1. Introduction

Photoacoustic imaging is an emerging technology that involves firing short laser pulses into tissue and recording acoustic signals due to light absorption-induced thermoelastic expansion. Image contrast is principally due to optical absorption. High-resolution photoacoustic imaging has captured considerable attention in the imaging community as its utility is extending to areas of interest to biologists and clinicians. One embodiment, termed Dark-Field Photoacoustic Microscopy (PAM) [1] achieves high resolution by using a mechanically-scanned high frequency, high numerical aperture single-element ultrasonic transducers to receive photoacoustic signals. In 2006, Zhang et al. [2] reported a lateral resolution of 45 μm at a maximum depth of 3 mm by using a 50 MHz focused ultrasonic transducer to demonstrate functional imaging of hemoglobin oxygen saturation. To realize micron-scale acoustic resolution, the transducer frequency would need to be in the hundreds of MHz to GHz range, where penetration depth is limited to less than ~100 μm due to high ultrasonic attenuation in tissue at high frequencies. An alternative technology to achieve high resolution is referred to as Optical-Resolution Photoacoustic Microscopy (OR-PAM). OR-PAM is capable of realizing high optical resolution images because its lateral spatial resolution is determined by the focused optical spot size (limited by the diffraction-limited focal spot size) rather than the width of the ultrasound focal zone. OR-PAM systems are depth-limited to approximately one transport mean-free path (~1 mm in tissue). Because of multiple scattering, diffraction-limited focusing is difficult to achieve past this depth. Beginning with the work of Maslov et al. [3], OR-PAM has been used for structural [4] and functional [4–8] imaging in mice. Oxygen saturation imaging, and imaging of blood velocity has been demonstrated [6] [7], and stunning trans-cranial images of whole-brain murine cortical capillary networks have shown the potential power of the technique for neuro-functional imaging [8]. Other applications have included in vivo imaging of amyloid plaques in a transgenic murine model of Alzheimer’s disease [9], high-resolution functional imaging and chronic monitoring of angiogenesis in a transgenic mouse model [10], ocular microvasculature [11], and others. E. Zhang et al [12] demonstrated a unique Fabry-Perot etalon-based approach for OR-PAM. Xie et al. [13], developed a laser-scanning OR-PAM (LS-OR-PAM) system based on the pioneering work of L.V. Wang’s group. In Xie’s studies, a lateral resolution of 7.8 μm in a 6 mm-diameter circular field-of-view (FOV) was achieved.

For clinical applications, OR-PAM should be developed into a realtime technique. Unfortunately, present volumetric OR-PAM systems have low imaging frame-rates. Since the separation between successive pixels within the image area should be no bigger than the optical spot size, even a relatively small image area <1 mm × 1 mm requires tens of thousands of PA signals (A-scan lines) for each frame. Laser PRR and/or high scanning speeds are thus keys to high imaging frame-rates. Recently, Wang et al. [14], reported a real-time OR-PAM B-scan imaging system with voice-coil scanning speed up to 40Hz over a 1mm range or 20 Hz over 9mm. The LS-OR-PAM system described by Xie et al. showed no frame-rate advantages over previous OR-PAM systems due to limitations of the laser PRR. Presently, the PRR of flashlamp-pumped laser systems can reach 10-100 Hz while that of diode-pumped solid-state Q-switched lasers is at kHz range. For real-time volume-scanning frame-rates which are desirable for clinical applications, these repetition rates are still not high enough. Recently we demonstrated the use of custom-built passively Q-switched microchip (10’s of kHz PRRs) and fiber lasers (100’s of kHz PRRs) operating at the frequency-double wavelength of 532-nm for OR-PAM [15]. Since then, Wang et al. [16], reported a 50-kHz fiber-laser operating at 1064nm for OR-PAM. Unfortunately, optical absorption of hemoglobin at 1064nm is comparatively low. Here we present experimental data to demonstrate the feasibility of performing high frame-rate OR-PAM by employing 532-nm fiber laser source with high repetition rate of up to 600 kHz. Combined with a fast-scanning mirror oscillating at 800 (B-scan) lines per second, we demonstrate an OR-PAM system capable of volumetric imaging at 4 fps. The imaging frame-rate can be higher if the FOV is sacrificed. This imaging speed may be adequate for some clinical applications and is one to two orders of magnitude faster than previous systems. Besides frame-rate advantages, we use a unique light-delivery acoustic probe system. Previous laser-scanning OR-PAM systems used an unfocused transducer positioned a significant distance from the sample to collect photoacoustic signals. Signal attenuation, diffraction and lack of transducer focal gain are significant disadvantages of this technique. In contrast, other OR-PAM systems use clever probe designs [17] to attain high numerical aperture focusing, offering considerable signal-to-noise advantages, but relying on mechanical scanning of the probe, limiting frame-rates. In our approach, we combined the signal-to- noise- ratio-advantages of focused transducers with the speed advantages of laser-scanning by using a unique low-loss light-delivery and ultrasound-detection probe. A laser-spot from our high-repetition-rate fiber laser is scanned across the focal zone of the ultrasound transducer. Due to our near realtime imaging frame-rates, our system may pave the way to more widespread acceptance by biologists and clinicians, and provide new opportunities for studying dynamic processes in clinical and pre-clinical settings.

