target tissue, off-target dose and method of excretion. Radiolabelled drug carriers may also fulfil this role and can be monitored non-invasively by PET [167] or SPECT, although this approach has increased risk, complexity and costs. Drug carriers can also be tagged with MRI sensitive agents but the equipment required for quantification of the carrier and monitoring of delivery may be prohibitive.

1) Cargo types a) Genetic material
Genetic material is defined here as any form of long or short chain of nucleotides, single or double stranded, including, but not limited to, DNA [168], RNA, mRNA, siRNA [95,169], shRNA and, potentially, some forms of simple virus. The choice of genetic material will depend upon application. From a loading perspective, the most important feature shared between different types of genetic material is the presence of a negatively-charged sugar-phosphate backbone. Cationic drug carriers, such as PEI [170], will load genetic material by electrostatic attraction. This loading can also protect from nuclease degradation in vivo [171,172]. Additionally, the neutralization of the negative can condense the genetic material to improve loading and trans-membrane passage [173].
In genetic material, the phosphate (P) groups confer a strong negative charge. Cationic drug carriers are normally formulated with protonated nitrogen groups (N), e.g. amine, imizadole, trimethyl ammonium, etc. The N/P molar ratio is commonly used to optimize electrostatic loading for genetic material [174,175]. As the backbone is a repeating structure, P can be calculated from the length of the genetic material, which is either known or measurable by standard electrophoretic techniques. For N, values for standard cationic drug carriers, e.g. PEI, are available in the literature. Cationic lipid components could be estimated from the initial ratios of chemical components, however accurate estimates can only be achieved via analysis of the lipid constituents post production, as mentioned earlier. Alternatively, and specifically for protein microbubbles, there are many tests available which can detect protonated nitrogen groups to estimate N, for example the TNBS assay for primary amines [143][144][145].
Loading can be qualitatively demonstrated by measuring changes in zeta potential charge as previously mentioned, or by the use of nucleic acid specific dyes or stains, for example; ethidium bromide, SYBR derivatives, 4',6-diamidino-2phenylindole (DAPI), YOYO-1, Hoescht, propidium iodide [171,176]. Many different options are available for quantitative measurement of genetic material. For lipid microbubbles without aromatic components, the simplest technique available is UV spectroscopy after microbubble destruction [171,172]. However, the correct calibration curve must be chosen, or the genetic material purified, as different forms of genetic material, its structure, contaminants and buffer will affect the absorption profile. Sensitivity of spectrophotometric methods can be increased and contamination issues avoided by the use of the aforementioned dyes, although they do increase cost and are usually hazardous. These stains may be essential for protein microbubbles without purification as some amino acid residues absorb in the UV range [177]. Alternatively, genetic material can be radiolabelled to allow sensitive quantification on intact microbubbles without issues of interference or altering loading properties [178][179][180]. However, this does further increase cost, complexity and potential risks. An alternative method for both quantification and checking for degradation of the genetic material is gel electrophoresis.
Separation of genetic material by electrophoresis can highlight structural abnormalities by differences in migration speed against controls [181,182]. This method can also be used to quantify genetic material if run against known concentrations [183]. For all the techniques mentioned above, it should be noted that microbubbles may leave behind charged shell remnants after destruction which could interfere with subsequent analyses.
Removal of electrostatically bound material may be required, for instance dissolution in 4 M guanidine hydrochloride [184]. Alternatively, standalone kits and techniques are available to isolate genetic material from proteins. Increasing the alkalinity of the sample could be attempted to the deprotonate the nitrogen groups but this should be done at low temperatures as a high pH will result in degradation of most genetic material, particularly RNA [185].
(1) Chemotherapy drugs An active area of microbubble research is loading of established chemotherapy drugs, such as doxorubicin [109,169,[186][187][188][189][190], 5-fluorouracil [191] and paclitaxel [98,192,193]. As chemotherapy drugs have high systemic toxicity, the ability to direct delivery to tumor tissue is an attractive therapy. Quantification of established chemotherapy drugs will be well-described in the literature due to their clinical use and the following will be general guidance on techniques. For some chemotherapy drugs, direct quantification may be possible by simple fluorescence spectrophotometry, for example: rose Bengal [194,195], protoporphyrin IX [196,197], topotecan [198,199], doxorubicin [200], irinotecan [201], etc. This also allows qualitative demonstration of loading and delivery by fluorescent microscopy and histology techniques.
For non-fluorescent drugs, alternative techniques must be considered. Release and separation of the drug from the microbubble or drug carrier can be followed by quantification. Drug carriers, such as liposomes can be disrupted by heating to the lipid phase transition temperature, whereas polymeric nanoparticles like PLGA can be dissolved in solvents or broken down with sodium hydroxide. For most drugs, literature will be available for a variety of techniques and separation criteria. Automated methods for separation, common in most laboratories, are high performance liquid chromatography (HPLC), gas chromatography (GC) or asymmetrical flow field flow fractionation (AF4). Detection is usually performed by UV absorbance, although many different detectors are available, including mass spectrometry, evaporative light scattering (ELS), multi-light scattering detector, etc. For example, HPLC-UV methods have been described for cisplatin [202], mitomycin C [203] and 5-fluorouracil [204]. In addition, certain drugs, such as photodynamic or sonodynamic therapy drugs, generate detectable free radicals and reactive oxygen species upon activation [205,206], which could be used for drug quantification. (

