1. Introduction

Photoacoustic lifetime (PALT) imaging is an attractive neo-molecular imaging technique [1–4]. By measuring a specific molecule’s optical absorption in the excited state through a pump-probe-based photoacoustic method, this hybrid imaging technique realizes high specific molecular/functional detection at a centimeter-scale penetration depth [5,6], bringing it excellent prospects in a broad spectrum of biomedical applications.

Conventional photoacoustic imaging (PAI) maps a chromophore in the biological tissue by directly collecting the photoacoustic signals generated by the laser pulses, whose wavelengths are set according to the absorption spectrum signatures of the chromophore [7–10]. However, the strong background signal from the blood significantly reduces the sensitivity and specificity of the target chromophore and thus lowers the reliability of the acquired molecular/functional images [11–13]. Although spectroscopic methods can potentially unmix different contrast agents and remove the background signal, their quantitative accuracy cannot be guaranteed in deep tissues due to the difficulties in compensating the optical influences for different wavelengths [5,14].

In contrast, PALT imaging is mainly based on the pump-probe detection of the triplet lifetime of a phosphorescent dye with two pulsed lasers [2–6,15]. The first pulsed laser pumps the dye molecules to an excited triplet state, and the second pulsed laser of a different wavelength generates photoacoustic signals for detection. By subtracting the photoacoustic signal before and after the pump pulse, the resulting image is proportional to the optical absorption by the triplet-state dye molecules. This means PALT imaging is of high specificity and free of background signal intervention. More importantly, by adjusting the pump-probe delay, a series of photoacoustic images can be generated, and the pixel-wise decay of the photoacoustic amplitudes represents the triplet-state lifetime of the dye. Because the lifetime is an intrinsic parameter of the dye molecule, PALT provides robust molecular/functional images insensitive to dye concentration, excitation intensity, and light absorption in tissue.

Methylene blue (MB) is an FDA-approved water-soluble dye commonly used in PALT imaging [4,6,15]. It can be optically activated into a triplet state with a 650 nm laser, in which its optical absorption peak changes to 810 nm (which can be used as the wavelength of probe pulse). The lifetime of this triplet state reaches about 79.5 µs due to a parity-forbidden transition mechanism [4]. However, oxygen can significantly shorten the relaxation of MB from this triplet state to the ground state through a fret phenomenon. Thus, the lifetime of the triplet state MB is highly sensitive to the oxygen partial pressure (pO₂), a critical functional parameter in the tissue. Measuring the lifetime of triplet MB with PALT can provide information about tissue hypoxia and evidence of oxygen availability in the circulatory system, which is crucial in photodynamic therapy (PDT) [15].

However, the transient triplet differential (TTD) based PALT generally needs more than 100 times averaging to suppress the laser fluctuation and the noise from the data acquisition system. Besides, a typical PALT imaging system contains a transducer array and two optical parametric oscillator (OPO) lasers. The data acquisition is composed of three sequences with different laser excitation modes. Switching the OPO lasers between the three sequences with conventional methods such as electrical shutter takes considerable time and needs complex programming control. It isn’t easy to fully synchronize the two OPO lasers and the data acquisition with this kind of laser switching method. Currently, most PALT imaging systems carry out the three data acquisition sequences separately, and the switching of the OPO lasers is taken manually in between [2,4,16]. This causes the data acquisition time of PALT generally be long. More importantly, this kind of data acquisition method is easily influenced by the long-time power drifts of the OPO lasers. So far, only 2D PALT imaging results have been reported. The inability of PALT-based fast 3D background-free imaging and pO₂ imaging has dramatically hindered the application of this imaging technology.

To overcome the abovementioned problems, this work proposes a novel automatic interleaved data acquisition method [17–19] for controlling the two OPO lasers to speed up the acquisition. The advantage of this method was demonstrated with PALT-based 3D background-free imaging and pO2 imaging in-vivo on a rat-bearing tumor model.

