2.1.

Hardware of the System

The diagram of the USF imaging system is shown in Fig. 1(a). The major components have been introduced in our previous publication.^11,12 For integrity, we will very briefly introduce the entire system and emphasize the difference between the current system and the previous ones.^11,12 Similar to the previous ones,^11,12 the current USF subsystem includes the following modules: (1) an ultrasound heating and its driving module, (2) an excitation light source, (3) a sample module, (4) an optical detection module, and (5) a scanning module.

Fig. 1

(a) The experiment block diagram. (b) A schematic diagram of the time sequence of different signals. (c) Sample configuration including $x–y$ (horizontal) and $x–z$ (vertical) regions of USF Imaging. MT: a pulse delay generator serves as an MT (T-1) with 0.1-Hz frequency; FG-1: a function generator for gating the second function generator FG-2 to drive the HIFU transducer module (2.5 MHz; HIFU), and generating a reference signal (RS-2) to LIA-2; FG-2: a function generator for driving the HIFU transducer by generating a 2.5 MHz sinusoidal signal and amplified by an RFPA; FG-3: a function generator to modulate the excitation laser at a 1-kHz frequency and generate a reference signal (RS-1) for the lock-in amplifier; W: a water tank to immerse the HIFU transducer and partially immerse the sample (S); ST: a silicone tube with an inner diameter (ID) of 0.31 mm and an outer diameter (OD) of 0.64 mm; 3D-TS: the three-dimensional (3-D) translational stages; TS-CU: 3-D translational stage motorized control units; CT-1: a collimation tube to focus the excitation laser into an optical bundle (OB-1); CT-2: an optimized collimation tube to guide the collected fluorescence from the optical bundle (OB-2) to the photodetector (PMT) and best eliminate the excitation light and pass the emission light; PMT: a photomultiplier tube to detect the optical fluorescence signal; Preamp: a preamplifier to amplify and filter the signal from the PMT; Laser: a 671-nm laser to irradiate the sample (S); LIA-1: a lock-in amplifier to detect the 1-kHz optical signal; LIA-2: a second lock-in amplifier to detect the signal with the HIFU-modulation frequency; NI-DAQ: a national instrument data acquisition card; CB-1: serial communication bus to control TS-MCU; CB-2: communication bus to transfer data; RS-1: a 1-KHz reference signal for LIA-1; RS-2: a reference signal with the modulation frequency for LIA-2; T-1: a trigger signal for the NIDAQ card; T-2: a single cycle digital pulse signal to trigger the movement of the 3D-TS.

2.2.

System Operational Flow Diagram

The system operational flow diagram is shown in Fig. 2. An MT started the entire system by triggering both the FG1 and the NI-DAQ. The typical duration of the optical signal acquisition was 4 to 8 s. Once all the data were acquired and stored, the MATLAB GUI will generate a voltage pulse via the output port of the NI-DAQ and trigger the Velmex 3-D translational stages to move to the next location. Care must be taken to ensure that the summation of the DAQ acquisition duration (4 to 8 s) and the movement duration of translational stages ($<2\mathrm{s}$) do not exceed the duration between the two adjacent MTs. Hence, for this study, the MT was set to 0.1 to 0.05 Hz frequency (i.e., 10 to 20 s duration). The duration between the two MTs determined the duration between the two scanning locations and thereby the overall duration of the system scan.

Fig. 2

A time sequence diagram of the operation of the USF imaging system.

2.3.

High-Intensity Focused Ultrasound Modulation

The HIFU modulation is schematically displayed in the second row in Fig. 3. The shadowed areas represent that the HIFU is turned on and its duration is indicated as $t_{\text{on}}$. Otherwise, it means the HIFU is off. The period of each cycle ($T$) is also indicated and equal to the inverse of the modulation frequency ($1/f_{\mathrm{M}\text{-}\mathrm{HIFU}}$). The duty cycle is defined as the percentage of the HIFU is turned on in one cycle, which is $t_{\text{on}}/T$ multiplied with 100%. The total HIFU exposure time ($T_{\text{on}}$) is equal to the product between $t_{\text{on}}$ and the number of the cycles.

