3.1.

Preparation of Fluorophore-Labeled Microbubbles

Microbubbles were formulated with a lipid suspension of 90 mol % DSPE (1,2-Distearoyl-sn-glycero-3-phosphoethanolamine, ME-8080, NOF America Corp., New York) and 10 mol % DSPE-PEG [N-(carbonyl-methoxypolyethyleneglycol 2000)-1,2-distearoyl-sn-glycero-3-phosphoethanolamine, DSPE-020CN, NOF America Corporation, New York] at $2\mathrm{mg}/\mathrm{mL}$ in 100 mL of phosphate-buffered saline (PBS, pH 7.2). Perfluorobutane (APF-N2HP, FluoroMed, Texas) gas was encapsulated as the gas core.³⁹ Microbubbles with sizes distributed between $\sim 4$ and $\sim 10\mu \mathrm{m}$ were selected for use. The size distribution of the microbubbles was determined by laser light obscuration and scattering (Accusizer 780A, NICOMP Particle Sizing Systems, Santa Barbara, California). To conjugate fluorophores on the microbubble surface, an ATTO532-NHS (Sigma-Aldrich, Missouri) dye solution was added to a diluted microbubble solution. To control the initial surface concentration of the dye on the microbubbles, several NHS-to-amine molar ratios were adopted; they are listed in Table 1. The mixture was reacted in a pH 8.5 PBS buffer (adjusted pH with 0.1 M NaOH) for 1 h at room temperature with constant gentle agitation. After that, the unreacted ATTO532-NHSs were removed through three rounds of centrifugal washing. The purified fluorophore-labeled microbubbles were diluted and injected into a glass chamber at the bottom of a water tank for imaging, as shown in Fig. 2. The chamber was made with two cover glasses (12-548-B, Fisher Scientific, Pittsburgh) stuck together by double-sided Scotch tape, creating a space distance of $\sim 200\mu \mathrm{m}$. Note that this fluorophore was selected because of its high quantum yield and high photon stability and because our picosecond (ps) laser for fluorescence lifetime image has a wavelength of 532 nm.

Table 1

Fluorophore-labeled microbubble (MB) solution.

Fig. 2

Preparation of fluorophore-labeled microbubbles.

3.2.

Characterization of Fluorescence Intensity and Lifetime of Fluorophore-Labeled Microbubbles

All experiments were conducted with a Nikon inverted microscope (Ti-U, Nikon), as shown in Fig. 3(a). A 532-nm ps laser (Katana, Onefive, Zurich, Switzerland) was synchronized with a gated and intensified charge-coupled camera (ICCD) system (Picostar HR, LaVision, Goettingen, Germany) (a minimum gate width is 200 ps) as a fluorescence lifetime imaging microscope system (FLIM). This FLIM system was used to measure the fluorescence intensity and lifetime of the fluorophore-labeled microbubbles. The ps pulsed laser was coupled into a multimode optical fiber ($62.5\text{-}\mu \mathrm{m}$ core diameter). The output light from the fiber was collimated and delivered to a filter cube where an excitation filter, a dichroic mirror, and an emission filter were installed. The dichroic mirror reflected the light into a $100\times$ oil immersion objective [$\mathrm{NA}=1.3$, working distance $(\mathrm{WD})=0.16\mathrm{mm}$] to illuminate the sample. The emitted fluorescence from the microbubbles was collected by the same objective lens and was detected by the gated ICCD camera system after passing through the same dichroic filter, an emission filter, and a flip mirror. In the filter cube, a 525/40-nm band pass filter (FF02-525/40-25, Semrock, New York), a 552-nm dichroic filter (FF552-Di02, Semrock, New York), and a 578/28-nm band pass filter (FF01-572/28-25, Semrock, New York) were used as the excitation, dichroic, and emission filters for dye ATTO532-NHS, respectively. Note that the excitation filter was not necessary when the ps-pulsed laser (532 nm) was used for lifetime imaging. However, it was necessary to select the appropriate light to excite the fluorophores when a lamp was used for UMF signal detection (see the next paragraph). The ICCD camera system incorporated a high-rate imager (Kentech Instruments Ltd., Oxfordshire, United Kingdom) and a ps delay unit, which enabled it to generate a time gate as short as 200 ps that could be temporally scanned with a step size down to $\sim 10\mathrm{ps}$. Here, we chose a gate width of 300 ps and a temporal step size of 100 ps, which was narrow enough to image fluorescence lifetime in a range of nanoseconds (ns). Images were saved in the computer and processed later with MATLAB. In order to calculate the fluorescence lifetime, the acquired images were fitted to a single exponential decay function pixel by pixel. An iterative numerical procedure was done until the best agreement between the experimental decay curve and the theoretical model decay curve was achieved. For each microbubble, the fluorescence lifetime of every pixel was calculated; therefore, the fluorescence lifetime image of the microbubble was obtained. The fluorescence intensity image of the same microbubble was obtained by selecting the image with the highest intensity in the sequences. For simplicity, the fluorescence lifetime and intensity of a bubble were defined as the mean lifetime and intensity of all the pixels in that bubble image. For statistical analysis, at least 10 microbubbles were randomly selected in the population. The averaged lifetime and intensity with standard deviation were calculated based on those microbubbles.

