2.1.

System Architecture

Figure 1 illustrates the overall architecture of the high-speed IVPA system which can perform both IVPA and IVUS imaging. A ns-pulsed OPO laser (EKSPLA, NT242) with a repetition rate of 1 kHz, operating at a tunable wavelength ranging from 1185 to 1235 nm (a lipid absorption band), was used for photoacoustic excitation. The pump source in NT242 is a diode-pumped solid–state Q-switched laser operating at 1064 nm. Upon third harmonic generation, the ultimate pump energy is about 3.5 mJ at 355 nm (used as the pump wavelength for the entire wavelength tunable band). The efficiency of the OPO crystal is about 4% at 1210 nm. The output laser beam was reshaped by an iris, attenuated by a neutral density filter, and then focused by a condenser lens into the multimode fiber (Nufern, MM-S105/125-22A) of the catheter. A photodiode (PD) (Thorlabs, SM05PD4A) was used to record the fluctuation of the laser pulse energy for fluence compensation during postexperiment data processing. The laser pulse energy out of the catheter was maintained to be $\sim 30\mu \mathrm{J}$ throughout the study in this paper. A custom-made miniature single-element ultrasonic transducer (Blatek) was used to detect the photoacoustic waves from the sample as well as to perform pulse-echo ultrasound imaging (a short delay of $\sim 5\mu \mathrm{s}$ was set to separate each photoacoustic and ultrasonic A-line). Both the photoacoustic and ultrasound signals were amplified (39 dB) using an ultrasound pulser receiver, and then digitized with a data acquisition card (12-bit digitization; 250-MHz sampling rate) in a personal computer. To form an A-line image, the acquired (photoacoustic and ultrasonic) signals were band-pass filtered (1–50 MHz for photoacoustic signal and 10–50MHz for ultrasound signal, with a 128-th order zero-phase forward and reverse finite impulse response filter), followed by a Hilbert transform for envelope detection. Finally, the image was converted into a polar coordinate for display (signal processing performed with MATLAB®). To acquire cross-sectional images (B-scans), the catheter was rotated at a speed as fast as 300 RPM via an optical-electric rotary joint. Furthermore, three-dimensional (3-D) images can be obtained via a pull-back stage driven by a step motor. As a demonstration, a video was recorded in real time while the system was working at a B-scan rate of 5 Hz (with 200 A-lines per B-scan) (Video 1).

Fig. 1

Illustration of the high-speed intravascular photoacoustic (IVPA) system. (a) Overall architecture of the system. (b) Enlarged view of the dashed box in (a). (c) Video of the high-speed IVPA system (Video 1, MOV, 10. 2 MB) [URL: http://dx.doi.org/10.1117/1.JBO.20.6.065006.1]. Optical parametric oscillator (OPO); neutral density filter (NDF); beam splitter (BM); photodiode (PD); ultrasonics (US); photoacoustics (PA); multimode fiber (MMF); 3-D Scanner, consisting of an optical-electric rotary joint, a motorized rotary stage and a motorized pull-back stage; PR, ultrasound pulser/receiver; MD, motor driver.

2.2.

Catheter Design

To enable the use of a high-repetition-rate OPO laser—which is of a relatively low energy output—for high-speed IVPA imaging, we designed an IVPA/IVUS dual-modality imaging catheter using quasifocusing light illumination. As shown in Fig. 2, the catheter consists of a multimode fiber, a gradient-index (GRIN) lens (GRINTECH), a prism, and a high-frequency miniature ultrasound transducer, developed based on the design of our previous work.¹⁷ Figures 2(a) and 2(b) show the schematic and the photo, respectively, of the distal part of the catheter. The outer diameter of the catheter is 0.9 mm. All the optical and ultrasonic components were ultimately housed in a metal housing. To increase the laser fluence in the targeted imaging region, the excitation laser beam was quasifocused onto the surface of the imaging target using a custom-fabricated GRIN lens of 0.5 mm in diameter. Specifically, the cross-sectional diameter of the laser beam out of the catheter tip was confined to be $\sim 0.3\mathrm{mm}$ within a 3-mm distance from the GRIN lens. The laser beam was then delivered onto the sample surface by a mirror with an angle of 35 deg [Fig. 2(a)]. A metal ferrule was used to ensure that the fiber and GRIN lens were coaxially aligned. A custom-made single-element ultrasonic transducer (dimension: $0.6\times 0.5\times 0.2{\mathrm{mm}}^3$; made from lead zirconate titanate composite) with a center frequency of 40 MHz, a pulse-echo fractional bandwidth of 60%, and an insertion loss of $-72\mathrm{dB}$ was sequentially aligned with the optical components, while a metal base was used to position the ultrasonic transducer for effective detection of the photoacoustic waves. The illumination angle of the laser out of the catheter tip is 16 deg in water, resulting in an optical and acoustic beam overlap range of $\sim 3\mathrm{mm}$ (from 1.5 to 4.5 mm, measured from the center of the catheter), as shown in Fig. 2(c). A flexible coil that can transmit the rotational torque from the rotary stage was used to house the rest of the catheter (besides the distal end).

