1. Introduction

Photoacoustic microscopy (PAM) provides high contrast imaging based on physiological differences in the optical absorption of tissues [1–5]. In brief, optically absorbing molecules convert light into heat causing thermoelastic expansion which, in turn, causes emission of acoustic pressure waves detectable with ultrasound transducers. As the predominant endogenous optically absorbing molecule in tissues, hemoglobin has been imaged using PAM to provide functional imaging of blood oxygen saturation and consumption in vivo. Acoustic-resolution PAM derives its lateral spatial resolution from acoustic focusing, while optical-resolution photoacoustic microscopy (OR-PAM) can achieve micron-scale resolution by using optical focusing to determine lateral resolution. By taking advantage of the high optical absorption of hemoglobin and micron-scale optical spot sizes, OR-PAM has proven invaluable for visualizing superficial capillary networks non-invasively and label-free, in vivo [6,7]. Furthermore, OR-PAM has proven useful for quantifying functional parameters down to capillary sizes [8]. Recent OR-PAM developments include real-time systems [9,10], small handheld systems [11,12], endoscopy systems [13–15], and novel detection schemes [16,17]. Fiber and microchip lasers have recently been introduced as high-repetition-rate sources for realtime OR-PAM [9, 18]; however, the wavelength tunability was limited.

Supercontinuum source for OR-PAM has been reported by using a high-nonlinearity fiber injected with nanosecond pulses from a microchip laser [19]. The supercontinuum was filtered into bands for multiwavelength imaging, however, energy per band was low due to broad distribution of power over such a wide spectral range. In 2011, Koeplinger et al. [20] demonstrated photoacoustic imaging using stimulated Raman scattering (SRS) in optical fiber. They used a 6-m polarization-maintaining fiber and a Nd:YAG microchip laser with 7.5-kHz pulse repetition rates (PRR) for generation of four wavelengths with energy exceeding 80nJ [20]. While limited to a few discrete spectral bands, energy per band was higher than in the super-continuum case. This concept can be extended to using large-mode area photonic crystal fibers for SRS-based generation of multiple wavelength peaks [21].

Our group previously demonstrated SRS multi-wavelength generation from a fiber-laser source and used chromatic aberration advantageously to improve the depth of field of OR-PAM by focusing simultaneous discrete wavelengths at different depths [22]. This was done at the expense of optimal settings for functional imaging. In this paper we aim to mitigate chromatic aberration and optimize outputs at each key wavelength.

In previous SRS multi-wavelength systems, the number of wavelength bands was limited, as was the pulse-energy. Near-infared wavelengths were not demonstrated, nor were in vivo imaging. Our goal was to significantly increase the number of output wavelengths to extend to the near-infrared range, to increase energy per-band, and to ensure minimal SRS peak spectral-widths. This is the first report of in vivo functional imaging using a multi-wavelength fiber laser source. Previous reports may not have had sufficient pulse energy to achieve functional imaging capabilities. Our work reports on an optimization strategy to maximize pulse energies per key wavelength and is not as simple as maximizing input power. The functional imaging capabilities were assessed by determining the dye concentration in tubes and estimating the oxygen saturation levels in the Swiss Webster mouse ear. Previous microchip laser pump sources suffered from pulse-to-pulse stability, timing jitter, low repetition rates, and lack of trigger-ability. These issues are now mitigated with the present fiber laser boasting trigger-ability, tunable repetition rates as high as 600KHz, and pulse-to-pulse stability ~1%. Number of wavelength peaks, narrower spectral linewidths, and tunable repetition-rates are also advantages compared to previous work which may lead to many new possibilities for an all-fiber functional photoacoustic imaging source.