2. Methods

2.1 Basic principle of the TTD-based PALT imaging method

PALT with phosphorescence dye such as MB has been introduced in greater detail elsewhere, and here is a brief review. MB is widely used in clinical diagnostic and therapeutic applications. Besides, it is relatively stable in stained tissue for up to several hours, much longer than the PALT data acquisition time. As seen in the energy diagram of PALT in Fig. 1(a), the MB molecule in the ground state (S0) has a strong absorption peak around 650 nm (extinction coefficient is 72000 cm⁻¹/M). After being excited into the excited singlet state (S1) by the pump laser, it undergoes an “intersystem crossing” (ISC) energy decay into a triplet state (T1) with a high quantum yield of 0.5 [15,20]. Compared with the fluorescence lifetime (usually less than 100 ns [15]), this T1 state has a relatively long phosphorescence lifetime due to the spin-forbidden nature of the T1 to S0 transition. However, oxygen can significantly reduce this lifetime by collisional quenching. The conversion formula of the phosphorescence lifetime T and the tissue pO₂ is by the following Stern-Volmer equation [4,6]:

 

Fig. 1. The TTD-based PALT imaging method with interleaved data acquisition. (a) The principle of the TTD-based PALT imaging method and the diagram for demonstrating the generation of photoacoustic signals from the triplet state of MB. (b) The advantage of the interleaved data acquisition method. Here is an experimental recording of the 6000 consecutive pulses from an OPO laser (black line, 20 Hz). We see the fluctuations in laser pulse energy with 50 times averaging (red line, 120 data points). If we do the time differential of the mean pulse energy (50 times averaging) like the conventional acquisition method, we see the fluctuations in the differentiated mean pulse energy reach 15%. If we do time differential like the interleaved acquisition method, the fluctuation is only about 2%. (c) The schematic of the APLI system. The system mainly comprises two OPO lasers, one Verasonics data acquisition system, one digital delay/pulse generator, and one transducer array. (d) The synchronization control of the interleaved data acquisition method. We applied a 1 ms delay to the pump (probe) laser when we only needed to acquire the probe (pump) laser’s data so that the OPO lasers could work steadily at 20 Hz.

Download Full Size | PDF

Here T° is the phosphorescence lifetime when the pO₂ is 0 mmHg, and k_Q is the quenching rate constant. For MB, T° is 79.6 µs, k_Q is 0.0036 µs⁻¹mmHg⁻¹ [4]. Applying a probe pulse at 810 nm allows the MB molecule in the T1 state to be excited to a higher state, which generates a photoacoustic signal during the relaxation to T1. Because the pump laser and the probe laser will also give photoacoustic signals due to the background absorption, the photoacoustic signal from the T1 state MB can be calculated as:

Here S_{pump + probe,t} is the photoacoustic signal obtained with both the pump and probe lasers on, and t indicates the delay time of the pump laser ahead of the probe laser. Since the probe laser usually fires at the start of the data acquisition, the time t is also the firing time of the pump laser ahead of the data acquisition. S_pump,t is the collected photoacoustic signal when the pump laser emits alone, and S_probe is the photoacoustic signal when the probe laser is on and the pump laser is off. Because only the T1 state MB can generate this transient TTD signal, the resulting image is highly specific and background-free without intervention from the endogenous contrast agents.

By varying the delay time t between the pump and the probe laser pulses, a pixel-wised decay curve of the TTD signal can be obtained, which gives the PALT image of the phosphorescence lifetime by fitting the following exponential decay function:

Here A(t) is the TTD signal with the delay time of t, A° is the fitted TTD amplitude, and T is the estimated phosphorescence lifetime. For reconstructing the PALT-based pO₂ image with Eq. (1), the coefficient of determination R² was computed at each pixel to reflect the goodness of the fit, and the maximum value of the fitted A° image was normalized to 1. Then an A° > 0.01 and an R² > 0.9 were used as selection criteria to ensure the validity of lifetime estimates. Only pixels that met both criteria were converted to pO₂ values. The default delay time of the background-free TTD images was 0 µs in this work, otherwise noted.