2. Methods

The system diagram for our unique system is shown in Fig. 1 . Laser pulses are generated by a diode-pumped pulsed Ytterbium-doped fiber laser (GLP-10, IPG Photonics Corporation.) with PRR ranging from 20 kHz - 600 kHz. The 1064-nm fundamental wavelength is frequency-doubled to 532-nm via a built-in compact laser head with output collimator for free-space light delivery. Laser pulse widths are ~1 ns and pulse energies are programmable up to 20 μJ by adjusting the amplifier pump power, referred to by the manufacturer as the set-point. A glass slide was inserted in the beam path to reflect a small amount of light onto a high speed custom photodiode, used for triggering and pulse-to-pulse energy normalization if required. Raster scanning is realized by using a 2D galvanometer scanning mirror system (6230H, Cambridge Technology Inc.) with XY mirrors driven by analog sinusoidal signals from two function generators (AFG3101, Tektronix Inc.). The scanning speed of each mirror can reach hundreds of Hz depending on the scanning angles required. The amplitude of the function generator output determines the maximum scanning angle, which in-turn determines the FOV in the image. An 18-mm-focal-length objective lens was positioned ~3.6 cm below the centers of the scanning mirrors to focus light through our unique light-delivery probe adapted from our previous work [15] [18]. As seen in Fig. 1(a) 10 MHz focused ultrasound transducer (19-mm focus, 6-mm active element, f# = 3.17, CD International Inc) faces a downfacing 10-mm optical prism which reflects the upward photoacoustic signals to the transducer [18]. The acoustic reflectivity is almost 100% for the incident acoustic signal angle within the angular acceptance angles of the transducer [18]. The optical index-matching fluid (Catalog # 19569, Cargille Labs, Cedar, Grove, New Jersey) used for ultrasonic coupling, allowed top-down laser illumination to be directed to the tissue surface without optical refractive path variation. Acoustic attenuation of this fluid was measured to be only slightly higher than that of water [18]. A thin transparent plastic membrane was used to hold the index-fluid in place using O-rings to make a water-tight seal.

 

Fig. 1 Experiment setup for the photoacoustic imaging system with focused transducer (FL: fiber laser; FLD: fiber laser driver; RM: high reflectivity mirror; GS: glass slide; PD: photodiode; FG1, FG2: function generator; DX, DY: galvanometer scanning mirror driver; SM1, SM2: scanning mirror; OL: 18-mm objective lens; P: prism; IF: index-matching fluid; M: membrane; UST: ultrasound transducer; Amp: amplifier). A diode-pumped pulsed Ytterbium fiber laser generates 532 nm output. A glass slide was inserted to reflect some light into high speed photodiode detector. We used 2D fast scanning mirrors and an objective lens to realize scanning and focusing. We used a uniquely designed minimal-loss probe with a 10 MHz ultrasound transducer (US-TX) for photoacoustic signal detection.

Download Full Size | PDF

An eight channel PCI data acquisition card (CS8289, Gage Cobra, Gage Applied Systems, Inc.) with 12-bit dynamic range and up to 125 MSamples/s sample rate was used to acquire the photodiode signal (to trigger data acquisition), the feedback positions of the scanning mirrors (to determine laser spot position on the image plane), and the photoacoustic signals which were detected by an ultrasonic transducer, and amplified by an ultrasound pulser-receiver (5900PR, Olympus NDT Inc). Further data processing and analysis were conducted by using MATLAB programs (Mathworks, Inc) in the data acquisition PC.

