Single all-fiber-based nanosecond-pulsed supercontinuum source for multispectral photoacoustic microscopy and optical coherence tomography

Abstract

We report the usefulness of a single all-fiber-based supercontinuum (SC) source for combined photoacoustic microscopy (PAM) and optical coherence tomography (OCT). The SC light is generated by a tapered photonic crystal fiber pumped by a nanosecond pulsed master oscillator power amplifier at 1064 nm. The spectrum is split into a shorter wavelength band (500–800 nm) for single/multispectral PAM and a longer wavelength band (800–900 nm) band for OCT. In vivo mouse ear imaging was achieved with an integrated dual-modality system. We further demonstrated its potential for spectroscopic photoacoustic imaging by doing multispectral measurements on retinal pigment epithelium and blood samples with 15-nm line-width.

Over the past decades, the advancement in multi-modality optical imaging has benefited biomedical research and clinical diagnosis by providing complementary contrasts. The contrasts originating from two major optical properties of biological tissue, absorption and scattering, can be revealed by using photoacoustic microscopy (PAM) and optical coherence tomography (OCT), respectively. PAM is an emerging imaging technique that allows mapping of tissue optical absorption properties by detecting ultrasonic wave generated from thermoelastic expansion following pulsed laser beam excitation. OCT, on the other hand, can reconstruct the three-dimensional (3D) distribution of scattering properties within tissue by low-coherence interferometry. Integrated PAM and OCT systems have already proven their usefulness in retinal, skin, and ovarian tissue imaging, and microcirculation studies [1–4].

In most of the previous dual-modality designs, two different light sources were employed, since no single commercially available light source is suitable for both PAM and OCT. PAM typically uses narrowband pulsed laser excitation and prefers tunability of dye lasers or optical parametric oscillators (OPO) to perform multispectral imaging for functional information [3,5]. OCT uses broadband light sources, among which super-luminescent diode (SLD) is dominant. Involvement of two light sources complicates the system, hampers portability, and increases cost. Therefore, developing a compact and versatile light source supporting both PAM and OCT would benefit both research applications and potential commercial product development.

An initial trial on implementing PAM and OCT with a single light source was reported by Zhang et al., who built a dye laser pumped by a frequency-doubled Q-switched Nd:YAG laser [6]. However, the output light only had a 20-nm bandwidth centered at 580 nm, which limited wavelength selection range for multispectral PAM, as well as the axial resolution in OCT.

A shared light source for PAM and OCT can potentially be achieved by a supercontinuum (SC) source, which relies on spectral broadening through nonlinear processes, and has proved its usefulness in both PAM and OCT applications. Nanosecond-pulsed SC sources are used in multispectral PAM, and wavelength selection is achieved by bandpass filtering. Compared to dye lasers and OPOs, SC sources are more stable and less bulky; they also exhibit a fast-tuning potential and offer a broader wavelength range [7]. SC sources also provide a much more uniform output spectrum and, thus, wider wavelength selection than stimulated Raman scattering lasers, whose major energy is distributed on a series of fixed individual peaks [8]. Besides PAM applications, SC sources also demonstrate gratifying performance in OCT systems [9]. Though stronger relative intensity noise of SC sources leads to lower OCT detection sensitivity, compared with a system built with SLD [10], the much broader SC spectrum provides extra wavelength selection flexibility and access to higher axial resolution [11].

To work as a shared light source for both PAM and OCT, a SC source needs to be pulsed at kilohertz range for PAM, though OCT, on its own, prefers a megahertz pulse repetition rate. Lee et al. developed such a SC source and, subsequently, tested it on a PAM/OCT system [12]. However, the SC spectrum was not broad enough, and the spectral power in the visible range was not sufficient. As a result, both PAM and OCT subsystems exploit NIR light. Since PAM relies on strong and distinct optical absorption spectra of oxy- and deoxy-hemoglobin in the visible range to extract oxygen saturation, using NIR light limits the capability of PAM in hemodynamics investigation [3].

In this Letter, we demonstrate the feasibility of dual-modality PAM/OCT system built with a single homebuilt all-fiber-based SC source. PAM uses a shorter wavelength band (SWB, 500–800 nm), while the OCT uses a longer wavelength band (LWB, 800–900 nm). We characterized the integrated system and achieved in vivo imaging on mouse ears. In addition, we performed multispectral PAM measurement on ex vivo retinal pigment epithelium (RPE) and blood samples, which paves the way for functional image analysis, such as quantification of RPE melanin concentration and vascular oxygen saturation.