Fig. 3

A time sequence diagram to show the HIFU modulation and related parameters.

2.4.

Representations of Ultrasound-Switchable Fluorescence Signal Strength

In our previous studies, a single HIFU exposure pulse (with a typical pulse width of 300 ms) was adopted.^11,12 The processing of previous USF data was as follows. The NI-DAQ card acquired the dynamic change of the amplitude of the 1-kHz signal before, during, and after the HIFU heating. The difference between the maximum amplitude and the background amplitude (acquired before HIFU exposure) was used as the USF signal strength at each location. After scanning the HIFU transducer or sample, an image was acquired based on the calculated USF strength.

In this study, because the HIFU exposure was different from that in our previous work,^11,12 the data acquisition and processing were different too. The time constant of the LIA-1 was kept at 10 ms when the HIFU modulation frequency ($f_{\mathrm{M}\text{-}\mathrm{HIFU}}$) was 1 and 2 Hz. The sensitivity of the LIA-1 was kept at 200 to 500 mV nA. The sensitivity of the Pre-Amp is set to $50\mathrm{nA}/\mathrm{V}$ for low-noise mode.

At each location of the HIFU transducer, three sets of data were simultaneously acquired. They were the data acquired from: (1) the Pre-Amp, which included the signals with both modulation frequencies ($f_{\mathrm{M}\text{-}\mathrm{OPT}}$ and $f_{\mathrm{M}\text{-}\mathrm{HIFU}}$); (2) the LIA-1, which was sensitive to the signal with the frequency of $f_{\mathrm{M}\text{-}\mathrm{OPT}}$ (i.e., 1 kHz) and detected its amplitude variation (at the frequency of $f_{\mathrm{M}\text{-}\mathrm{HIFU}}$); and (3) the LIA-2, which was sensitive to the signal with the frequency of $f_{\mathrm{M}\text{-}\mathrm{HIFU}}$ and detected its strength. The relationship among these three sets of data is shown in Fig. 4 and more details will be discussed in Sec. 3.

Fig. 4

An example to show the modulation of USF signal at 1 Hz with three cycles and a 10% duty cycle acquired from different locations: (a) acquired from the Pre-Amp, (b) acquired from the LIA-1, and (c) acquired from the LIA-2.

To quantify the USF strength at each location, Table 1 shows four methods (a to d) to process the acquired data. For data acquired from the Pre-Amp, the amplitude of the Fourier transform of the acquired signal at the frequency of $f_{\mathrm{M}\text{-}\mathrm{HIFU}}$ was adopted to represent the USF strength, which was the method (a) in Table 1 [Note that the raw signal acquired from the Pre-Amp is an amplitude-modulated (at the frequency of $f_{\mathrm{M}\text{-}\mathrm{HIFU}}$) sinusoidal wave (at a frequency of $f_{\mathrm{M}\text{-}\mathrm{OPT}}$)]. For data acquired from the LIA-1, the amplitude of its Fourier transform at the frequency of $f_{\mathrm{M}\text{-}\mathrm{HIFU}}$ was adopted to represent the USF strength, which was the method (b) in Table 1 (Note that the data acquired from the LIA-1 is a signal with a frequency of $f_{\mathrm{M}\text{-}\mathrm{HIFU}}$, which represents the amplitude profile of the signal from the Pre-Amp). For data acquired from the LIA-2, the maximum and the mean of the acquired signal (within a time window of the total duration of the HIFU exposure) were, respectively, used to represent the USF strength, which were the methods (c) and (d) in Table 1.

2.5.

