2.1.

Multimodal Imaging System

The schematic of the multimodal system is shown in Fig. 1. The system allowed for three modes of operation: pulse-echo US imaging, PAI, and PALI. The pump laser was a pulsed Nd:YAG laser (Surelite I, Continuum), equipped with a third harmonics generation module, emitting short pulses (5 ns) at 355 nm. An optical parametric oscillator (OPO, MagicPRISM, OPOTEK, Carlsbad, CA) was used in conjunction with the laser to generate a wavelength of 650 nm to match the peak absorption of MB in its ground state. A frequency-doubled Nd:YAG pulsed laser (Brilliant, Quantel, Bozeman, MT, 532 nm, 5 ns) commercially integrated with a second OPO unit served as the probe laser. The integrated laser system (Rainbow, OPOTEK) provided a tuning range from 680 to 960 nm. The output wavelength of this laser system was tuned to 810 nm to match the excited state triplet–triplet peak absorption of MB. The pulse energies at 650 and 810 nm were measured to be 8 and 10 mJ, respectively. A set of prisms was used to align the two beams such that they overlapped at the surface of the imaging object. The light intensities of the two beams were approximately 8 and $13\mathrm{mJ}/{\mathrm{cm}}^2$, which are below the ANSI limit.³⁸

Fig. 1

Schematic of in vivo multimodal imaging system. The system was capable of generating ultrasound (US), photoacoustic imaging (PAI), and photoacoustic lifetime imaging (PALI) images using the same hardware. The animal was illuminated by two laser systems triggered by an FPGA module. Both pulse-echo US and PA signals were acquired by a conventional phased-array US transducer and then amplified, digitized, and stored in the US system. Data were transferred to a PC for image processing and display.

Both the lasers were externally triggered by an field-programmable gate array (FPGA) module (Saxo FPGA board, KNJN, California) at a 10 Hz repetition rate. The time delay between the pump and probe pulses was controlled by a PC via a parallel port connection to the FPGA module. The module also triggered the signal acquisition of a 64-channel US system (OPEN system, Lecoeur Electronique, Chuelles, France). The US system was connected to a 64-element US phased-array transducer (P7-4, ATL). This transducer has 64 elements arranged in a linear array configuration with interelement spacing of 0.18 mm. The center frequency of the transducer is 5 MHz and its bandwidth is 3 MHz. The same transducer was used as a transmitter–receiver in pulse-echo US imaging mode and as a receiver-only in PAI and PALI. All the 64 channels were digitized simultaneously at 12-bit resolution. The amplification gain of each channel can be programmed from 0 to 79.9 dB. The data were stored in an internal buffer and then transferred to a PC for image reconstruction and processing. The 80-MHz internal clock of the US system was utilized as a common time-base for all other clocks and triggering signals in the multimodal imaging system, thereby ensuring minimal jitter between laser pulses and PA signal acquisition.

2.2.

US Imaging

In US imaging mode, the lasers were turned off, and the US system triggering was set to internal. The US images were obtained by the synthetic transmit aperture method.^39,40 At each step, only one channel of the phased array emitted an US pulse, and the reflected backscattering signals were simultaneously recorded on all 64 channels. This step was repeated until the echo signals from every emitting channel were acquired. The signal acquisition time for a complete scan was $<100\mathrm{ms}$. This method generates a data set from which signals corresponding to any pair of emitting and receiving elements can be extracted. For each emitting element, an amplitude image was reconstructed by the delay-and-sum method,⁴¹ followed by Hilbert transformation along the axial direction for conversion to a complex, analytical signal. Finally, all complex amplitude images were summed, and the absolute values were presented using dB scales.

2.3.

PAI and PALI

When working in PAI and PALI modes, both lasers were turned on, and the US system was then set to external triggering mode with the US pulse generator turned off to reduce electronic noise. The PA signal acquired by the US array was averaged over 100 measurements to compensate for the fluctuations in the OPO output energy. A filtered backprojection algorithm was applied to reconstruct the PA images.⁴² The PA signal was filtered by a zero phase-shift finite impulse response filter with bandwidth of 3 to 7 MHz and then processed through a coherent summation procedure to generate the PA amplitude at each pixel in the field of view. Envelope detection was then applied by taking the absolute value of the Hilbert transform of the PA amplitudes along lines in the depth direction. The final step was required for generating a unipolar PA image.

