1.

Introduction

Optical resolution photoacoustic microscopy (OR-PAM), which provides high sensitivity by detecting optical absorption contrasts, has been widely used for imaging biological systems, from single cells to organisms.^1–5 A conventional OR-PAM setup typically excites molecules using a focused Gaussian laser beam, which can have a micron- to submicron-sized focal spot but suffers from limited focal depth. In order to acquire millimeter-thick volumetric images with consistent lateral resolution in the depth direction, we need three-dimensional (3-D) raster scanning to compensate for the short focal depth, which reduces the imaging speed. In comparison, a Bessel beam is essentially diffraction free, i.e., the focal depth can be extended greatly without compromising the focal spot size, which avoids the tradeoff between imaging resolution and focal depth.^6–8 However, Bessel beams have many strong side lobes, which generate ghost signals, limiting their application in optical imaging. Similarly, in photoacoustic (PA) imaging, ultrasound transducers usually have large focal diameters (tens to hundreds of microns), and thus they receive most of the ghost signals from the side lobes of the Bessel beam,⁹ which inhibits the direct application of Bessel beams in PA imaging.

Until recently, nonlinear optical techniques, e.g., two-photon excitation in fluorescence microscopy, have been successfully used to suppress the side-lobe effect of Bessel beams and then achieve rapid volumetric imaging with a long focal depth.^10–14 Nonlinear effects exist in PA imaging as well.^15–21 One of them is the Grueneisen relaxation (GR) effect,²⁰ a thermally induced nonlinear effect that has been explored to improve both the axial and lateral resolution for PAM.^20,21 Here, we utilize the GR effect in Bessel-beam PAM to suppress the side-lobe effect and achieve long-focal-depth volumetric imaging.

2.

Methods

GR-PAM employs dual-pulse excitations to obtain nonlinear PA signals.²⁰ The first laser pulse generates a PA signal and, at the same time, increases the local temperature. The second laser pulse is almost identical to the first laser pulse and excites the same absorber within the thermal confinement time. Owing to the temperature rise induced by the first pulse, the local Grueneisen parameter is increased, and the second PA signal is stronger than the first one. The nonlinear PA signal is the amplitude difference between these two signals and is proportional to the square of the optical fluence, thereby achieving nonlinear effects. Alternatively, we can also realize GR-PAM using a two-step measurement scheme. In the first step, one probe laser pulse generates a PA signal; in the second step, a heating laser is applied before firing the second probe laser pulse. The resulting nonlinear PA signal is the difference of the two signals from the two-probe laser pulse excitations. In this way, the heating laser could be either a continuous-wave (CW) or a pulsed laser, as long as the turn-on time and time delay between the heating and probing fulfill the thermal confinement requirement. In the two-step measurement scheme, the first probe laser pulse generates an initial pressure rise:

In the second step, after the heating laser is applied, the Grueneisen parameter increases owing to the local temperature rise:

The corresponding PA signal is

Subtracting Eq. (2) from Eq. (5), we obtain a differential PA signal ($\mathrm{\Delta }\mathrm{PA}$):

When the heating laser and probe laser have the same optical beam profile, then their optical fluences are linearly proportional to each other: $F_2(x,y)=g{F}_1(x,y)$, where $g$ is a constant depending on the power ratio of these two lasers. This simplifies Eq. (6) to

Let the optical fluence follow a Bessel profile:

3.

Experimental Setup

Figure 1(a) is a schematic of a Bessel-beam PAM system. The probe laser is a 532-nm diode-pumped solid-state pulsed laser (Innoslab BX2II-E, Edgewave GmbH), which generates 10-ns laser pulses at repetition rates in the range of hundreds of hertz. The heating laser beam, from either a CW laser or a pulsed laser, is combined with the probe laser beam and then spatially filtered by a $50\text{-}\mu \mathrm{m}$ tungsten pinhole. The filtered laser beam has a 3-mm diameter and a near-Gaussian transverse profile. Finally, an axicon lens (Thorlabs AX255-A, 25 mm, 5.0 deg) transforms the laser beam into a Bessel beam. The sample holder, a petri dish filled with water for ultrasound coupling, is mounted on a 3-D scanning stage (PLS-85, PI miCos GmbH) and it is about 7 mm away from the tip of the axicon lens. When the sample is irradiated by the Bessel beam, PA waves are generated through light absorption and then detected by a custom-made focused ultrasonic transducer (48-MHz central frequency, 80% detection bandwidth, 11.6-mm focal length, and 0.25 NA). The focal zone of this transducer is about 2 mm, encompassing the long focal depth of the Bessel beam. The PA signal from the transducer is further amplified by 40 dB (ZKL-1R5+, Minicircuits) and digitized at 500 MHz using a data acquisition card (ATS9350, AlazarTech). A LabVIEW programmed FPGA card (PCI-7830R, National Instruments) synchronizes the laser triggers, scanner motion, and data acquisition for imaging. A microcontroller (Arduino Due) programs the CW laser TTL trigger and the interpulse delay.