2.2 Interleaved data acquisition method

As mentioned above, to measure the transient optical absorption of the MB in the triplet state, three sequences should be performed: S_{pump + probe,t}, S_pump,t, and S_probe. Because this TTD signal is generally much smaller than the background signals, all three sequences require multiple averaging. Conventionally, the averaging of the three sequences were executed sequentially. For example, the S_{pump + probe,t} is first averaged N time with both OPO lasers on, and the corresponding delay t is applied. Then S_pump,t is averaged N times with the probe laser blocked manually or with an electrical shutter. Finally, S_probe is averaged with the probe laser on and the pump laser blocked.

However, this data acquisition method is generally realized by manually turning off or blocking the lasers. Or, it may be automatically accomplished by using electrical shutters but requires complex programming control. More importantly, it assumes the laser pulse energy is independent of each other, which fluctuates around a mean value. This means that the mean power of the pump laser is expected to be the same in the sequences of S_{pump + probe,t} and S_pump,t, and this is the same case for the probe laser in S_{pump + probe,t} and S_probe. Such an assumption neglects the long-time fluctuations of the laser power, so the accuracy of the TTD image is generally low with this kind of conventional acquisition method. Therefore, this work proposes an interleaved acquisition method for S_{pump + probe,t}, S_pump,t, and S_probe. In this method, the acquisitions of the three different signals are performed pulse-by-pulse, giving a TTD image in a group of three triggers. The averaging of the TTD image is carried out by repeating this kind of group acquisition.

For better illustration, we recorded the pulse energy change of an OPO laser (650 nm, 20 Hz) for 5 min, as seen by the black curve in Fig. 1(b). It’s seen that the fluctuations still reached 15% after 50 times averaging (the red curve). We divided these 6000 pulses into two groups A and B in two different ways to simulate the conventional and interleaved acquisition method. We termed the mean energy of every 50 pulses in the two pulse groups as GA and GB, respectively. Then we calculated the change of their differential 2(GA-GB)/(GA + GB) with time. This is analog to calculating the difference between the mean pump laser power in S_{pump + probe,t} and S_pump,t (or the mean probe laser power in S_{pump + probe,t} and S_probe). As we see, the fluctuation of the mean pulse power change with the conventional acquisition method still reaches about 15% after 50 times averaging (the yellow line), but it drops to 2% with the interleaved acquisition method (the green line).

2.3 System setup and synchronization control

As shown in Fig. 1(c), the PALT system mainly consists of two tunable OPO lasers, a digital delay/pulse generator (DG645, Stanford Research Systems, CA), a stepper motor, a 7.8 MHz ultrasound linear array (L11-5v), and a 256/256 channel Verasonics ultrasound imaging system. One OPO laser (SpitLight OPO 600 broad-band, pumped with 355 nm, 20 Hz) is operated at 650 nm for MB excitation, and the other OPO laser (SpitLight OPO 600 mid-band, Innolas, München, Germany, pumped with 532 nm, 20 Hz) was tuned to 810 nm. The pump and probe lasers were coupled into a 1 × 2 fiber bundle with a dichroic mirror. The output end of the fiber bundle was formed into two rectangles of 1.5 mm × 3.5 cm, fixed at the two sides of the linear ultrasound array, as seen in Fig. 1(c). The resulting fluence on the sample was 15 mJ/cm² for the pump laser and 23 mJ/cm² for the probe laser, which were all within the ANSI limits [21].

To realize the interleaved acquisition of the three sequences, the Verasonics system not only determines the timing of the DAQ but also sends out a 20 Hz signal to synchronize the DG645, which triggers the two lasers and controls their pulse delay. We used four channels of the DG645, with each two as a group to control one OPO laser. The time delay between the flash lamp and the Q-switch of each laser was maintained at 215 µs. As seen in the data acquisition scheme of Fig. 1(d), when collecting the S_{pump + probe,t} at the first trigger pulse from the Verasonics system to the DG645, the pump laser was shot ahead of the probe laser with a time of t. The data acquisition (DAQ) was carried out simultaneously with the probe laser fires. Then, when the next trigger occurred to collect the S_probe background signal, the probe laser was shot at the same time as the DAQ, and the pump laser was fired one millisecond after the DAQ. In this way, both OPO lasers are operated at 20 Hz, which is the prerequisite for the stable operation of the OPO lasers. Still, the background signal of the pump laser will not influence the S_probe background signal since the time delay of the pump laser is much longer than the data acquisition length.