Ultrasound-Switchable Fluorescence Contrast Agents

The USF contrast agent used in this study was $\mathrm{ADP}{(\mathrm{OH})}_2$-encapsulated Pluronic-F98 based thermos-sensitive nanocapsules. The details can be found in our previous publications.¹² The initial concentration of Pluronic-F98 in deionized water is $50\mathrm{mg}/\mathrm{ml}$. The peak excitation and emission wavelengths were around 680 and 715 nm, respectively. The temperature threshold ($T_{\mathrm{th}1}$) is about 28°C, which is slightly lower than our previous results¹¹ and may be because of different fluorophores, some measurement errors and/or bench variation.

3.1.

Demonstration of Ultrasound-Switchable Fluorescence Modulation

To demonstrate how the USF signal is modulated by the variation of HIFU exposure, we show a set of USF signals in Fig. 4, which were acquired, respectively, from the three different locations, as described above. Specifically, the HIFU was turned on and off at a frequency of 1 Hz ($f_{\mathrm{M}\text{-}\mathrm{HIFU}}$) with a duty cycle of 10% and a total duration of 3 s. This means within 1 s, the HIFU was turned on 0.1 s and off 0.9 s and this on-and-off cycle was repeated three times. To be able to clearly visualize the modulation of the USF signal, a relatively high SNR was adopted.

Figures 4(a), 4(b), and 4(c), respectively, show the acquired USF signals from the Pre-Amp, LIA-1, and LIA-2. The signal acquired from the Pre-Amp is an amplitude-modulated 1-kHz signal (i.e., $f_{\mathrm{M}\text{-}\mathrm{OPT}}=1\mathrm{kHz}$ and $f_{\mathrm{M}\text{-}\mathrm{HIFU}}=1\mathrm{Hz}$). The amplitude modulation can be seen from the three large and negative peaks [Fig. 4(a)]. Note that the negative sign is caused by the PMT because usually a PMT generates a negative signal when optical intensity is increased. As discussed above, the LIA-1 functions as a demodulator to extract the variation of the amplitude of the 1-kHZ signal. Figure 4(b) clearly shows the 1-Hz USF signal with three cycles, which represents the profile of Fig. 4(a) after inverting the signal. Similarly, the LIA-2 functions as the second demodulator to extract the profile of the 1-Hz modulation signal. Figure 4(c) displays a quasi direct current signal, which represents the amplitude profile of the signal from the LIA-1 and proportional to the strength of the 1-Hz modulated USF signal.

3.2.

Ultrasound-Switchable Fluorescence Images

Figure 5 shows the normalized 2-D USF images of a silicone tube ($\mathrm{ID}=0.31$ and $\mathrm{OD}=0.64\mathrm{mm}$) embedded in a piece of porcine muscle tissue with a thickness of $\sim 12\mathrm{mm}$. The 2-D USF images are obtained along the $x–y$ plane across the silicone tube, as shown by the red dashed line in Fig. 1(c). Figures 5(a)–5(d), respectively, display the USF images processed using the four methods discussed in Tables 1 and 2. The details about the sample have been described in the section of the sample configuration. The HIFU modulation frequency is 2 Hz with a 10% duty cycle and a total of three cycles (controlled by FG-1). Thus, the total HIFU exposure time is 150 ms (i.e., the sum of three 50-ms on cycles). The driving signal of the HIFU is a 2.5-MHz sinusoidal waveform that is generated by FG-2, gated by FG-1, and amplified by RFPA. The peak-to-peak voltage of the driving signal from the FG-2 is $\sim 0.1\mathrm{V}$, and after being amplified 50 dB by the RFPA, the signal is $\sim 32\mathrm{V}$ at the HIFU transducer.