Fig. 3

(a) A schematic diagram showing the imaging system for characterization of a single microbubble’s oscillation and its UMF signal; PA: power amplifier; FG: function generator; PDG: pulse delay generator; UST: ultrasound transducer; OS: oscilloscope. (b) An example to show the microbubble oscillation measured by the ICCD camera system, and the synchronization between the bubble oscillation and the ultrasound-driven UMF signal. Note that the images displayed were selected from the actually collected images and the delay time between two sequential images is $0.5\mu \mathrm{s}$.

3.3.

Detection of Ultrasound-Driven Microbubble Oscillations and UMF Signal from Individual Microbubbles

Figure 3(a) was designed to study the ultrasound-driven microbubble oscillation and the corresponding UMF signal. In the acoustic part [see the green blocks on Fig. 3(a)], a 1-MHz single element ultrasound transducer (UST, V314-SU-F-1.00-IN-PIT, Olympus NDT, Texas) with a focal length of 25.4 mm was mounted onto a three-dimensional translational stage (PT1, Thorlabs) and focused on the microbubble sample. The driving signal was generated by an arbitrary function generator (FG, Agilent 33220A, Agilent Tech., California) and amplified by a radio-frequency power amplifier (PA, 2100L, Electronics & Innovation Ltd., New York). In this study, the driving signal was a three-cycle 1-MHz sinusoidal electronic wave with a repetition rate of 5 Hz. It generated an ultrasonic pressure oscillation with a cycle number slightly larger than 3 (or a duration time slightly longer than $3\mu \mathrm{s}$) due to the finite bandwidth of the ultrasound transducer [see Fig. 3(b)].

In the optical part, the ICCD camera was employed to detect microbubble oscillation [see the blue blocks on Fig. 3(a)]. The principle was introduced in our previous publication.⁴⁰ Briefly, to observe the fast bubble oscillation, a bright xenon strobe light source (AC-4020-C, Electromatic Equip’t Co., Inc., New York) was used to illuminate the oscillating bubble via a fiber bundle (40-644, Edmund Optics, New Jersey). The strobe light source illuminated a $10\text{-}\mu \mathrm{s}$ light pulse when receiving a trigger signal from the pulse delay generator (PDG) that was used to synchronize the strobe light, ultrasound pulse, and ICCD camera recording. In each ultrasound pulse (slightly longer than $3\mu \mathrm{s}$), the strobe light was triggered on ($\sim 10\mu \mathrm{s}$ and the ICCD camera was gated on with a window width of 20 ns to capture a frame image of the oscillating bubble. To reconstruct the entire bubble oscillation event during the period of the ultrasound pulse, the above procedure was repeated (with a repetition rate of 5 Hz). Compared with each previous step, the time to trigger on the ICCD camera for 20 ns was delayed by 100 ns in each repeated ultrasound pulse. Thus, each frame image acquired by the ICCD camera had a time interval of 100 ns. Eventually, a total of 44 ultrasound pulses were repeated, and the ICCD camera captured a total of 44 frames of images, which covered a total duration of $4.4\mu \mathrm{s}$ ($44\times 100\mathrm{ns}$). Those frames were saved and processed with MATLAB. In each frame, the contour of the microbubble was delineated, and the bubble diameter was calculated. By repeating this procedure for all 44 frames, a diameter versus time curve was generated and shown on Fig. 3(b).