Similarly, when collecting the S_pump,t background signal at the third trigger, the pump laser was shot with a time of t ahead of the DAQ, and the probe laser was fired one millisecond after the DAQ. After this, the S_{pump + probe,t} is collected at the fourth trigger, and S_probe is collected at the fifth trigger. This data collection cycle goes on until the average time is reached. Then, the delay of the two OPO lasers is adjusted for lifetime imaging, or the step motor drives the ultrasound array to scan the next position. The Verasonics system monitors the triggering of the DG645 and controls its delays of the four channels with an RS232 serial port through an external function, which also governs the step motor. The ultrasound images were obtained using the synthetic aperture imaging method at the beginning of the PALT imaging. The default averaging time in this work was 50, otherwise noted. The total scanning time for such a 50-times-averaging data set was only 7.5 seconds, which was limited by the OPO lasers’ repetition rate.

2.4 Experimental designs

The proposed method was validated with two tube phantom experiments and three in-vivo experiments. In the first tube phantom experiments, there were four PVC tubes placed perpendicular to the transducer array. The concentrations of MB for the four tubes were 0, 100, 200, and 400 µmol/L, and the first three tubes were mixed with bovine hemoglobin (120 mg/mL) to simulate the blood, except for the fourth tube with the 400 µmol/L MB. The TTD images obtained with the conventional method and the interleaved data acquisition method were compared. In the second tube phantom experiment, there were three tubes imaged. The first tube contained air-saturated MB solution, so it was with a high pO₂. The second tube was with about 10 times diluted black ink. The third tube was also filled with MB solution but de-oxygenated by pumping nitrogen for more than 15 minutes to produce a low pO₂. The MB concentrations of the two MB tubes were both 400 µmol/L. We show the PALT and pO₂ images of the three tubes, to demonstrate the ability of the system to distinguish different pO₂ levels.

The in-vivo experiments were carried out with C57BL mice for about eight weeks. The mice were mounted on a 45° tilted acrylic holder, with hair on the hindlimb removed, and kept anesthetized with a dose of 2% isoflurane and a flow rate of about 0.6 L/min. In the first in-vivo experiment, an amount of 0.2 mL MB of 5 mM was injected into the hindlimb of the mouse. Data acquisition was initiated after 10 to 15 min to allow the dye to diffuse sufficiently in the tissue. In this experiment, 0.1 mL carbon nanoparticles (CNPs) suspension injection with a concentration of 50 mg/mL solution (Lummy Pharmaceutical Co., Ltd, China) was also injected to generate a strong background signal. We obtained the 2D TTD images to show the ability of background signal removal with our proposed interleaved TTD method. In the second in-vivo experiment, only MB was injected, and we obtained 3D TTD images of the mice’s hindlimbs. In the third in-vivo experiment, a tumor-bearing mouse on the hindlimb was employed for the 2D lifetime and pO2 imaging. The tumor model was developed by subcutaneous injection of 2 × 10⁶ Lewis lung carcinoma cells (LLC) in 100ul PBS into the mouse’s hindlimb. After two weeks of the tumor growth, a tumor diameter of 5 to 10 mm was considered appropriate for further experiments. MB was injected into the tumor to obtain the pO₂ with our proposed interleaved PALT method. All animal experiment procedures were approved by the Department of Laboratory Animals of Central South University.

3. Results

Figure 2 compares the TTD imaging results between the conventional and interleaved data acquisition methods for background removal. As shown in Fig. 2(c), the four tubes had an outer diameter of 2.2 mm and an inner diameter of 1.6 mm, and the MB concentrations were 0, 100, 200, and 400 µmol/L, respectively. Only the tube with 400 µmol/L MB was not mixed with hemoglobin, showing a lower signal in Figs. 2(a) and (b) due to the lack of high optical absorption from the hemoglobin. The relative signal intensity of this tube is even lower in Fig. 2(b) as compared with Fig. 2(a) because 810 nm is not the absorption peak of MB.