Figure 1 shows the schematic and output spectrum of the SC source. As Fig. 1(a) illustrates, the SC is pumped at 1064 nm using a master-oscillator-power-amplifier configuration with a homebuilt ytterbium-doped fiber amplifier. The laser is triggered at 25 kHz and the pulse width is 2 ns. Spectrum broadening is induced by a photonic crystal fiber from NKT Photonics A/S, whose tapered design (starting from a 10.7 μm mode field diameter down to 5 μm) allows enhanced power handling and, therefore, stronger SC generation. Based on this design, the entire SC spectrum covers a broad 500–2300 nm wavelength range, and a good uniformity for the spectral emission is obtained between 600 and 1600 nm, as shown in detail by Bondu et al. [13]. In this Letter, however, as a proof of concept, we only selected SWB for PAM, targeting blood and melanin absorption, and LWB for OCT, which is among the most frequently used bands in both research and clinical applications. Figure 1(b) shows the output power spectrum between 500 and 900 nm of the light source. Figure 1(c) depicts the stability of the source power characterized by a relative standard deviation measured over 100 power spectra sequentially acquired by a commercial spectrometer (USB2000+, Ocean Optics) with a 190-μs integration time.

Fig. 1.

(a) Schematic of the SC source. YDFA, ytterbium doped fiber amplifier; PCF, photonic crystal fiber; OL, outlet. (b) Output spectrum used by the dual-modality system. (c) Relative standard deviation (RSD) across the spectrum.

With the source described above, we built a combined PAM and OCT system, as shown in Fig. 2(a). The SC was split into SWB and LWB for PAM and OCT, respectively. The LWB was delivered to a fiber-based spectral-domain OCT, where it was split into a sample arm and a reference arm by a 50 by 50 fiber coupler (FUSED-22-850, OZ optics). The backscattered light from the sample arm interfered with the mirror-reflected light from the reference arm, and the interference pattern was collected by a homemade spectrometer consisting of a diffraction grating (1200 lines/mm, Wasatch Photonics), a focusing lens (150 mm, Thorlabs), and a line camera (spL2048-140 km, Basler). The SWB was delivered to a homebuilt PAM, where the beam was expanded by a 1:2 Keplerian telescope to improve the system’s lateral resolution and to superimpose the foci of both subsystems in the case of chromatic focus shift between the two bands. The LWB and SWB were recombined by a dichroic mirror (FM02, Thorlabs), deflected by a pair of galvanometers (QS-7, Nutfield Technology), and focused by an objective lens (Plan Apo Infinity, 10 × , Mitutoyo) onto the sample. The illumination power on the sample was 3 mW for the SWB and 0.25 mW for the LWB. The entire SWB power was used for the following PAM imaging experiment on the resolution target and the mouse ear. The illumination numerical aperture (NA) was 0.1 for PAM and 0.06 for OCT. A customized unfocused needle ultrasonic transducer (30 MHz center wavelength; 60% bandwidth; 0.4 mm active element) was used to detect the photoacoustic (PA) signal coupled through water, which was then amplified, digitized at 200 MHz and recorded by an acquisition card (CS1622, Gage).

Fig. 2.

(a) Schematic of PAM/OCT system setup. SC, supercontinuum; DM1-DM3, dichroic mirrors; GM, galvanometer; UT, ultrasound transducer; AMP, amplifier; ADC, digitizer; SM, spectrometer; DC, dispersion compensating slab; M1-M2, mirrors; BD, beam dump. The red dashed square is where spectral filtering components are inserted for multispectral PAM measurement, which is further demonstrated in Fig. 4. (b) Representative PA A-line generated by a 1951 USAF resolution test chart. (c) Measurement of axial resolution, sensitivity, and roll-off effect in OCT. (d)–(e) En face and B-scan images of a resolution chart by PAM and OCT, respectively. The positions of B-scan images are indicated by vertical lines in the en face images. Scale bar, 150 μm.

SC pulsing, raster scanning, and PAM/OCT image acquisition were controlled by the same analog output board (PCI-6731, National Instruments). The light source was pulsed at 25 kHz and synchronized with the scanner, though the latter was triggered at only 5 kHz. Therefore, each scanning position received five SC pulses at a time during imaging. The digitizer acquired a PA A-line signal for each pulse delivered to the sample, and five A-lines were averaged. The spectrometer for OCT acquisition was synchronized with scanning at 5 kHz, and the exposure time was 190 μs, during which the energy from five SC pulses were integrated. Briefly, a single A-line from either channel contained information from five SC pulses, which was designed to improve signal to noise ratio.

We further characterized the system performance. The axial resolutions of the dual-modality system are demonstrated by A-line images of a calibration target and a mirror surface for PAM and OCT, respectively. Figure 2(b) shows a PA A-line from a black bar on a 1951 USAF resolution test chart. The colored layer on the resolution chart is less than 5-μm thick, has strong optical absorption and, thus, serves as an axial impulse for PAM. The two peaks in the PA signal arose from two oppositely propagating acoustic waves generated by the laser excitation. The one traveling away from ultrasound transducer was reflected back by the underlying glass slide. The axial resolution of PAM is quantified as the product between the temporal full width at half-maximum (FWHM) of the PA signal envelope and the speed of sound, giving 51 μm, which is limited by the bandwidth of the needle ultrasonic transducer. Figure 2(c) characterizes the axial resolution of the OCT subsystem and its sensitivity roll-off. A silver mirror was used as the testing sample. We inserted a neutral density (ND) filter (OD = 1.2) in the path of the sample arm so that the interferogram did not saturate the spectrometer. We adjusted the reference arm and fixed the delay between the two arms at 250 μm, our typical imaging depth. The OCT A-line of the mirror reflection displays an axial resolution of 4.6 μm measured in air. The resolution agrees with our expectation for a light source with 75-nm FWHM bandwidth. In addition, the peak of the reconstructed A-line rises 53 dB above the noise, indicating a system sensitivity of 77 dB, after allowance for the double path through the ND filter. We further translated the reference arm and recorded mirror signals at different imaging depths. The roll-off was around 10 dB/mm. We quantified lateral resolutions of the dual system by imaging a 1951 USAF resolution test chart, as shown in Figs. 2(d) and 2(e). The lateral resolutions were 4.5 μm for PAM and 9 μm for OCT. The better resolution of PAM was due to shorter wavelength and larger beam size.