Fig. 1

Schematic of Bessel-beam photoacoustic microscopy (PAM). (a) Experimental system setup. L1, L2: lens. (b) Continuous wave (CW)-heating scheme. A CW laser is used as the heating laser. The heating duration ($\mathrm{\Delta }t$) of the CW laser must be less than the thermal relaxation time. (c) Pulsed-heating scheme. Both the heating and probe laser beams are pulsed, and the interpulse delay ranges from 100 ns to $10\mu \mathrm{s}$.

The axicon lens converts a Gaussian beam to a good approximation of a true Bessel beam.^10,22–24 According to Ref. 25, the radius of the central lobe ($\rho$) is defined as the first zero of the Bessel function: $\rho =2.40\lambda /(2\pi \beta )$, where $\lambda$ is the wavelength, $\beta ={\sin }^{-1}(n\sin \alpha )-\alpha$, $\alpha$ is the axicon angle, and $n$ is the refractive index of the axicon lens material. The depth of field is $L=0.8w_0/\beta$, where $w_0$ is the beam diameter of the incident Gaussian beam. In our system, the wavelengths of both the CW laser and pulsed lasers are 532 nm, and the beam diameter of the incident Gaussian beam is 3 mm. The beam profile, measured by a beam profiler (BeamGage, Ophir), as shown in Fig. 2(a), is the transverse intensity profile of the Bessel beam taken at a relative axial position $z=2\mathrm{mm}$ (The axial position at $z=0\mathrm{mm}$ is about 7 mm away from the tip of the axicon lens). The central lobe has a full-width at half-maximum (FWHM) of $6.5\mu \mathrm{m}$, while the side lobes spread widely beyond a $100\text{-}\mu \mathrm{m}$-diameter circular area. The central lobe diameter varies depending on the axial position relative to the axicon lens. The axial distributions of the central lobe diameter (FWHM) and peak intensity of the central lobes are shown in Fig. 2(b). In the region $z>0\mathrm{mm}$, the transverse beam profile can be described as a quasi-Bessel beam and the central lobe diameter is around 6 to $7\mu \mathrm{m}$. In the region $z<0$, i.e., closer to the axicon lens, the central lobe diameter diverges quickly; hence, this region should be avoided for imaging due to its poor resolution. In practice, the peak intensity is also an important factor, since a low peak intensity not only attenuates the PA signal but also attenuates the heating effect. Therefore, for the Bessel beam shown in Fig. 2, only a 2-mm focal depth ($0<z<2\mathrm{mm}$) is used for nonlinear Bessel-beam PAM. This focal depth is much larger than that of conventional Gaussian beams, e.g., the focal depth of a Gaussian beam with the same focal spot size $6\mu \mathrm{m}$ at 532 nm is only about $106\mu \mathrm{m}$.

Fig. 2

Bessel beam profile. (a) Transverse intensity profile taken at $z=2\mathrm{mm}$ (at $z=0\mathrm{mm}$, it is about 7 mm away from the tip of the axicon lens). The central lobe diameter ($6.5\mu \mathrm{m}$) is the full-width at half-maximum (FWHM) of the central lobe. White peaks represent intensity profiles along white dashed lines, respectively. (b) Axial distributions of the central lobe diameter (red solid line, $Y$-axis is on the left side) and peak intensity (green dashed line, $Y$-axis is on the right side).