After the oscillation measurement, the UMF signal from the same microbubble was detected with the system shown in Fig. 3(a) (see the orange blocks). The light source was switched to a lamp (Lumen 200, Prior Scientific, Massachusetts) to excite the fluorophores on the microbubbles (note that a lamp provided a large and relatively uniform illumination area compared with the laser). An adjustable iris was positioned in front of the lamp and used to control the illumination area in such a way that only one microbubble was illuminated in the field of view. The fluorescent emission was detected by a cooled photomultiplier (PMT, H7422-20, Hamamatsu, New Jersey). After that, the electronic signal from the PMT was amplified by a broadband amplifier (SR445A, Stanford Research Systems, California), filtered by a low-pass filter ($\mathrm{BLP}\text{-}10.7+$, Mini-Circuits), and finally acquired by a high-speed digital oscilloscope. The oscilloscope was synchronized with the ultrasound pulse via the PDG so that it acquired and displayed the UMF signal when the microbubble was oscillating [see Fig. 3(b)]. The PDG with multiple channels (DG645, Stanford Research Systems, California) was used as the master clock to trigger the function generator, xenon light, ps laser, ICCD camera system, and oscilloscope.

3.4.

UMF Measurement from a Microbubble Population

Following the characterization of individual microbubbles, the UMF signal from a microbubble population was investigated. Specifically, a microchannel was filled with the solution of the UMF contrast agents, and the UMF signal was collected using an imaging system as shown in Fig. 4. The channel was covered by a slab phantom that was made of either a clear medium (water) or an optically scattering medium (intralipid solution). The major difference of this imaging system from the one shown in Fig. 3 was that a gated boxcar integrator (SR250, Stanford Research Systems, California) was used to increase detection sensitivity to the UMF signal. The contrast agents were injected into a $500\text{-}\mu \mathrm{m}$ PDMS microchannel (SynVivo, CFD Research Corporation, Alabama) with a concentration of ${5.84\times 10}^7\text{microbubbles}/\mathrm{mL}$, measured by a hemacytometer (bright-line, Hausser Scientific, Pittsburgh). A continuous-wave (CW) 532-nm laser (MGLII532, Dragon Lasers, Jilin, China) was used as the excitation light source. A $4\times$ objective (CFI Plan Achro, $\mathrm{NA}=0.1$, Nikon, New York) with a WD of 30 mm was adopted to deliver the excitation light and to collect the emission light. The same optical filters used for Fig. 3 were employed here. In comparison with the single microbubbles, a bulk solution of the microbubbles has a relatively strong background fluorescence because of the large illumination area and the existing residue of the free fluorophores in the volume. To extract the UMF signal from the strong background, the gated integrator was employed after the (electronic) low-pass filter and before the oscilloscope. The gate window of the integrator was set to $2\mu \mathrm{s}$, which was precisely controlled to be overlapped with a two-cycle UMF signal in response to the ultrasound wave [see Fig. 4(b)]. The UMF signal within the gate window was integrated, and the integrator output a voltage that was proportional to the average of the input signal. An asymmetric input signal relative to the baseline was desired to avoid a zero output. A moving exponential average of 300 samples was selected to increase the signal-to-noise ratio. (A moving exponential average is formed by putting more weight to the recent values, therefore, it has less time lag. The signal-to-noise ratio is increased by the square root of the number of samples in the average in the case of random white noise background.) The 1-MHz transducer mounted on the translational stages was scanned across the tube with a step size of 0.635 mm, and the UMF signal acquired by the gated integrator was displayed on the oscilloscope. In the phantom experiment, a 0.5% intralipid slab phantom with a thickness of 2 mm (${\mu }_s^{\prime }\sim =1.2{\mathrm{mm}}^{-1}$, and ${\mu }_a\sim =0.001{\mathrm{mm}}^{-1}$ was positioned between the microchannel sample and the objective lens to mimic a scattering tissue.

Fig. 4

(a) A schematic diagram showing the imaging system for UMF scanning across a tube; PA: power amplifier; FG: function generator; PDG: pulse delay generator. (b) A diagram showing the $2\mu \mathrm{s}$ gate window of the integrator that was overlaid with the signal.