A series of PA signals was obtained with pump–probe delays ($\tau$) of 0.25 μs, 0.5 μs, 1 μs, 2 μs, 4 μs, 8 μs, and 100 ms. The delay was accomplished by changing the time of the pump with respect to the triggering signal to US system. At each delay, PA signals were recorded with the probe beam both on (synchronized with start of US signal acquisition) and off. Taking into account the 10-Hz repetition rate and the time for data transfer, the total imaging acquisition time was $\sim 150\mathrm{s}$. This includes averaging of 100 samples for each PA signal. In order to extract the transient PA signal, the following steps were applied. First, the PA signal corresponding to the 810-nm laser ($S_{810}$) was calculated by subtracting the PA signal generated by the 650-nm laser only ($S_{650}$) from the PA signals generated with both lasers (650 and 810 nm) emitting ($S_{810+650}$) at a specific pump–probe time delay (${\tau }_0$), as described by Eq. (1).

The $S_{810}$ signal had two independent components, a background absorption signal due to light absorption in tissue (mostly in hemoglobin) ($S_{810,\text{background}}$) and a transient absorption signal of MB at ${\tau }_0$ ($S_{810,\mathrm{transient},\tau ={\tau }_0}$). The two components were separated using a background suppression method described by Huang et al.⁴³ in more detail below. The background absorption is independent of the pump–probe time delay. Therefore, the background PA signal was obtained by separating the two individual PA responses (650 and 810 nm) at a long $\tau$ (100 ms), a value that is much longer than the normal range of lifetime of MB ($<100\mu \text{s}$). The transient PA signal was then obtained by subtracting the background PA signal at 810 nm from the total PA signal at 810 nm.

A sequence of PA amplitude images of the transient absorption was then reconstructed from the transient PA signal $S_{810,\mathrm{transient},\tau ={\tau }_0}$. Each image in the sequence corresponds to a different time delay. The following set of time delays was used: 0.25, 0.5, 1, 2, 4, and 8 μs. For each pixel, the lifetime ($T$) and the transient PA amplitude ($A_0$) were computed by fitting the amplitude ($A$) to an exponential decay function of the pump–probe delay $\tau$:

Additionally, the coefficient of determination, $R^2$, was computed at each pixel to reflect the goodness of fit. The amplitude threshold was set to be a quarter of the global maximum amplitude of the PA image. An amplitude above threshold and an $R^2$ value $>0.8$ were used as selection criteria to ensure the validity of lifetime estimates. Only pixels that met both criteria were converted to ${\mathrm{pO}}_2$ values and displayed in a color scale. The Stern–Volmer relationship [Eq. (5)] describing the lifetime as a function of ${\mathrm{pO}}_2$⁴⁴ was used to convert lifetime to ${\mathrm{pO}}_2$:

2.4.

Phantom Imaging

To demonstrate the capability of multimodal imaging, we tested the imaging system using a phantom consisting of two tubes. The experimental setup is shown in Fig. 2(a) and 2(b). Two plastic tubes (inner diameter 0.7 mm and outer diameter 2.4 mm) containing MB aqueous solution at different ${\mathrm{pO}}_2$ values were placed perpendicular to the scanning plane. The solution in each tube was in closedflow circulation with an oxygenation cell, in which the ${\mathrm{pO}}_2$ level was controlled by gas bubbling. One cell was bubbled by air so that the ${\mathrm{pO}}_2$ of the solution was airequilibrated, and the other one was bubbled by pure nitrogen. The US transducer was placed near the tubes, with both lasers illuminating the tubes from the side [left side of Fig. 2(a)]. Both PA and US signals were recorded without moving the tubes and transducer. PA images at a series of time delays were reconstructed, and the lifetime was extracted and converted to ${\mathrm{pO}}_2$ for display.

Fig. 2

Experimental setup and multimodal imaging results of a two-tube phantom experiment. (a) The phantom consists of two plastic tubes containing methylene blue (MB) aqueous solution at high ${\mathrm{pO}}_2$ (150 mmHg, right) and low ${\mathrm{pO}}_2$ (50 mmHg, left). Both tubes were illuminated by the excitation beam (650-nm laser) and probe beam (810-nm laser). (b) Photo of the phantom and the US transducer. UPAT, ultrasound phased-array transducer. (c) Transient PA image of the two tubes by the 810-nm laser at a pump–probe delay of 0.5 μs displayed in a linear scale. The two clusters of PA signals represent the location of the wall–dye interface in the two tubes. The phantom was illuminated by two lasers from the left, resulting in higher PA amplitude in the left tube. (d) PALI image (color) superposed on US (gray scale) image. Inner and outer walls of both tubes are indicated by dashed lines.

2.5.