To generate nonlinear PA signals, we employ two schemes. In the first scheme, as illustrated in Fig. 1(b), a CW laser is turned on for a short period of time $\mathrm{\Delta }t$ to heat the sample and then a pulsed probe laser is triggered at the end of the heating. The heating time is within the thermal relaxation time. The pulsed probe laser generates PA signals, whereas the CW laser is only for heating. At each scanning point, we record two PA signals: the first PA signal (${\mathrm{PA}}_1$) is excited by a single laser pulse without CW laser heating, whereas the second PA signal (${\mathrm{PA}}_2$) is excited by another laser pulse right after the CW laser heating. Owing to the GR effect, the second PA signal is stronger than the first one, and their amplitude difference is the nonlinear PA signal ($\mathrm{\Delta }\mathrm{PA}$). Due to its low power ($<100\mathrm{mW}$ in our system) and the thermal confinement of the target, the CW laser is good for heating highly absorptive samples (e.g., carbon fiber, hair, etc.). In the second scheme, both the heating laser and the probe laser are pulsed, as shown in Fig. 1(c). The interpulse delay is a constant ranging from 100 ns to several $\mu \mathrm{s}$, which is much shorter than the thermal relaxation time. In this scheme, similar to the first scheme, two probe PA signals are recorded with and without the pulsed heating. Their computed amplitude difference is the nonlinear PA signal ($\mathrm{\Delta }\mathrm{PA}$). The pulsed heating laser, possessing higher pulse energy, provides better heating effect for low absorptive samples but is more expensive than the CW heating laser. In our experiments, we demonstrated the feasibility of both schemes. We employed the first scheme to measure a carbon fiber to quantify the effect of the nonlinear Bessel-beam PAM and then used the second scheme to image red blood cell (RBC) samples.

4.

Results and Discussion

In order to quantify the lateral resolution and effective axial focal depth, we imaged a carbon fiber ($6\text{-}\mu \mathrm{m}$ diameter) using the nonlinear Bessel-beam PAM system. The thermal relaxation time of a carbon fiber immersed in water with a $6.5\text{-}\mu \mathrm{m}$ diameter optical heating spot has been estimated to be 1 ms,²⁶ so we heated the carbon fiber using a 532-nm CW laser (40-mW power, $300\text{-}\mu \mathrm{s}$ duration, and equivalent heating energy $12\mu \mathrm{J}$) and excited PA signals using a pulsed laser (10-ns pulse width and 1200-nJ pulse energy). Because the energy distributes almost equally among the central and each side lobe of the Bessel beam, only a small portion of the energy can be absorbed by the thin carbon fiber. The carbon fiber was placed horizontally in the petri dish and immersed in water for coupling and a cross-sectional PA image was obtained by one-dimensional mechanical scanning across the carbon fiber. A typical measured PA amplitude profile across the carbon fiber is shown in Fig. 3(a), where the red line ${\mathrm{CW}}_{\text{on}}$ is the enhanced PA image with the CW laser heating, while the blue line ${\mathrm{CW}}_{\text{off}}$ is the conventional PA image acquired without laser heating. Both the ${\mathrm{CW}}_{\text{on}}$ and ${\mathrm{CW}}_{\text{off}}$ images have significant backgrounds, which are produced by the side lobes of the Bessel beam. By taking the amplitude difference between ${\mathrm{CW}}_{\text{on}}$ and ${\mathrm{CW}}_{\text{off}}$, we computed the differential PA image $\mathrm{\Delta }\mathrm{PA}$, which clearly reduces the background and suppresses the side-lobe effect. In Fig. 3(a), the FWHM of $\mathrm{\Delta }\mathrm{PA}$ is about $6.1\mu \mathrm{m}$, which is close to the value of $6.5\mu \mathrm{m}$ estimated from the convolution of the carbon fiber cross-section profile and the profile squared of the Bessel beam central lobe. Figure 3(b) shows the lateral resolution at different axial positions. In the region of $0.6\mathrm{mm}<z<2.2\mathrm{mm}$, the FWHM value is around $6.2\mu \mathrm{m}$ while it increases quickly for smaller $z$ ($<0.6\mathrm{mm}$, closer to the axicon lens) because the central lobe diameter becomes large, as shown in the Bessel light beam profile in Fig. 2(b). For $z>2.2\mathrm{mm}$ (further from the axicon lens), the FWHM also deteriorates, since the heating light intensity of the central lobe is attenuated significantly in this region. In summary, the effective nonlinear Bessel-beam imaging region in Fig. 3(b) is about 1.6 mm in which the system maintains a lateral resolution finer than $\sim 7\mu \mathrm{m}$.