Fig. 2

Distal end of the quasifocusing IVPA catheter. (a) Overall schematic. (b) Top view of the catheter. (c) Overlap between the acoustic and laser beams.

To estimate the laser fluence enhancement with such a design compared with that in conventional IVPA systems, a Monte Carlo simulation of photon transport in biological tissue was performed. The simulation code was modified based on the Monte Carlo modeling of light transport code developed by Lihong Wang. A three-layer tissue model, consisting of a 0.06-mm thin-cap layer (${\mu }_a=0.1{\mathrm{cm}}^{-1}$, ${\mu }_s=100{\mathrm{cm}}^{-1}$, $g=0.9$), a 1-mm lipid layer (${\mu }_a=1.1{\mathrm{cm}}^{-1}$, ${\mu }_s=100{\mathrm{cm}}^{-1}$, $g=0.9$), and a 3.94-mm elastin layer (${\mu }_a=0.1{\mathrm{cm}}^{-1}$, ${\mu }_s=50{\mathrm{cm}}^{-1}$, $g=0.9$) was used.^18,19 The illumination spot sizes at the surface in the two cases (quasifocusing and broad illumination beams) were set as 0.3 mm [Fig. 3(a)] and 1.2 mm [Fig. 3(b)] in diameter, respectively, while the incident photon energy ($2\times {10}^6$ photons) was maintained to be the same. Figures 3(a) and 3(b) show the optical fluence distribution in the two cases, while Figs. 3(c) and 3(d) are the quantification results along the depth direction of the central axis. From Fig. 3, it can be seen that the laser fluence of the quasifocusing beam is significantly higher than the broad illumination case in the “lipid core” region of the numerical phantom (from 0.1 to 1 mm). As the simulation here does not account for the illumination angle difference between the two laser beams (approximately parallel beam for quasifocusing versus divergent beam for broad illumination), the actual difference of the laser fluence between the two cases should be more significant, further favoring the quasifocusing beam design.

Fig. 3

Monte Carlo simulation of photon transport in a numerical plaque phantom. (a) and (b) Laser fluence distributions with quasifocusing and broad illumination beams, respectively. (c) Quantitative comparison of the laser fluence along the central axis between the two cases in (a) and (b). (d) Ratio of the laser fluence between the cases in (a) and (b).

3.1.

Resolution and Signal-to-Noise Ratio Measurement

We characterized the performance of the high-speed IVPA imaging system using a variety of targets, including a black tape of $\sim 800\mu \mathrm{m}$ in width and three carbon fibers of diameters ranging from $20–100\mu \mathrm{m}$. First, it can be seen that all the targets are imaged with both IVPA and IVUS [Figs. 4(a) and 4(b)]. The dynamic range is $\sim 34\mathrm{dB}$ in both the IVPA and IVUS images. To quantify the spatial resolution, one carbon fiber of $\sim 20\mu \mathrm{m}$ in diameter was used as the target. The axial resolution and transverse resolution as a function of depth are shown in Fig. 4(c). From the figure, it can be seen that, in general, the transverse resolution of IVPA is better than IVUS at various depths, presumably due to the more strictly confined optical excitation area [Fig. 2(c)]. The axial resolution of IVPA is quite similar to that of IVUS ($\sim 100\mu \mathrm{m}$ at various depths) as it primarily depends on the bandwidth of the ultrasound transducer. Figure 4(d) shows the signal-to-noise ratio (SNR) of IVPA and IVUS measurements as a function of depth (axial position) using a carbon fiber as the target.

Fig. 4

Photoacoustic and ultrasonic resolution and signal-to-noise ratio (SNR) of the high-speed IVPA system. (a) IVPA and (b) intravascular ultrasound (IVUS) B-scan images of the phantom with targets of different sizes. (c) Resolution of IVPA and IVUS at different depths. (d) SNR of IVPA and IVUS as a function of depth. (1), black tape ($\sim 800\mu \mathrm{m}$); (2), carbon fiber ($\sim 100\mu \mathrm{m}$); (3), carbon fiber ($\sim 20\mu \mathrm{m}$); (4), carbon fiber ($\sim 100\mu \mathrm{m}$).

3.2.

Imaging of a Stent

To demonstrate the potential of the high-speed IVPA system for 3-D intravascular imaging, an iliac stent (Abbott) inside a healthy rabbit vessel was imaged with both the photoacoustic and ultrasonic modalities.²⁰ Figure 5(a) shows a photo of the stent. Figures 5(b) and 5(c) show the reconstructed 3-D IVPA and IVUS images, respectively, of the stent. Moreover, representative IVPA (SNR 46 dB), IVUS (SNR 41 dB), and IVPA/IVUS fused B-scan cross-sectional images are also shown in Figs. 5(d)–5(f), respectively. From the images, it can be seen that, in this case, the contrast of IVPA is generally better than that of IVUS—the fine structure of the stent struts can be clearly visualized in the IVPA images. However, IVUS can be used to visualize the structure of the vessel. Thus, the fusion of the two (IVPA and IVUS) imaging modalities, as demonstrated here, should be advantageous over either modality alone for guiding and evaluating stenting. When 200-A lines were used to form a B-scan, one cross-sectional image took 0.2 s to complete. Therefore, when a fine pull-back step of $100\mu \mathrm{m}$ (smaller than half of the lateral resolution) was used, $\sim 20\mathrm{s}$ was needed to scan the entire stent of 10 mm in length.