Animal Preparation

Tumor-bearing nude mice were used as a cancer model in this study. Tumors were induced by injecting LNCaP cells into the hindlimbs. The LNCaP cell line was derived from human prostate adenocarcinoma and has been commonly used in cancer research.^45–47 Specifically, ${1\times 10}^6$ cells suspended in 0.1 mL Matrigel matrix (50% Matrigel and 50% LNCaP growth medium) were subcutaneously injected into the hindlimb of each 24-g nude mouse. Tumors grown for 3 to 5 weeks with a diameter of 5 to 10 mm were considered appropriate for further experiments.

Methylene blue (MB hydrate, Fluka, Sigma-Aldrich, St. Louis, MO) was dissolved in physiological saline (DPBS $1\times$, Mediatech, Manassas, VA) to a concentration of 5 mM. After the animal was anesthetized by ketamine ($100\mathrm{mg}/\mathrm{kg}$) and xylazine (10 mg/kg), the MB solution was injected into the hindlimb. The amount of MB injection was 0.2 to 0.6 mL at each spot (5 to 10 mm between injection spots, total injection volume was limited to 1 mL for a given hindlimb). Data acquisition was initiated after 10 to 15 min to allow the dye to sufficiently diffuse in the tissue. All animal procedures and care were performed according to protocols approved by the University of Minnesota Institutional Animal Care and Use Committee in accordance with federally approved guidelines.

2.6.

In Vivo Oxygen Imaging and Measurement

The animals were divided into two groups: mice with hindlimb tumor (group $A$, $n_A=8$) and control non–tumor-bearing mice (group $B$, $n_B=10$). Two animals from each group were randomly chosen for imaging, whereas the rest were used for direct measurements using an oxygen probe for comparison. In group $A$, both the PALI and oxygen probe were applied to quantify the ${\mathrm{pO}}_2$ of the tumor and normal tissue at the hindlimbs. In group $B$, ${\mathrm{pO}}_2$ of the non–tumor-bearing hindlimb of control mice were measured by both methods. The ${\mathrm{pO}}_2$ values in PALI were obtained from individual pixel values within regions of interest in the images, whereas the ${\mathrm{pO}}_2$ values by the oxygen probe were directly recorded. The statistics of ${\mathrm{pO}}_2$ values obtained by PALI and the probe with three types of tissues (tumor tissue, normal tissue, and control mice) were compared.

The animal chosen for imaging was transferred to the imaging platform as shown in Fig. 3. The imaging platform included an animal holder in order to keep the hindlimb of the animal submerged in water and the head above the surface for normal breathing. The transducer and platform were fixed by mechanical arms so that the relative position of the animal and US transducer did not change throughout the experiment. The water temperature was kept at $35°\mathrm{C}\pm 0.5°\mathrm{C}$ by a digital heating system (T3-150, Transworld Aquatic Enterprises, Inglewood, CA). The US images were generated to provide the anatomical structure of the hindlimb. PA images at each pump–probe delay were obtained and then processed to yield the PALI image. After the imaging procedures, the animal was euthanized.

Fig. 3

In vivo imaging platform for small animals. The nude mouse was fixed within a water tank with its head above the water level. Two laser beams overlapped on the tumor-bearing hindlimb, while the US transducer was positioned next to the illuminated spot and aligned for multimodal imaging. TBM, tumor-bearing mouse; UPAT, ultrasound phased-array transducer. Red dashed line, the 650-nm laser beam; green dotted line, the 810-nm laser beam.

To perform direct oxygen measurements, the animal was first anesthetized and transferred to a heater pad. The ${\mathrm{pO}}_2$ of the hindlimb was then assessed by a commercial single-point oxygen probe. The device consists of a retractable needle-type oxygen sensor (sensor tip diameter 50 μm, surrounded by a 40-mm-long and 0.8-mm-wide syringe needle, OXR50, Pyroscience, Aachen, Germany) and optical oxygen meter (Firesting ${\mathrm{O}}_2$, Pyroscience). The Firesting ${\mathrm{O}}_2$ system is a PC-controlled fiber-optic oxygen meter that measures ${\mathrm{pO}}_2$ by quantifying the oxygen-dependent phosphorescence of a dye-stained polymer membrane located at the tip of the optical fiber. The oxygen sensor was inserted into tissue with the needle. By pulling out the probe in a step-wise manner, ${\mathrm{pO}}_2$ values along the needle path at different depths were recorded. Typically, three needle penetrations were made in one hindlimb and two to four stable ${\mathrm{pO}}_2$ readings were recorded for each penetration. The animal was euthanized after the measurements.

3.1.