Fig. 3

Cross-sectional images of a $6\text{-}\mu \mathrm{m}$-diameter carbon fiber. (a) Red line: one-dimensional (1-D) photoacoustic (PA) image acquired when the CW laser is on. Blue line: 1-D PA image acquired when the CW laser is off. Green line: the differential PA image $\mathrm{\Delta }\mathrm{PA}$. Black line: Gaussian fit of $\mathrm{\Delta }\mathrm{PA}$. The FWHM of the $\mathrm{\Delta }\mathrm{PA}$ image is $6.1\mu \mathrm{m}$. To improve SNR, 20 measurements were averaged. (b) Axial distribution of the FWHM of the cross-sectional carbon fiber images. The axial coordinate $z$ is relative to the beginning measurement position ($z=0$ is about 7 mm away from the tip of the axicon lens).

Besides the carbon fiber, we also imaged RBC samples, which are relatively low absorptive. Bovine RBCs (Quad Five, Ryegate, Montana) were first fixed with methanol and then mixed with agar gel (1% weight ratio of agarose and water) at 40°C. The volume ratio between the RBC and agar gel was $1:500$. A single-layer RBC sample was prepared by simply smearing the RBC-agar gel onto a cover glass. A thick-layer (0.5 to 1.0 mm) RBC sample was formed by dropping the gel into a hole in the bottom of a petri dish sealed with a cover glass. Finally, the RBC-agar gel was cooled to room temperature (25°C) to form solid samples. The single-layer RBC sample was used to quantify the side-lobe suppression effect. The thick-layer sample was used to demonstrate the volumetric imaging capability of nonlinear Bessel-beam PAM. To provide sufficient power to heat the highly diluted RBC sample, we exchanged the CW heating laser for a high-power pump laser (Innoslab IS8II-E, Edgewave GmbH, $20\text{-}\mu \mathrm{J}$ pulse energy) and triggered a probe laser (Innoslab BX2II-E, Edgewave GmbH, $14\text{-}\mu \mathrm{J}$ pulse energy) to generate PA signals about 300 ns after triggering the heating pulsed laser.

We performed two-dimensional (2-D) raster scanning to image the single-layer RBC sample (actually four separated RBCs) in the agar medium. The regular probe-only PA image ${\mathrm{PA}}_1$, thermally enhanced PA image ${\mathrm{PA}}_2$, and the differential PA image $\mathrm{\Delta }\mathrm{PA}$ (${\mathrm{PA}}_2-{\mathrm{PA}}_1$) are displayed in Figs. 4(a)–4(c). We took the maximum amplitude projection (MAP) along the depth (or axial) direction to produce these images. In Fig. 4(a), the regular PA image ${\mathrm{PA}}_1$ with probe-only detection rendered severely ring-shaped artifacts, which reveal the side-lobe problem in regular (linear) Bessel-beam PAM. Especially for RBCs 2 and 3 as labeled in Fig. 4(a), the side lobe signals significantly interfere with other main lobe signals. After irradiation by the heating laser, the PA signal ${\mathrm{PA}}_2$ and thereby the signal-to-background ratio were enhanced owing to the GR effect, as indicated in Fig. 4(b), but the side lobe signals still exist. The maximum amplitude of ${\mathrm{PA}}_2$ was 43% greater than that of ${\mathrm{PA}}_1$ at the same position, which showed that there was a temperature rise of 13°C in the RBCs due to Grueneisen effect. According to this temperature rise, the effective optical fluence around the central lobe is estimated to be $195\mathrm{mJ}/{\mathrm{cm}}^2$.²⁷ In Fig. 4(c), the differential image $\mathrm{\Delta }\mathrm{PA}$ shows significantly reduced side lobe signals and reveals each RBC clearly with a higher signal-to-background ratio.

Fig. 4

Single-layer red blood cell (RBC) imaging. (a) Two-dimensional (2-D) PA image (${\mathrm{PA}}_1$) of four RBCs in agar medium. No heating laser was applied. (b) 2-D thermally enhanced PA image (${\mathrm{PA}}_2$) of the same RBCs. The probe laser, triggered at 300 ns after the release of pulsed heating laser shot, detected the PA signal. (c) 2-D differential PA image of the same RBCs. $\mathrm{\Delta }\mathrm{PA}={\mathrm{PA}}_2-{\mathrm{PA}}_1$. All images are normalized to the maximum amplitude of all pixels. There was no averaging for each image. (S/B: signal-to-background ratio).