 

Fig. 2. Tube phantom experimental results of comparison between the conventional and interleaved data acquisition method for background signal removal. (a) and (b) are the conventional photoacoustic images with the pump (650 nm) laser and probe (810 nm) laser, respectively. (c) is the photo image of the four tubes. (d) and (e) are the TTD images by the conventional and interleaved data acquisition method. (f) shows the normalized peak photoacoustic intensity for different tubes in (a), (d), and (e). PA, the conventional photoacoustic image; Conv-TTD, the conventional TTD method; Inter-TTD, the interleaved TTD method.

Download Full Size | PDF

Due to the MB concentration distributions in the four tubes, the TTD signal intensity should be close to zero for tube 1 and consistently increase from tube 1 to tube 4. However, the TTD image by the conventional data acquisition method (Fig. 2(d)) fails to show this trend, as seen in the red curve in Fig. 2(f). Comparatively, the TTD image by the interleaved method (Fig. 2(e)) suits the MB concentration distribution well, and its signal intensity for tube 1 is quite close to zero (the green curve in Fig. 2(f)), indicating that the photoacoustic signal of the hemoglobin is almost wholly removed. Numerical analysis shows that the TTD signal intensity of tube 4 is about 330 times higher than that of tube 1 (142.86 to 0.43) in Fig. 2(e), while this ratio is only 5 times (2.05 to 0.43) in Fig. 2(d). These results indicate that the interleaved data acquisition method is much more efficient than the conventional data acquisition method for background signal removal.

Figure 3 is to prove that the proposed interleaved method can effectively suppress the background signal in in-vivo experiments. Here both MB and CNPs were injected into the mouse’s hindlimb, as shown in the photograph of Figs. 3(a) and (b). The position of the 2D cross-sectional images and the drug injection sites are noted in Fig. 3(b). Figure 3(c) is the ultrasound image. Figures 3(d) and (e) show the conventional PA images with the pump and probe lasers, respectively. Compared with Fig. 3(d), because the absorption of MB is much lower than CNPs for the probe laser, only the CNPs’ signal is clearly noted in Fig. 3(e). With the interleaved data acquisition method, the distribution of MB is revealed in Fig. 3(g), where the background signal from the blood and CNPs are removed. In this image, we can see the MB signal indicated in Fig. 3(d) and some weak signals from the MB near the tissue surface. Figure 3(h) is the overlayed image of Fig. 3(g) and the ultrasound image. Comparatively, the background signal removal is not complete with the conventional TTD data acquisition method in Fig. 3(f), so the signal from the CNPs remains in the image (as indicated with a solid red circle), which is good evidence that the interleaved TTD method performs better than the conventional TTD method. Furthermore, it’s noticed that with the interleaved TTD method, the images with 25, 15, and 10 averaging times are close to the TTD image in Fig. 3(h) (averaged 50 times). The structural similarity (SSIM) values between these four images are all around 0.8. The SSIMs between Figs. 3(i)-(k) and Figs. 3(h) are 0.841, 0.832, and 0.797, respectively, with a slightly decreasing trend with the reduced averaging numbers. This proves the excellent stability of the interleaved TTD method and implies that this new method can reduce the averaging time to improve the scanning speed compared with the conventional TTD method.

 

Fig. 3. In-vivo TTD imaging results of a mouse hindlimb tumor with both MB and CNPs injected. (a), the photo images of the experimental design. (b) Enlarged image of the mouse’s hindlimb after drug injection. (c) Ultrasound image of the hindlimb, whose position is noted with a dotted white line in (b). (d) and (e) The conventional PA images with the pump laser and probe laser wavelengths, where the signal from the MB and CNPs are indicated. (f) The conventional TTD image with 50 averaging times. (g) The interleaved TTD imaging results of the MB, with 50 averaging times. (h) Overlayed image of (e) onto the ultrasound image. (i)-(k) The interleaved TTD imaging results with 25, 15, and 10 times of averaging. US, ultrasound image.