2) Proteins
Loaded proteins can include hormones [127] or growth factors [207,208]. Potentially, the proteins also could act as a drug carrier, for instance, albumin nanoparticles [99,100,116,209]. Protein detection can be performed by UV spectroscopy at 280 nm but the sensitivity is low, extinction coefficient is dependent upon sequence, and many contaminants absorb in the same range. Instead, many colorimetric or fluorescent assays are available for protein quantification, e.g. the Lowry method, bicinchoninic acid assay, Quant-iT™, etc. (see Noble JE and Bailey MJA, 2009 for a review [210]). Fluorescence methods also allow qualitative assessment of loading on to intact microbubbles. Protein samples can also be examined on HPLC-UV machines by several methods [211,212]. Finally, if the protein has some intrinsic enzymatic activity, an assay could be used to quantify both loading and integrity of the sample.
(3) Cells Stem cell delivery using acoustic radiation force and microbubbles is another area currently under research [190,213,214]. In these cases, loading is "reversed", with one cell potentially loading multiple microbubbles via antibodies. Optical microscopy is usually sufficient to determine microbubble loading, although fluorescent staining of the bubble with lipophilic dyes or fluorophore conjugation would allow automated techniques such as fluorescence assisted cell sorting (FACS) to be used. FACS would also allows sorting of cells with desired levels of microbubble loading and a rough estimate of cell number can be determined from absorbance at 600 nm, although this technique should be performed without microbubbles to avoid light scattering effects.

VII. DISCUSSION AND CONCLUDING REMARKS
This paper presents the means by which different research groups have sought to characterize and compare microbubbles for diagnostic and therapeutic applications. It is important to note that for each of the characteristics discussed, comparison across research labs or indeed between different investigations within the same lab are only possible when careful consideration of the environmental and set-up conditions which may influence microbubble behavior has been made. In addition, where results are to be used to gain insight to behavior in vivo, careful control of temperature, gas saturation and preparation and handling methods used are essential to give reliable results. It is clear that while the case for developing a standardized method of comparison for microbubble contrast agents is strong, several challenges remain to support translation of new formulations for therapy and emerging quantitative imaging techniques from research to practice. The characterization of polydisperse microbubble suspensions by mean diameter does not capture the complexities of size distribution or echogenic behavior. There is certainly a case for including alongside this information details of the gas volume distribution and description of the suspensions polydispersity characteristics. There is no consensus on a reliable and useful means of comparing microbubble acoustic behavior. The development of the STAR method has contributed to this discussion by providing a useful and easily understood means of comparison, but has not seen wide adoption. This may be in part because commercial agents, which are considered to be well characterized, remain dominant for in vivo investigations. It is also possible that the STAR/nSTAR value could be similar for a contrast agent suspension which does not attenuate or scatter effectively to one which both attenuates and scatters effectively. However, our growing understanding of how environmental factors influence microbubble behavior, the continued development of new, and particularly in-house fabricated microbubbles which are particular to each research group, plus our demands for improved insight to microbubble biophysics to understand therapeutic applications and extend their utility ensures that the need to accurately characterize microbubbles will remain important. Before her move to Glasgow she was a post-doctoral research assistant in the Imaging Sciences Department, Imperial College London and prior to this worked in industry as a Mechanical Design Engineer and Project Leader. Dr Mulvana's research focuses on the use of ultrasound and ultrasound contrast agents to improve the diagnosis and treatment of cancer. Her particular interests concern improving the understanding of how ultrasound and microbubbles interact with tissues and cells and developing new imaging methods to aid in the accuracy and availability of ultrasound diagnostic information to clinicians in the pursuit of personalized cancer treatment. Her research is funded by the Royal Society, The Carnegie Trust and Bowel and Cancer Research UK. She is an associate editor of the IEEE Transactions on Ultrasonics, Ferroelectrics and Frequency Control and is on the Technical Program Committee of the IEEE International Ultrasonics Symposium. Helen is Chartered through the Institute of Mechanical Engineers.