As a short summary, four light sources were used in the imaging system according to their unique features. The ps-pulsed laser was used for ns-lifetime imaging of the fluorescent contrast agents. The strobe light source was used to provide strong illumination in a 20-ns exposure window for imaging the oscillating bubbles. The lamp was used to provide preferred large and relatively uniform illumination under a $100\times$ objective for measuring the UMF from individual microbubbles. The 532-nm CW laser was employed to provide a stronger illumination under a $4\times$ objective for measuring the UMF from a microbubble population.

4.1.

Quantification of Self-Quenching on Individual Fluorescent Microbubbles

As listed in Table 1, six groups of fluorophore-labeled microbubble solutions were prepared by mixing the solutions of the ATTO532-NHS dye and the amine-attached microbubble solutions with an NHS-to-amine mole ratio of 0.01, 0.03, 0.1, 0.3, 1, and 3, respectively. Figure 5(a) shows a set of the representative fluorescence lifetime images of the six groups of microbubbles. Averaged lifetime and intensity were plotted as a function of the mole ratio. The averaged lifetime decreases as the dye concentration increases, as shown in Fig. 5(b). Using group #6 as an example, it has the highest NHS-to-amine ratio and shows the shortest fluorescence lifetime of $\sim 0.78\mathrm{ns}$, revealing a strong self-quenching. By contrast, group #1 has the lowest NHS-to-amine ratio and shows the longest fluorescence lifetime ($\sim 3.8\mathrm{ns}$), indicating no or little quenching occurs because the 3.8 ns is the lifetime of the free ATTO532-NHS. This is also confirmed by the fluorescence intensity result in Fig. 5(c). Generally, the fluorescence intensity first increases as the surface dye concentration increases and then drops at a higher dye concentration where the self-quenching dominates.

Fig. 5

(a) Fluorescence lifetime images of individual fluorophore-labeled microbubbles in the six groups, as listed in Table 1. The scale bar is $6\mu \mathrm{m}$, and the label of the color scale is from 0 to 4 ns. (b) The averaged fluorescence lifetime and (c) the averaged fluorescence intensity with standard deviation (shown as error bar) of 10 randomly chosen microbubbles were plotted as a function of NHS-amine mole ratio.

4.2.

Quantification of Individual Fluorescent Microbubble Oscillations under Different Ultrasound Pressures

The amplitude of microbubble oscillation under different ultrasound pressures was measured for all six groups. We use peak–peak pressure in this paper. The oscillation amplitude was calculated as the difference between the maximum diameter during the bubble expansion and the minimum diameter during the bubble contraction. The results are plotted as a function of the ultrasound pressure in Fig. 6. The results show that the oscillation amplitude increases almost linearly with ultrasound pressure. An averaged maximum of $\sim 2\text{-}\mu \mathrm{m}$ oscillation amplitude is observed at 675 kPa, indicating a 33% size change when considering that the average bubble size is $6\mu \mathrm{m}$. No significant difference was observed among the six groups, which implies that the bubble oscillations are independent of the surface dye concentrations. In this study, the ultrasound pressure was controlled to be no larger than 675 kPa to minimize bubble translation or fragmentation. When the ultrasound pressure was further increased, microbubbles were moved from their resting positions and became difficult to track with the gated ICCD camera system. Some of the observed microbubbles still underwent translation even at the pressure of 675 kPa. The results displayed in Fig. 6 are based on at least 10 randomly selected microbubbles; bubbles that experienced moving, shrinking, or breaking during experiments were discarded from the analysis.

Fig. 6

The oscillation amplitude of individual fluorescent microbubbles as a function of the applied ultrasound pressure for the six samples.

4.3.