In general, all four methods clearly showed the tube via the modulation of USF. The full-width-at-half-maximum (FWHM) and SNR of each USF image were quantified and summarized in Table 2. The FWHMs of the USF images fall into the range from 0.95 to 1 mm, which is slightly larger than the tube OD and in agreement with our previous study.¹¹ This is because both studies used the same HIFU transducer, which is the major factor determining the USF resolution. The SNRs of these USF images in Fig. 5 are in the range from 62 to 114. The method (b) achieved the highest SNR of 114. Currently, it is unclear why the SNRs achieved by the methods of (c) and (d) are lower than the methods of (a) and (b). Further studies should be conducted to address this question in the future.

Fig. 5

Normalized 2-D USF images of the small tube embedded in the tissue sample. The images were processed using the four different methods of (a) to (d) in Tables 1 and 2. The HIFU modulation pattern is 2 Hz with a 10% duty cycle and a total of three cycles.

Unlike FWHM, these SNRs are higher than what we achieved in our previous study where no HIFU modulation was adopted and other experimental conditions were similar.¹¹ Specifically, in our previous study,¹¹ when the HIFU exposure was a single 300-ms burst (no modulation), the SNR was achieved $\sim 40$ in the similar sample using a time constant of 30 ms for LIA-1 (only one LIA was adopted in Ref. 11). After using a correlation method in our previous studies,¹¹ the SNR was improved from $\sim 40$ to $\sim 95$, which falls into the range of the current SNR values. This result indicates that the method developed in the current study (i.e., the modulation of USF) is comparable to the correlation method developed in our previous study.¹¹ Mathematically, this is understandable because both modulation and correlation follow the similar rule to recognize signals, which is to encode the USF signal with a unique temporal pattern and use a reference to extract the signal. The method of modulation encodes the USF signal with a specific frequency ($f_{\mathrm{M}\text{-}\mathrm{HIFU}}$) while the correlation method takes advantage of the natural dynamic change of the USF signal to differentiate the USF signal from the background noise. Note that the correlation method in our previous study¹¹ can be considered as a software-based off-line processing method, while the current method encodes the USF signal at a specific modulation frequency and can be considered a hardware-based online processing method. Another major difference is that the method of modulation has a specific modulation frequency so that the detection bandwidth may be narrower than that of the correlation method. Thus, the SNR (or sensitivity) of the modulation method should be relatively higher. If the SNR or sensitivity is similar, the applied HIFU power and/or exposure time should be lower or shorter. This is true if we compare the HIFU exposure parameters in our current and previous studies.¹¹ In the current study, the HIFU was turned on a total of 150 ms during a period of 1.5 s. In our previous study,¹¹ the HIFU was turned on a total of 300 ms while the HIFU exposure power is similar. Thus, the HIFU-induced temperature in the focal volume should be lower in the current study. Finally, we conclude that, compared with the nonmodulation method but processed with the correlation algorithm,¹¹ the current modulation method can achieve the similar SNR but shorten the HIFU exposure time by half. Furthermore, compared with the nonmodulation method without using the correlation algorithm, the current method not only achieves higher SNR but also uses shorter exposure time. Therefore, although the modulated HIFU exposure may lead to lower temperature rise compared with that of continuous wave HIFU exposure, the SNR may not be sacrificed because of the higher sensitivity.

3.3.

Effect of the Modulation Frequency ($f_{\mathrm{M}\text{-}\mathrm{HIFU}}$) and Duty Cycle on Ultrasound-Switchable Fluorescence Images

Figure 6 shows the time sequence diagram of four different HIFU modulation patterns. Figures 6(a) and 6(b) have a modulation frequency of 1 Hz with a duty cycle of 10% and 20%, respectively. Figures 6(c) and 6(d) have a modulation frequency of 2 Hz with a duty cycle of 10% and 20%, respectively. There are three cycles for each HIFU exposure modulation. Figure 7 illustrates the normalized 2-D USF image acquired from the four HIFU exposure patterns. Note that these images were processed using the method of FFT of the LIA-1 output [method (b) in Table 1] since this method has relatively higher SNR in comparison to other methods.