Phantom Imaging

An example of multimodal imaging results of a phantom is given in Fig. 2(c) and 2(d). Figure 2(c) shows the PA image of the phantom. Figure 2(d) shows the superposition of the PALI image onto the US image. Both the PA and US images correctly reflect the structure of the phantom, with the PA image showing the optical absorption of the contents inside the tubes and the US image indicating the inner and outer walls. The different levels of ${\mathrm{pO}}_2$ in the two tubes are shown in the PALI image. The ${\mathrm{pO}}_2$ within the left and right tubes were set to 50 and 150 mmHg, respectively. The PALI image clearly shows the levels in the range of 40 to 60 mmHg and 130 to 160 mmHg for the left and right tubes, respectively.

3.2.

In Vivo Imaging

The setup for the in vivo measurements is shown in Fig. 3. PA and PALI images were superposed onto the US image. As can be seen from Fig. 3, the laser beams illuminated only a portion of the hindlimb due to the limited power of the lasers. Therefore, both PA and PALI images covered less area than the US image [Fig. 4(a)]. The PA image shows the amplitude distribution resulting from the background absorption [Fig. 4(b)] and MB’s transient absorption [Fig. 4(c)]. The transient PA amplitudes associated with two pixels, along with their exponential decay determined by curvefitting, are shown in Fig. 4(e). The two pixels correspond to points with high and low ${\mathrm{pO}}_2$. The hypoxic region imaged by PALI [Fig. 4(d)] is consistent with the site of the tumor imaged by US. We have also applied the same technique to the control mice without tumor. The PALI images indicate inhomogeneous tissue oxygenation within the normal muscle tissue, and the ${\mathrm{pO}}_2$ values were within the normal physiological range.

Fig. 4

Multimodal imaging of the tumor-bearing mice. (a) US image of the left tumor-bearing hindlimb of a mouse. The area of the tumor is enclosed by a red dashed line. (b) PAI representing the amplitude of background absorption at 810 nm. Amplitude is displayed in linear scale. (c) PAI of the transient absorption of MB with an 810-nm laser at a pump–probe delay of 0.25 μs. (d) PALI of ${\mathrm{pO}}_2$ in color scale superimposed on US image. (e) Transient PA amplitudes of two representative pixels within the tumor and in normal control tissue, respectively. The triangles and circles are the averaged transient PA amplitudes of tumor and normal control tissue, respectively. The error bars represent the standard deviation of 100 recordings. Both the sets of data were fitted with an exponential curve as shown by the dashed line. (f) Animal after imaging procedure. An arrow indicates the MB stained imaging region. (g) Open-skin view of the tumor site. Note that the tissue was still stained with MB after the imaging process, thereby confirming stable MB staining of the tissue for a period of $>1\mathrm{h}$. The red arrow indicates the site of the tumor.

3.3.

Correlative In Vivo Measurement

We have performed correlative direct measurements using the oxygen needle sensor in tumor-bearing mice and control mice. The total number of valid ${\mathrm{pO}}_2$ readings for tumor tissue, normal tissue of tumor-bearing mice, and control mice was 38, 39, and 121, respectively. The three groups of measurements were then compared with the ${\mathrm{pO}}_2$ values extracted from PALI images. The frequency distributions of the ${\mathrm{pO}}_2$ values are displayed in a histogram shown in Fig. 5(a). The three regions for statistical analysis of ${\mathrm{pO}}_2$ distribution via PALI are labeled in Fig. 5(b) and 5(c). We performed two-sample $t$-tests. The results show a statistically significant difference in ${\mathrm{pO}}_2$ distribution between tumors and other tissues, with $p$ values $<0.05$ by both PALI and the needle probe. Both PALI and direct oxygen measurements confirmed that the tissue oxygen level in the tumors is significantly lower ($\sim 20\mathrm{mmHg}$) than that in normal tissue and control mice.

Fig. 5

Histogram of ${\mathrm{pO}}_2$ in tumor tissue, normal tissue, and control mice. (a) Top row: histogram of ${\mathrm{pO}}_2$ values extracted from PALI. Bottom row: histogram of ${\mathrm{pO}}_2$ values measured by the oxygen probe. Data from tumor tissue, normal tissue (adjacent to tumor site), and control tissue (no tumor present) are displayed in left, middle, and right columns, respectively. (b) PALI showing the regions of tumor (red dashed line) and normal (yellow dashed line) sites in a tumor-bearing hindlimb of a mouse. (c) PALI showing the control tissue of a tumor-free hindlimb. The image does not show a substantial area of low ${\mathrm{pO}}_2$ as compared to Fig. 4(d).