Quantification of the UMF Signal from Individual Fluorescent Microbubbles

Figure 7(a) shows a typical UMF signal acquired from a single fluorescent microbubble in group #5. A clear fluorescent oscillation was observed when the ultrasound was applied. With the absence of ultrasound pulse, the fluorescent emission was nearly constant. The asymmetric oscillation around the baseline is related to the microbubble’s initial quenching status. In group #5, the bubbles initially presented a strong quenching. When ultrasound was applied, the bubble compression increased the quenching effect to a limited degree and caused a relatively small fluorescence decrease. Conversely, the bubble expansion caused a significant signal increase. The strength of the UMF intensity was calculated as the peak-to-peak voltage. As Fig. 7(b) illustrates, the UMF intensity was strongly dependent on the surface dye concentration. Groups #4 and #5 showed the strongest UMF signal, followed by group #6 and then groups #3, #2, and #1. Note that nonzero voltage noise ($\sim 2.5\mathrm{mV}$) was observed when the ultrasound pressure was 0 kPa. The noise was mainly due to the baseline fluctuation and was calculated as a peak-to-peak value in the same time window of the UMF presence. Therefore, any signal below or around $\sim 2.5\mathrm{mV}$ was considered to be noise. Thus, group #1 showed almost no UMF signal and can be neglected. To better compare the results, the UMF modulation depth was calculated and is shown in Fig. 5(c). The modulation depth is a relative value and is defined as the ratio of the UMF strength to the corresponding unmodulated fluorescence (i.e., the baseline of the fluorescence or the DC fluorescence signal when there is no ultrasound). Group #1 is not compared here since it presents a negligible UMF signal. As shown in Fig. 7(c), the UMF modulation depth was proportional to the surface dye concentration. Group #6 had the highest dye concentration and exhibited the highest UMF modulation depth, followed by groups #5, #4, #3, and #2. This can be interpreted as follows: when the initial dye concentration is high and the quenching is significant, the bubble fluorescence intensity (i.e., the baseline of the fluorescence) is weak. Thus, even a small UMF caused a significant modulation depth. The results in Figs. 7(b) and 7(c) also imply that the UMF signal strength and modulation depth have a strong correlation to the applied ultrasound pressure. A larger oscillation induced a stronger UMF signal strength and modulation depth. An averaged maximum modulation depth of $\sim 42\%$ was observed at a pressure of 675 kPa with the microbubbles in group #6. We believe that the modulation depth ($\sim 42\%$) is mainly limited by the small oscillation of the microbubbles ($\sim 33\%$). At the risk of destroying microbubbles, a higher ultrasound pressure may be applied to induce an even higher UMF signal. It has been shown that a microbubble diameter can be expanded approximately three times before collapse.⁴¹ We observed an approximately four times expansion under a high acoustic pressure of 1 MPa (results not shown). However, the bubbles were quickly broken. To get a stable UMF signal, repeatable oscillations of bubbles were desired. Therefore, the applied ultrasound pressure in this section was maintained $\le 675\mathrm{kPa}$ for investigating more stable microbubble oscillations and the corresponding UMF signals.

Fig. 7

(a) A typical UMF signal of one fluorophore-labeled microbubble in Group #5; UMF signal intensity (b) and modulation depth (c) of fluorophore-labeled microbubbles as a function of applied ultrasound pressure in different groups.

4.4.

UMF Signal from a Population of Fluorescent Microbubbles

In the first experiment, the slab phantom shown in Fig. 4 was made of water (a clear medium). The group #5 contrast agents were employed and injected into the $500\text{-}\mu \mathrm{m}$ PDMS microchannel tube. We chose those contrast agents because they have strong UMF modulation depth and relatively bright fluorescent emission. When the ultrasound transducer was focused onto the tube, the UMF signal was measured under different ultrasound pressures. The results were normalized and displayed on Fig. 8(a). The UMF signal increased as the applied pressure increased until it reached the maximum at 405 kPa. After that, the UMF signal decreased due to bubble breakage caused by the strong ultrasound pressure. At 765 kPa, the signal decreased to 0, and no bubbles remained in the solution, as observed under the microscope. These results were averaged based on three repeated experiments. After each experiment, the transducer was moved to a different location along the tube direction where microbubbles were intact, as shown in the inset on Fig. 8(a). Compared to the previous results in Figs. 6 and 7, bubbles in the tube presented a lower fragmentation threshold ($\sim 405\mathrm{kPa}$). This can be explained as follows: the cover glasses used in the individual microbubbles experiments (Figs. 6 and 7) had a larger acoustic impedance than the PDMS used in the microbubble population studies (Fig. 8).⁴² Thus, the boundaries formed between the cover glasses and the surrounding water reflected more acoustic energy than the boundaries formed by the PDMS and the surrounding water. Therefore, a larger acoustic energy was needed to oscillate bubbles. After testing, the energy transfer efficiency is $\sim 80\%$ through the cover glass and $\sim 96\%$ through PDMS. After calculation, the actual energy applied to the microbubbles sample was $\sim 540\mathrm{kPa}$ after the cover glass and 390 kPa after the PDMS medium. It shows that the actual pressure applied to the bubble solution was a little smaller than that applied for oscillating individual bubbles. This difference can be induced by other factors, such as the tube geometry and microbubble concentration.