Fig. 5

Imaging of a stent deployed in a healthy rabbit vessel. (a) Photo of the stent. (b) 3-D IVPA and (c) IVUS images of the stent. (d) IVPA, (e) IVUS, and (f) fused IVPA/IVUS B-scan images, respectively, of a representative cross section [labeled by the arrows in (b) and (c)] of the stent.

3.3.

Imaging of Lipid-Mimicking Phantom

To demonstrate the capability of identifying lipid content with the high-speed IVPA system, a lipid-mimicking phantom fabricated by placing a small amount of butter inside a plastic tube was used. Five wavelengths (1185, 1195, 1210, 1225, and 1235 nm) were employed to perform spectroscopic IVPA imaging of the phantom. These wavelengths were selected to reflect the characteristic points of the local absorption peak of lipid, primarily based on the well-measured lipid absorption spectrum reported by several other groups.^9,21 A photo of the lipid-mimicking phantom is shown in the inset of Fig. 6(d). Figures 6(a)–6(c) show the IVPA images of the phantom at three representative excitation wavelengths of 1185, 1210, and 1235 nm, respectively. It can be seen that the signal amplitude is significantly higher at 1210 nm than that from the other two wavelengths, which is consistent with the optical absorption spectrum of lipid. Figure 6(d) shows the spectroscopic IVPA result computed from the above-mentioned five excitation wavelengths, where each data point represents the integrated signal strength corresponding to the region labeled by the dashed box in Fig. 6(b). As a reference, the optical absorption spectrum of lipid is also shown in Fig. 6(d).²¹ The result shows that the IVPA spectrum agrees well with the lipid absorption spectrum.

Fig. 6

Spectroscopic photoacoustic imaging with the high-speed IVPA system. (a)–(c) IVPA images of a lipid-mimicking phantom [inset in (d)] with excitation wavelengths at 1185, 1210, and 1235 nm, respectively. (d) Comparison between the IVPA spectroscopic results with the absorption spectrum of lipid. (e) Lipid map computed based on the correlation between the IVPA and optical absorption spectra of lipid, using five excitation wavelengths.

The lipid distribution can be obtained by computing the correlation between the lipid optical absorption spectrum and the spectroscopic IVPA (sIVPA) data. Specifically, first, the sIVPA data were band-pass filtered without averaging; then, the correlation coefficient $R_{X,Y}$ between the sIVPA and absorption spectra was calculated at each pixel. The correlation coefficient between the two data sets $X$ and $Y$ with expected values $U_X$ and $U_Y$ and standard deviations ${\sigma }_X$ and ${\sigma }_Y$ is defined as

3.4.

Imaging of Peri-Adventitial Adipose Tissue

To demonstrate the capability of identifying lipid in a more natural form, peri-adventitial adipose tissue excised from a porcine aorta segment was imaged. Previous studies have shown that both the $1.2\text{-}\mu \mathrm{m}$ and $1.7\text{-}\mu \mathrm{m}$ spectral bands can be used for IVPA imaging of lipid, and each band may have its own advantages and disadvantages.^16,22,23 Here, as a simple demonstration of the system’s spectroscopic imaging capability, we performed IVPA imaging at both spectral bands. Figures 7(a)–7(c) show the IVPA images of the adipose tissue (placed in a plastic tube) at three representative excitation wavelengths—1185, 1210, and 1235 nm—at the $1.2\text{-}\mu \mathrm{m}$ spectral band, while Figs. 7(d)–7(f) show the IVPA images of the tissue at three representative excitation wavelengths—1680, 1720, and 1750 nm—at the $1.7\text{-}\mu \mathrm{m}$ spectral band. Further, the five-wavelength spectroscopic IVPA data at the location of the arrows are shown in Figs. 7(g) and 7(h); the absorption spectra of lipid at both spectral bands are also depicted for comparison. In this case, it can be seen that the peri-adventitial adipose tissue can be detected via IVPA at either spectral band. All animal experiments were conducted under protocols approved by the Laboratory Animal Centre of Shenzhen Institutes of Advanced Technology and carried out ethically and humanely.

Fig. 7

Spectroscopic IVPA images of the peri-adventitial adipose tissue from a porcine aorta segment [inset in (g)]. (a)–(f) Representative IVPA B-scan images at both spectral bands. (g) and (h) Comparison between the IVPA spectroscopic results and the absorption spectrum of lipid. The arrows indicate the location of the adipose tissue, while the dashed green circles show the contour of the plastic tube.