2. Methods

Figure 1 depicts the experimental setup for generation of SRS peaks and for C-scan imaging. The output of a 1-ns pulse width, ytterbium-doped fiber laser (IPG Photonics) capable of PRR from 20 to 600 kHz was coupled into a 6-m polarization-maintaining single-mode fiber (PM-SMF) (HB-450, Fibercore Inc., UK) to generate SRS peaks using a fiber launch system (MBT621D/M, Thorlabs Inc.). SRS peaks are formed from inelastic nonlinear interaction between incoming photons through the fiber and the molecules in the fiber itself [23]. A fiber optic spectrometer (USB4000, Ocean Optics Inc.) measured the SRS peaks and confirmed the filtered wavelengths. The output of the PM-SMF was collimated using a collimator lens (F280APC-A, Thorlabs Inc.) and bandpass filters (FB, Thorlabs Inc.) were used to select the desired wavelength. A systematic optimization study was performed to measure the outputs from varying fiber lengths, input pulse energies, and pulse-repetition rates to maximize SRS wavelength peaks without burning the fiber. Filtered light was scanned across samples using a 2D galvanometer scanning mirror system and focused tightly using an objective lens (Leitz Wetzlar 10X/0.25 160/- EF microscope objective, Germany). Chromatic aberration with this lens was negligible compared to the lens used in our previous work aimed at harnessing chromatic aberration for improved depth of field [22]. An unfocused ultrasound transducer (A3125-SM, 10MHz/0.25”, Olympus Inc.) was positioned beneath the sample to receive the photoacoustic signals. A customized holder was used to contain water between the target and transducer. The position feedback signals from the two mirrors and the amplified RF signals from a pulser-receiver unit (5900PR, Olympus Inc.) were collected using a data acquisition card (CS8289, Gage Applied Systems, Inc.).

 

Fig. 1 A) Experimental setup of multi-wavelength OR-PAM. FLD: Fiber laser diode, OL: Objective lens, PM-SMF: Polarization maintaining single mode fiber, CL: Collimator lens, UST: Ultrasound transducer B) Photograph of the generated multi-wavelength spectrum in a PM-SMF.

Download Full Size | PDF

For maximum energy at the desired imaging wavelengths (532, 545, 558, and 590nm), a laser PRR of 40 kHz and a 3-m fiber were used. All the experimental procedures were carried out in conformity with the laboratory animal protocol approved by the University of Alberta Animal Use and Care Committee. Authors are also trained and certified in order to use mice in the research work. During the imaging session the animal was anaesthetized using a breathing anaesthesia system (E-Z Anesthesia, Euthanex Corp.).

3. Result and discussion

Table 1 shows the measured power at different PRR and input power level for 3-m and 6-m fiber length. The coupling efficiency for all of these experiments was ~70%. The pulse energy of each output wavelength can be maximized by choosing a specific input power into the fiber which means, for example, in order to generate optimized pulse energies for 3 different wavelengths, 3 different input powers are necessary. While a higher input power is desirable to generate higher output power levels, input power levels will be limited due to the fiber damage threshold. In this experiment, the input power was varied between 30 mW-150 mW depending on the fiber length, the PRR, the coupling efficiency and the desired wavelength. The fiber damage was first observed at 160-kHz PRR and 150 mW average input power. The back-reflected light from the filter may also cause damage at the end of the fiber; this can be solved by introducing a small angle to the filter. Longer interaction lengths induce a wider range of wavelengths; however, for the proposed in vivo application the 3-m fiber length at 40-kHz PRR produced maximized pulse energies for four wavelengths under the fiber damage threshold as shown in Table 1.

Table 1. Measured power of SRS peaks generated in varying fiber lengths and at different PRR

View Table | View all tables in this article

Using a 160-kHz PRR and 15-m PM-SMF, we were able to obtain SRS peaks at 489, 499, 511, 522, 532, 545, 558, 572, 587, 603, 621, 639, 656, 675, 695, 710, 740, 765 and 788nm (Fig. 2(A)). The spectrometer verified that the bandpass filter accurately selected for the desired imaging wavelength (Fig. 2(B)). Far-red and infra-red spectral peaks have estimated pulse energies as high as ~100 ± 50nJ, higher than pulse energies used in previous OR-PAM experiments [11, 13]. Four-wave mixing may be one source of SRS peak broadening as two or more wavelength may interact in order to produce an output at various sum or difference frequencies [23].

 

Fig. 2 A) SRS peaks for 160kHz PRR and a 15m PM-SMF. B) Unfiltered (dashed) and filtered (solid) SRS peaks for 160kHz PRR and a 6m PM-SMF (the input power varied between 55 and 100mW).

Download Full Size | PDF

Broad spectral peaks could result in ambiguous oxygen saturation estimates if spectral bins are not sufficiently narrow in multi-wavelength imaging, however, this could be dealt with using narrower-line filters, at the expense of energy. For functional OR-PAM, only wavelengths of 532, 545, 558, and 580nm were necessary to be optimized therefore to minimize four-wave mixing and for maximum energy, a 40-kHz PRR and a 3m PM-SMF was used to generate 532, 545, and 580 nm with pulse energy between 300 and 500 nJ.