Download Full Size | PDF

Figure 4 shows the results of the second tube phantom experiment to demonstrate the system’s ability for distinguishing different pO₂ levels with the interleaved data acquisition method. Figures 4(a) and (b) are the ultrasound and photoacoustic (with the pump laser) images. From left to right, the three tubes are MB with high pO₂, the ink (acting as a control target), and MB with low pO₂, respectively. In the TTD images of Figs. 4(d)-(f), only the signal of MB can be seen, despite the three tubes having similar signal intensity in the conventional photoacoustic images in Figs. 4(b) and (c). These three images show a consistent decrease in the MB tubes’ TTD signals with the increasing time delay. The overall changes of the two MB tubes in the peak TTD signal with the time delay (the solid black dots) are shown in Fig. 4(g). The two curves are further fitted with Eq. (3) to calculate the lifetime, as seen in the solid lines (red for the MB tube with high pO₂, and green for the MB tube with low pO₂). The coefficients of determination R² for both curves reached 0.95, which proves the excellent stability of the interleaved data acquisition method. However, the calculated lifetimes of the two tubes are notably different, which were about 2.6 and 6.0 µs, respectively, as seen from the lifetime image in Fig. 4(h). The lifetime of the air-saturated MB tube (the most left one) is close to previously reported studies [15]. It’s also noted that the curves in Fig. 4(g) remain above zero even after a long decay time, which is due to the existence of the background noise residual. Figure 4(i) is the image of pO₂ calculated with Eq. (1), which reliably revealed the difference in the pO₂ level of the two tubes.

 

Fig. 4. PALT tube phantom imaging results with the interleaved data acquisition method. (a) Ultrasound image of the phantom. (b) Conventional PA image of the phantom with the pump laser wavelength. (c) Overlayed PA image onto the ultrasound image. (d)-(f) TTD images with a time delay of 1, 4, and 8 µs, respectively. (g) The changes of peak TTD signal amplitudes with the time delay for the two MB tubes. (h) The lifetime image. (i) The image of pO2. Here (a) and (c) share the same Gray colormap for ultrasound imaging, and (b)-(f) share the same Jet colormap for the conventional PA images and the TTD images.

Download Full Size | PDF

Figure 5 demonstrates the lifetime and pO₂ imaging ability of the proposed method in in-vivo experiments. The tumor’s position is indicated with the dotted red lines in the ultrasound image in Fig. 5(a). Figure 5(b) is the conventional photoacoustic image with the pump laser, and Fig. 5(c) is the overlayed photoacoustic and ultrasound image. Figures 5(d)-(f) are the TTD images with different time delays between the pump and probe lasers. The decay rate of the TTD signal varies at different tumor positions, possibly due to the different lifetimes of MB caused by the local distribution of pO₂. Figure 5(g) shows the TTD signal decay of two different pixels, whose positions are indicated in Fig. 5(d), and their lifetimes are calculated to be 20 us and 0.5 us, respectively. Figures 5(h) and (i) show the overlayed lifetime and pO₂ images onto the ultrasound images, respectively. It’s seen that the remaining pixels on the lifetime and pO₂ images are quite continuous. More importantly, the pO₂ image shows a clear low pO₂ in the tumor, compared with the relatively high pO₂ in the tumor’s periphery regions, as indicated in Fig. 5(i). This hypoxia is due to the high oxygen consumption rate of the cancer cells, as explained and validated by many reported studies [22–24]. Here the color is highly saturated in the hyperoxia region in Figs. 5(h) and (i), because the two colormap ranges are intentionally set to highlight the hypoxia regions in these two images.

 

Fig. 5. In-vivo PALT imaging of a mouse-bearing tumor with the interleaved data acquisition method. (a) Ultrasound image of the tumor. (b) Conventional PA image of the tumor with the pump laser wavelength. (c) Overlayed PA image onto the ultrasound image. (d)-(f) TTD images with a time delay of 0 us, 2 us, and 8 us, respectively. (g) The peak TTD signal amplitudes decay with the time delay of two representing pixels, as noted in the TTD images. (h) The lifetime image. (i) The image of pO₂. (b)-(f) share the same jet colormap for conventional PA and TTD images.