3. Results

Figure 2(a) shows a maximum-amplitude-projection (MAP) image of a human hair target positioned 1.5mm below the probe membrane in a clear medium. By setting the 2D scanning galvanometer mirror system with scanning frequencies of 2 Hz and 400 Hz in the Y direction (slow scanning axis) and X direction (fast scanning axis), respectively at optical angles within ± 1.6 degrees, the OR-PAM system realized raster scanning at 4 fps. The image in Fig. 2(a) shows the target diameter of around 100 μm which is roughly the size of a hair. For resolution studies, an MAP image of a ~7.5-μm carbon fiber target positioned ~1.5 mm below the probe membrane was obtained as shown in Fig. 2(b). To analyze the photoacoustic signal resolution, we extracted out a slice of data on the image perpendicular to the carbon fiber. The circles in Fig. 2(c) indicate the photoacoustic signals along the slice direction, and the line curve is its Gaussian fitting which shows the photoacoustic signal full width at half maximum (FWHM) as ~9 μm. Since the measured FWHM shown in Fig. 2(c) are partly due to the width of the carbon fiber itself, we computed the convolution of a 2D Gaussian beam with a carbon fiber (simulated as a 2D rectangular absorption region) [15], which shows the corresponding lateral resolution as ~7μm.

 

Fig. 2 (a) Image of a human hair; (b) Image of a 7.5-μm carbon fiber target positioned 1.5 mm below the probe membrane; (c) Photoacoustic signal (circles) of a slice of data selected perpendicular to the carbon fiber on 2D image with Gaussian fit (line) in clear water, FWHM = 9 μm.

Download Full Size | PDF

For in vivo studies, we used pulse energies of ~0.15 μJ (measured after the scanning mirror system). While 4 fps were achieved in phantom studies, we chose to image a larger field of view (1 mm × 1 mm) for in-vivo studies. We set the 2D scanning galvanometer mirror system with scanning frequency in Y direction as 1 Hz, and in X direction as 400 Hz at optical angles of ± 1.6 degrees, which enabled raster scanning at 2 fps. Frame-rates for this FOV are limited by the fast-scanning mirror system but higher frame-rates are possible for smaller field of view. All experimental animal procedures were conducted in conformity with the laboratory animal protocol approved by the University of Alberta Animal Use and Care Committee. The Swiss Webster mouse was anesthetized using a breathing anesthesia system (E-Z Anesthesia, Euthanex Corp.) during image acquisition. Figure 3 shows snapshots from 2 volumetric images of the microvasculature in a Swiss Webster mouse ear in vivo. The volumetric images were processed by using 3D Vesselness filtering [19] and displayed using Volview software (Kitware, Inc., Volview 3.2). Figure 3(a) clearly depicts a pair of parallel arteries or veins surrounded by many capillaries, while a corresponding volumetric rending is shown in Media 1. Figure 3(b) shows another snapshot with photoacoustic signal FWHM of ~6 μm obtained from resolution studies, while a corresponding volumetric rending is shown in Media 2.

 

Fig. 3 (a) in vivo image of microvasculature in a Swiss Webster mouse ear acquired at 2 fps, which shows an image of a pair of parallel arteries or veins surrounded by capillaries; Media 1 shows the volumetric visualization of the Swiss Webster mouse ear microvasculatures by OR-PAM; (b) Another in vivo image of microvasculature in a Swiss Webster mouse ear; Media 2 shows corresponding volumetric visualization of the Swiss Webster mouse ear microvasculature.

Download Full Size | PDF

4. Discussion

In our studies on photoacoustic imaging systems with fiber laser, we have demonstrated an OR-PAM system with lateral resolution of ~6μm, which is close to the objective lens diffraction-limited focal spot size of ~4.3 μm at 532 nm. A smaller optical focal spot size can be achieved by using an objective lens with higher numerical aperture; however, this would require adjustment of the light delivery system.

Also, our OR-PAM system is capable of C-scan and 3D imaging at 4 fps by using a high PRR 532 nm fiber laser and a high speed 2D galvanometer mirror system. Our in vivo images have a FOV of 1 mm × 1 mm, with an average pixel size of 2.5 µm × 2.5 µm, imaged at a frame-rate of 2 fps. Recently, Hu et al. [20] reported second-generation optical-resolution photoacoustic microscopy which takes 70 min. for 7.8 mm × 10 mm FOV at a pixel size of 2.5 µm × 2.5 µm. Taking into account the different number of pixels in each image, our setup would be on the order of 100 × faster. Ref [16]. uses a 50 kHz fiber laser (25 × faster than previous dye-laser-based systems but12 × slower than our 600 kHz maximum repetition-rate laser), however their imaging frame rates are further limited by mechanical scanning and some signal averaging due to the lower signal strength at 1064nm.