Fig. 8

(a) Normalized UMF signal from fluorescent microbubbles filled in a microtube through water as a function of ultrasound pressure. The arrows in the inset indicate that the ultrasound transducer was focused on three different locations on the tube. (b) Normalized UFM signal from fluorescent microbubbles filled in a microtube through a 2-mm-thick intralipid slab phantom as a function of the lateral location of the 1-MHz ultrasound transducer. The dotted square line shows ultrasonic echo data that were recorded based on the conventional pulse-echo method. The solid diamond line shows the UMF signal. The arrow in the inset indicates that the transducer was scanned across the tube repeatedly for three times.

In the second experiment, the slab phantom shown in Fig. 4 was replaced with the intralipid phantom. The ultrasound pressure was fixed to 405 kPa. Figure 8(b) displays the UMF signal strength when scanning the ultrasound transducer across the tube; the inset of Fig. 8(b) shows the measurement configuration. The dotted square line shows ultrasonic echo data from the empty PDMS tube that were recorded based on the conventional pulse-echo method. The solid diamond line shows the UMF signal from the tube through a 2-mm-thick intralipid slab phantom. The two data sets were normalized and displayed together. The results show a similar FWHM of $\sim 2\mathrm{mm}$, which indicates the feasibility of detecting UMF signal using microbubble contrast agents in scattering media.

Generally, the UMF modulation efficiency in the bulk solution deteriorated because of the increase in the unmodulated fluorescent emissions from the large illuminated area and unattached fluorophores. In addition, microbubble oscillation behavior could be constrained by the nearby boundaries in the surrounding environment,^43–45 which could also affect the overall modulation efficiency. The employment of a gated integrator greatly increased the detection sensitivity and system signal-to-noise ratio. It is worth mentioning that when choosing contrast agents, some factors need to be taken into consideration: (1) strong modulation efficiency is necessary to extract UMF from background; (2) asymmetric UMF oscillation is required for the gated integration detection; and (3) relatively strong fluorescent signals are necessary for the UMF detection in the scattering media.

4.5.

Further Discussions about the Fluorescent Microbubbles and the Imaging Systems

The previously developed UMF contrast agent⁵ has several limitations because the fluorophores were attached to the microbubble shell via the lipophilicity. First, the fluorophores can only be selected from the family of lipophilic dyes that have high affinity to the microbubble phospholipid shell. Unfortunately, the number of lipophilic dyes is limited. Second, the shell material of the adopted microbubbles has to be lipid. These limitations prevent the majority of fluorophores (not lipophilic) and microbubbles (with nonlipid shells) from being used as UMF contrast agents. The strategy of labeling fluorophores on the surface of a microbubble via different chemical or biochemical interactions, such as amine-NHS and biotin-streptavidin, can significantly broaden the selection of fluorophores and microbubbles. Currently, fluorophores and microbubbles attached with various functional groups have been widely reported in the literature and many of them are available commercially.^39,46–48 Thus, the synthesis of microbubble-based UMF contrast agents become straightforward, which is necessary for bringing this technique into practical application. More importantly, simultaneous multimolecule targeted imaging becomes straightforward by using fluorophores with different excitation and emission wavelengths, which will be investigated in the future.

The two imaging systems shown in Figs. 3(a) and 4(a) are unique in the following respects: (1) the combination of the ultrasonic and optical systems enables us to optically investigate the ultrasound-induced microbubble oscillation and UMF signal; (2) the high temporal ($<1\mu \mathrm{s}$ or $<1\mathrm{ns}$) and spatial ($<1\mu \mathrm{m}$) resolutions enable us to investigate the fast oscillation of individual microbubbles at a megahertz ultrasound frequency and the fluorescence lifetime on an individual microbubble’s surface at a level of nanoseconds; (3) the capability of both white light and fluorescent detections enables us to investigate the bubble oscillation, fluorescence lifetime, and UMF signal from the same individual microbubble; and (4) the high (optical and electronic) sensitivity enables us to detect the weak UMF signal from the scattering medium.