Download Full Size | PDF

The proposed automatic interleaved data acquisition method enables the fast 3D TTD imaging of the mouse’s hindlimb, as shown in Fig. 6. Here, the first row (Figs. 6(a)-(d)) shows the 3D ultrasound images, PA images with the pump laser, PA images with the probe laser, and the TTD images with 0 us time delay, respectively. The PA images (Figs. 6(b) and (c)) and the TTD images (Fig. 6(d)) are rendered together with the ultrasound images. The images in the second row (Figs. 6(e)-(h)) and the third row (Figs. 6(i)-(l)) are the corresponding images from two different views. Figure 6(j) shows the photograph of the mouse’s hindlimb after the MB injection. Here the scanning region was indicated in Fig. 6(m), and the whole 3D images are composed of 60 slices (x-z images) aligned along the y-direction with a total scanning length of 6 mm. Figures 6(n)-(q) show four selective cross-sectional TTD images overlayed onto the corresponding ultrasound images, whose positions have been indicated in Fig. 6(m). Compared with the conventional PA images, the TTD images are only sensitive to MB, providing more accurate information on MB distribution. As MB is frequently adopted as an effective photosensitizer in PDT, the 3D TTD images may significantly improve the treatment of PDT by monitoring the 3D distribution of MB in the tumor.

 

Fig. 6. The 3D TTD imaging results of a mouse’s hindlimb with MB injection. (a) The 3D ultrasound image. (b) and (c), the 3D PA images acquired with the pump laser and probe laser overlayed onto the ultrasound image, respectively. (d) The 3D TTD image overlayed onto the ultrasound image. (e)-(h) and (i)-(l), the corresponding images of (a)-(d) at two different views. (m) is the photo image of the mouse’s hindlimb, with the positions of the 2D TTD images of (n)-(q) indicated.

Download Full Size | PDF

4. Discussion

Minimizing the background blood signal for improved sensitivity and specificity has always been challenging in PAI. Reversibly switchable photoacoustic probes can modulate their optical absorption between two states by external stimuli, which provides a valuable way to completely remove the background signal by differentiating the photoacoustic signals of the two different states since the background signal is independent of the external stimuli [11,25].

So far, the study of the reversibly switchable photoacoustic probes is still in the preliminary stages. The two major categories of this kind of probe are genetically encoded photo-switchable proteins [13,25,26] and triplet-state phosphorescence dyes [2,3,16,27]. Although magnetic modulated nanoparticles have recently been invented for background-free photoacoustic imaging [28], the strong magnetic field (as high as 0.4 T) required in this method is difficult to build in ordinary labs. Compared with the genetically encoded photo-switchable proteins, background-free PALT imaging with the FDA-proved MB as a representative of the triplet-state phosphorescence dye has obvious advantages of easy drug administration (or experimental preparation), faster switchable time [13], and higher signal intensity. The high signal intensity of MB is due to the high allowable injection dose and quantum yields. More importantly, by measuring the triplet state lifetime of MB with PALT, the quantitative mapping of the pO₂ in the tissue can be obtained, which is crucial to many related diagnoses and treatments.

PALT imaging with MB or other small organic phosphorescence molecules has already been experimentally validated in several studies. However, this method has not been implemented widely due to its low accuracy and long scanning time. Current PALT imaging generally assumes that the outputs of the two OPO lasers remain constant during the scan, neglecting the long-time perturbation of the laser energy. Thus, the three signal acquisitions in most PALT imaging are performed sequentially [2,4,16]. According to our experiences, this acquisition method cannot guarantee good background signal suppression even with a considerable averaging number. Besides, it’s hard to validate whether the background signal is well suppressed in most in-vivo experiments. Therefore, the current PALT technique is not widely applied due to the lack of reliability and stability. For the same reason, evident exponential decay can only be seen in a few pixels when mapping the pO₂ with PALT, and it’s difficult to explain the obtained image of pO₂.