Initially, carbon fiber networks and double dye experiments were conducted to assess imaging potential at each SRS peak (Fig. 3). The measured power for each filtered wavelength was set to 5mW. Signal-to-noise ratios (SNR) for carbon fibers networks were ~41, 39, 43, and 42 dB and lateral resolution was 7, 7.5, 7.7, and 8µm at 532, 545, 560, and 590nm, respectively. The lateral resolution degrades due to spherical aberration; however this is insignificant and negligible [20]. Figure 3(A) shows an example of C-scan images of carbon fiber network using 560nm wavelength at 160 kHz PRR. Double dye experiments were conducted using red dye (Fiesta Red) and blue dye (Lake Placid Blue). In this experiment the photoacoustic signals were normalized by laser fluence at each wavelength and the photoacoustic spectrum was compared to the absorption spectrum of the dyes. Using several regions within the tube from the C-scan image as shown in Fig. 3(B), spectral demixing was able to accurately estimate the concentration of blue and red dye in each tube. Figure 3(B) is an example of the dye-filled tubes at 3 different wavelengths at 160 kHz PRR for 100% Red and 0% Blue dyes. The demixing results of 5 different concentration of red dye are shown in Table 2. Figure 4 shows the measured absorption of Fig. 3(B) images at 3 different regions. Different regions were chosen in order to improve the accuracy of the measurement. Figure 5 shows mock oxygen saturation estimation using mixtures of red and blue dye. The results show that the C-scan OR-PAM images are in good agreement with the spectrometer results.

 

Fig. 3 C-scan images of (A) carbon fiber networks and (B) the dye-filled tubes at 3 different wavelengths for 100% red and 0% blue dyes.

Download Full Size | PDF

Table 2. Spectral demixing of PA signals of tubes containing various concentrations dyes

View Table | View all tables in this article

 

Fig. 4 The average signal for selected regions within the tube shown in Fig. 3(B). This data has been used to determine the absorption spectrum.

Download Full Size | PDF

 

Fig. 5 Mock oxygen saturation estimation using mixtures of red and blue dye.

Download Full Size | PDF

In vivo imaging of a capillary network at 545 nm and 558 nm is shown in Fig. 6. Slight shifts between the two images were resolved with cross-correlation so that spectral demixing algorithms could be applied. Using a least squares demixing algorithm, arteries (orange in Fig. 6(C)) and veins (blue-violet in Fig. 6(C)) can be separated. The resolution of the system is measured ~7µm with ~2µm average scanning step size. The present all-fiber source is suitable for imaging at multiple wavelengths and, with minor modification to the setup, is capable of imaging at near-realtime frame rates. Our experimental work demonstrates that we can image with up to 160 kHz PRR for phantom applications and 40 kHz for in vivo applications and with pulse energies greater than 120 nJ – enough for typical photoacoustic imaging experiments. This system can easily be integrated with current OR-PAM table top systems, our endoscopy systems [13], and our handheld systems [11] providing C-scan frames for each SRS peak available. This will permit quantitative estimation of blood oxygenation and saturation in capillaries, as well as, imaging other absorbing reporter molecules. Further improvements can be made by optimizing the SNR, increasing frame rates, utilizing additional SRS peaks, implementing fast electronic wavelength switching or utilizing new nonlinear fibers to generate high-order SRS peaks.

 

Fig. 6 Multi-wavelength in vivo imaging using 545nm (A) and 558nm (B) pulses. Oxygen saturation estimations are shown in (C) for the area within the dashed rectangle.

Download Full Size | PDF

4. Conclusion

This paper presents the potential of a multi-wavelength optical-resolution photoacoustic microscopy system for in vivo imaging using stimulated Raman shifting. The output of a ytterbium-doped fiber laser with 1ns pulse widths and laser pulse repetition rates up to 160 kHz was coupled into varying lengths of single mode polarization maintaining fiber generating wavelengths peaks out to nearly 800 nm with pulse energies of hundreds of nJ. We anticipate that this cost effective multi-wavelength source will open up a whole range of new possibilities for functional imaging applications.