Fig. 6

The time sequence diagram of four different HIFU modulation patterns: (a) 1 Hz with a 10% duty cycle, (b) 1 Hz with a 20% duty cycle, (c) 2 Hz with a 10% duty cycle, and (d) 2 Hz with a 20% duty cycle. All modulation patterns have three cycles.

Fig. 7

Normalized 2-D USF images of the small tube embedded in the tissue sample with four different HIFU modulation patterns: (a) 1 Hz with a 10% duty cycle, (b) 1 Hz with a 20% duty cycle, (c) 2 Hz with a 10% duty cycle, and (d) 2 Hz with a 20% duty cycle. All modulation patterns have three cycles. The images were processed by using the method of the FFT of the LIA-1 signal.

The SNRs are summarized in Table 3 for the four different HIFU modulation exposures. For $f_{\mathrm{M}\text{-}\mathrm{HIFU}}=1\mathrm{Hz}$, when increasing the duty cycle from 10% to 20% [i.e., from (a) to (b) in Table 3], all the SNRs acquired from the four processing methods increase obviously and consistently. This result can be explained using the concept of temperature modulation depth, which is usually defined as the ratio between the difference and the sum of the maximum and minimum temperature values. The above result indicates that the modulation depth of the HIFU-induced temperature increases with the increase of the duty cycle for these two cases. This is understandable because the higher duty cycle means longer HIFU exposure time for each cycle and therefore higher temperature rise and stronger USF signals. On the other hand, the time period when HIFU remains off is longer enough for these two cases to allow the temperature decreasing to generate the modulation pattern. This is because the modulation frequency is 1 Hz and the time period when HIFU remains off in each cycle is 0.9 and 0.8 s for the 10% and 20% duty cycle, respectively. Both values are usually longer than the thermal diffusion time constant in the adopted tissue phantom. Thus, the difference between the maximum and minimum temperature is large, which leads to large modulation depth and therefore high SNR.

Table 3

Calculated SNRs of USF images using the different analysis methods at different HIFU modulation exposures.

However, when the HIFU modulation frequency increases to 2 Hz, raising the duty cycle from 10% to 20% [i.e., from (c) to (d) in Table 3] leads to slight and consistent decrease of SNRs. This result indicates that the modulation depth is reduced when increasing the duty cycle for 2-Hz exposure. This may be because the time period when HIFU remains off is not long enough to allow the temperature decreasing. For the modulation frequency of 2 Hz, the time period when the HIFU remains off in each cycle is 0.45 and 0.4 s for the 10% and 20% duty cycle, respectively. These two values may be close to the thermal diffusion time constant in the adopted tissue phantom. Thus, reducing the time period of off-cycle will limit the cooling time and therefore, the temperature may not be able to reduce down enough. Accordingly, both the modulation depth and SNR reduce.

Comparing all four modulation patterns, the SNRs achieve the highest level for all four processing methods when $f_{\mathrm{M}\text{-}\mathrm{HIFU}}=1\mathrm{Hz}$ with the 20% duty cycle [i.e., (b) in Table 3]. By contrast, the SNRs fall down to the lowest level when $f_{\mathrm{M}\text{-}\mathrm{HIFU}}=2\mathrm{Hz}$ with the 20% duty cycle [i.e., (d) in Table 3]. Last, the SNRs of the remaining two cases [i.e., (a) and (c) in Table 3] are in between and also their SNRs are comparable. Therefore, the modulation pattern (a), $f_{\mathrm{M}\text{-}\mathrm{HIFU}}=1\mathrm{Hz}$ with the 20% duty cycle, is the best choice for this study. In addition, comparing all four processing methods, the FFT of the signal from the LIA-1 always provides the highest SNRs, followed by the methods of the FFT of the filtered USF signal. The SNRs of the remaining two methods, where two LIAs were used, are relatively lower for all the four HIFU exposure patterns. The reasons are unclear and should be investigated in the future.