Another consequence of this conventional data acquisition method is the long scanning time. This is partly because intensive averaging is required with this scanning mode and partially because most existing PALT systems don’t operate automatically. Because the OPO lasers must operate at 20 Hz, both pump and probe lasers will keep irradiation on the biological target. Mechanically redirecting or blocking the laser beam will complicate the synchronization between the lasers and DAQ. In this work, we realized the single acquisition of the pump (probe) photoacoustic signal by applying a time delay to the probe (pump) laser. This enabled the automatic interleaved data acquisition by simply programming the time delays. With this new data acquisition mode, one TTD image with 50 times averaging costs only 7.5 seconds. Comparatively, the conventional TTD method costs about 2.5 minutes manually switch the laser sources [4].

More importantly, the interleaved data acquisition eliminated the long-time perturbation of the laser energy. Ideally, the OPO laser pulses are independent, and each pulse’s energy should fluctuate around the same mean value with the same probability distribution. For such short-term perturbation in the laser pulse energy, the interleaved acquisition method will not differ from the conventional acquisition method. However, because many factors (such as the temperature) may change continuously in the laser, the mean pulse energy expectation of the OPO laser is time-varying. In this case, the interleaved acquisition method can effectively filter out this kind of “long-term” low-frequency fluctuations by doing a time differential, which ensures the accuracy of the PALT and helps to minimize the data averaging time. It’s seen from Fig. 3 that the results after averaging 10, 15, 25, and 50 times show high similarity, which is clear evidence that the new method offers stable results and less scanning time.

With this automatic interleaved acquisition method, we obtained the 3D background-free TTD images for the first time. In the pO₂ images, the selected pixels are more continuous due to the improved stability, and the interpretation of the pO₂ image is more straightforward, which shows good prospects for PALT imaging in biomedical applications. These results imply that compensating for the fluctuation of the laser energy is crucial to the accuracy of PALT imaging. Therefore, we plan to measure the intensity of each pulse and then normalize the corresponding photoacoustic signal with the pulse energy to minimize the fluctuations, which is also expected to reduce the scanning time dramatically. We will also continue obtaining the 3D distribution of pO₂ and monitor its change over a prolonged time. However, it’s noted that as averaging is needed, the motion effects may prevent acquiring high-quality PALT images. In this study, the tumor was implanted in the mice’s legs, so the motion effect could be neglected when the mice were well anesthetized. However, the motion effects due to breathing can significantly deteriorate the PALT images if the tumor is on the mouse’s back. In the following studies, we will record concurrent ultrasound images as guidance to compensate for the tissue displacement during a single TTD acquisition [19,29].

Another shortage of this work is that the MB as a reporter is not bonded to a recognition unit such as a monoclonal antibody or a fragment of nucleic acid for imaging a specific molecular event, which dramatically lowers the significance of the acquired PALT images. Therefore, we will collaborate with chemists to build MB-based probes and evaluate their targeting performance with in-vivo experiments. However, attention should be paid to the shortcomings of MB, including the susceptibility of enzymatic reduction [30], the aggregation-caused quenching, and the aggregation-caused changes in the optical absorption spectrum when synthesizing and testing the MB-based probes [20,31]. In addition, we only show that the proposed method can distinguish different pO₂ values. The quantitative accuracy of this method needs to be investigated in future studies. With all these efforts, the PALT-MB system probably becomes a powerful tool for in-vivo therapeutic and molecular imaging.

5. Conclusion

We have proposed a novel automatic interleaved data acquisition method for TTD-based PALT imaging. This new method not only improved the scanning efficiency but also eliminated the long-time fluctuations of laser energy, which significantly improved the stability of the acquired PALT images and helped reduce the number of averaging times. With this new method, we obtained the 3D background-free TTD images for the first time. We also observed distinct low pO₂ inside the tumor compared to its surrounding regions, which coincides with other reported results. All these results indicate that the proposed method laid a good foundation for the further application of PALT in biomedical applications.