Further, the nonlinear Bessel-beam PAM was used to image two thick RBC samples in agar medium. The first sample was $400\text{-}\mu \mathrm{m}$ thick. Figure 5 shows the MAP images.²⁸ The maximal amplitude of each A-line at every pixel from 2-D raster scanning constitutes the MAP image, which can be interpreted as stacked images along the $z$-axis. A regular probe-only Bessel-beam PA MAP image of the $400\text{-}\mu \mathrm{m}$ thick RBC sample is shown in Fig. 5(a), which is blurred due to the accumulation of side lobe signals. In addition, in this sample, some RBCs were separated from other cells, while others were aggregated. The side lobe signals from the big aggregated RBC clusters dominated over the main lobe signals of single RBCs or small aggregated RBC clusters, resulting in more artifacts than in the single-layer RBC sample. As a consequence, it is difficult to identify RBCs in Fig. 5(a). On the other hand, as shown in Fig. 5(b), the MAP of the nonlinear Bessel-beam image dramatically improved the image quality. The single or clustered RBCs in Fig. 5(b) can be clearly distinguished from the background. For example, comparing the lower-left corners of Figs. 5(a) and 5(b), we see a high background covering the actual main lobe signals in Fig. 5(a), while the background in Fig. 5(b) is dramatically reduced.

Fig. 5

2-D maximum amplitude projected (MAP) volumetric image of a $400\text{-}\mu \mathrm{m}$-thick RBC sample in agar medium. (a) Linear Bessel-beam PA image (${\mathrm{PA}}_1$) acquired without heating laser. (b) Nonlinear Bessel-beam PA image based on the differential PA image ($\mathrm{\Delta }\mathrm{PA}$).

The second sample was 1-mm thick. Figure 6 shows the linear (${\mathrm{PA}}_1$) and nonlinear ($\mathrm{\Delta }\mathrm{PA}$) Bessel-beam PA images obtained at different depths from $z=0$ to $960\mu \mathrm{m}$. Four sets of 2-D images were extracted from volumetric images of the sample. Comparing the linear (${\mathrm{PA}}_1$) and nonlinear ($\mathrm{\Delta }\mathrm{PA}$) images, it is clear that the nonlinear approach was able to highlight central lobe signals and suppress the side-lobe background over a 1-mm depth. For example, the image in Fig. 6(d1) shows high background arising from side lobes, almost “burying” the actual signal from the main lobe, whereas the nonlinear image in Fig. 6(d2) clearly shows a clear main lobe signal. Deeper images were noisy and have weak nonlinear effects due to the limited effective focal depth of the Bessel beam in the current experimental setup. Compared with two-photon Bessel-beam microscopy,^10,25 nonlinear Bessel-beam PAM not only provides a similar long imaging depth but also provides axial resolution along the $z$-axis, which is determined by the bandwidth of the ultrasonic signal. In this system, the axial resolution is estimated as $30\mu \mathrm{m}$.

Fig. 6

2-D image slices from a volumetric image of a 1-mm-thick RBC sample in agar medium. Upper row: regular Bessel-beam PA images (${\mathrm{PA}}_1$). Lower row: nonlinear Bessel-beam PA images ($\mathrm{\Delta }\mathrm{PA}$). Images at relative axial depths (a1 and a2) $z=0\mu \mathrm{m}$, (b1 and b2) $z=261\mu \mathrm{m}$, (c1 and c2) $z=741\mu \mathrm{m}$, and (d1 and d2) $z=960\mu \mathrm{m}$. The maximum signal amplitude was normalized to be 1 in each image.

5.

Conclusions

In summary, we have demonstrated the extended imaging depth and optical lateral resolution of nonlinear Bessel-beam PAM in imaging carbon fibers and RBC samples. The imaging depth is determined by the effective depth of field of the Bessel beam, and the resolution is mainly determined by the central lobe diameter of the Bessel beam. In linear Bessel-beam PAM, the image quality is poor due to the side-lobe effect, whereas in nonlinear Bessel-beam PAM, the side lobe signals are effectively suppressed, thereby highlighting the main lobe signal and offering optical lateral resolution. The experimental system can be converted to reflective mode for in vivo imaging by following the design for previous OR-PAM systems.¹ Although the optical fluence used in this work ($195\mathrm{mJ}/{\mathrm{cm}}^2$) is higher than the ANSI standard ($20\mathrm{mJ}/{\mathrm{cm}}^2$) but lower than the damage threshold, improving the ultrasound transducer’s sensitivity to increase the signal-to-noise ratio can reduce the optical fluence. Nonlinear Bessel-beam PAM effectively extends the focal depth of PA imaging beyond that of conventional Gaussian-beam PAM and therefore reduces the time to image thick samples.