The diode-pumped pulsed Ytterbium fiber laser with PRR ranging from 20 to 600 kHz is of compact size, potentially inexpensive, and should offer considerable advantages due to fiber coupling. It should be noted that the distance traveled by photoacoustic signals during the time between laser pulses at 600 kHz PRF is ~2.5mm, only slightly more than the optical transport mean-free path (which defines the penetration depth of OR-PAM). Hence, repetition-rates much higher than this may be complicated by deep-tissue signals from previous pulses interfering with those due to superficial structures.

The pixel separations along the Y-axis were determined by the ratio of the fast axis scanning speed over the slow axis scanning speed. The maximum scanning speed of around 500 Hz within optical angles of ± 1.6 degrees limits the maximum frame rate to 2.5 fps for an image area of 1 mm × 1 mm at average pixel pitch of around 2.5μm. We chose not to push these scanning-speed limits to prevent heat buildup in the galvanometer system. If active cooling higher speed 2D galvanometer mirror systems are used, with our fiber laser operating at its maximum PRR of 600 kHz, this system is capable of achieving a maximum frame rate of ~15 fps at an average pixel pitch of 2.5 μm for an image size of 500 μm × 500 μm, which provides near-realtime volumetric imaging. Realtime or near-realtime frame-rates will be possible in the near future, which will permit clinical applications.

For our 2 fps in vivo images, each 3D frame takes nearly 96 Mbytes. This is calculated as follows: we used 100 samples per A-scan, and 160,000 laser-shots per 3D image (with laser rep-rate of 320 kHz at 2 fps, with a higher number of laser shots needed for 600 kHz PRR). Note that we also acquire position feedback signals and laser-diode signals for a total of 4 channels. Thus each frame requires (100 × 160,000 × 4) samples/image × 12 bits/sample / (8 bits/byte) = 96 Mbyte/image. Since we have only 128 Mbytes on-board storage capacity we are limited in the number of frames we can acquire before transferring data to the PC RAM. To sustain realtime imaging rates, very high data transfer rates between the data acquisition card and the PC RAM are required and these transfer rates are beyond the capabilities of our current hardware. Future work will aim to store only peak-to-peak values, reduce the number of data channels required, increase the onboard RAM, and use higher data-throughput data acquisition hardware for sustained realtime acquisition and display.

In our in vivo studies, the laser pulse energy was tuned to 0.15 μJ. Assuming that the depth of the laser focus is ~120 μm below the tissue surface, the calculated surface laser fluence is 18 mJ/cm², which is less than the 20 mJ/cm² safety standard set by American National Standards Institute (ANSI) [21], also less than that reported in other OR-PAM studies [20]. The average power delivered to a 1 mm × 1 mm FOV using the maximum laser PRR is ~9 W/cm², significantly higher than the 100 mW/cm² (ANSI-recommended exposure for CW light delivery), but comparable to other in vivo microscopy methods such as fluorescence confocal microscopy. In addition, focal peak power densities (assuming lossless focusing) of 540 MW/cm² are still less than those used in 2-photon microscopy [22]. In our work, light delivery is confined to a localized area and no tissue damage is visible after imaging.

Future work will involve demonstrating imaging of dynamic processes, minimizing the channel count and system complexity by eliminating mirror position feedback, increasing the imaging FOV, developing multi-wavelength high-repetition-rate sources for functional imaging by using nonlinear photonic crystal fibers [23] or Raman shift crystals, exploring nonlinear photoacoustic phenomena for single-wavelength functional imaging, minimizing the system footprint [24] and expanding clinical and biological applications.

5. Conclusion

In our research on OR-PAM imaging system with a diode-pumped pulsed Ytterbium fiber laser, we have demonstrated: a) a photoacoustic imaging system presenting optically-defined lateral resolution of around 6 μm; b) an OR-PAM system capable of volumetric imaging at 4 fps by using a high repetition rate 532nm fiber laser and high speed 2D scanning galvanometer mirror system; c) our system has adequate laser parameters in terms of PRR, pulse energy, and spatial resolution to be used for near real-time optical-resolution photoacoustic microscopy.