The in vivo imaging capability of the dual system was tested through imaging mouse ears. We initially anesthetized a nude mouse (25 g, Charles River) by 2.5% isoflurane mixed with 3.0 standard liter per minute (SLPM) air and then transferred it to a homemade animal holder. We flattened one of its ears and attached it to the imaging plate. A drop of water was applied to its surface for ultrasound coupling. During the following experiment, the rodent was kept in anesthesia by a mixture of 1.5% isoflurane and 1.5 SLPM air. We used a heating pad to maintain the mouse’s body temperature. All experimental procedures were approved by Northwestern University IACUC.

Figure 3 shows the in vivo imaging results. Due to the limited sensing area of a stationary ultrasonic transducer under optical scanning, we manually translated the imaging stage and sequentially acquired 10 images at different positions and stitched them for a large field of view. Figure 3(a) is the en face PAM image, which was the projection of maximum signal amplitude from each scanning position after Hilbert transform. Arteries and veins bear the major signals, and some of their small branches can be observed. A typical cross-sectional B-scan and an A-line profile are shown in Figs. 3(b) and 3(c), respectively. The OCT en face image, calculated as the mean projection of reconstructed 3D dataset, reveals scattering features, as shown in Fig. 3(d). Glands appear as round dark spots. The OCT B-scan image in Fig. 3(e) shows a more detailed anatomy of the mouse ear. The whole ear is penetrated, while the epidermis, dermis, and cartilage are resolved.

Fig. 3.

Results of in vivo mouse ear imaging. (a) En face PAM image, stitched from 10 acquisitions. (b) PA B-scan taken from location indicated by the green line in (a). (c) Typical PA A-line and its signal envelope obtained by a Hilbert transform. (d) En face OCT image. G, gland; BV, blood vessel. (e) OCT B-scan taken from location indicated by yellow line in (d). ED, epidermis; CT, cartilage; D, dermis. Scale bar, 150 μm, applies to (a), (b), (d), (e) in the 3 dimensions.

Finally, we tested the multispectral measurement capability of PAM, as shown in Fig. 4. We built an adjustable bandpass filter using an equilateral dispersive prism (DP), as shown in Fig. 4(a). The beam from the dispersed SC source was spatially filtered by a pinhole (PH1), allowing only a narrow band within the spectrum passing through. We manually rotated the prism to select different bands and adjusted the aperture size of PH1 to vary the bandwidth. The rotating axis of the prism was carefully chosen to minimize change of beam direction after filtering. We added a second pinhole (PH2) and slightly adjusted the mirror (M3), when necessary, to avoid misalignment. After optimization, we fixed PH1 and M3, and recorded spectra of illumination light on a sample as we traversed through the SWB of the spectrum. As shown in Fig. 4(b), the bandwidth increased from 10 to 25 nm due to nonlinear dispersion. We further adjusted the aperture size between measurements using different bands to maintain a fixed 15-nm bandwidth in the following multispectral experiment. We carefully controlled the illumination power to be 45 μW before measuring the PA signals of the samples using each single narrow band.

Fig. 4.

Multispectral PA measurement with the SC source. (a) Schematic of adjustable bandpass filter inserted in the red dashed square in Fig. 2(a). DP, dispersive prism; PH1-PH2, pinholes; M3, mirror. (b) Spectra of illumination light on sample after passing through the bandpass filter. (c) Comparison between the multispectral PA measurement and absorption spectrum of eumelanin. (d) Comparison between the multispectral PA measurement and the blood absorption spectrum convolved with 15-nm excitation bandwidth.

We placed ex vivo porcine RPE samples (dissected from pig eye, Carolina Biological) and bovine blood (Quad Five, Inc.) on glass slides, covered them with transparent plastic wrap (Glad), added a drop of water for ultrasound coupling, and performed multispectral measurements. We acquired PA A-lines of the sample repeatedly without scanning the illumination beam laterally and recorded the averaged PA signal from 32 acquisitions for each wavelength narrow band to reconstruct the absorption spectrum. Eumelanin, the dominant optical absorber in RPE, and hemoglobin in blood, are the two major contrast sources in PAM [14]. We measured their absorption spectra through optical transmission to compare them with multispectral PAM measurements. Since the absorption spectrum of eumelanin decreases linearly within the visible spectral range, we performed multispectral measurements from 550 to 775 nm in 25-nm steps. Good agreement was demonstrated, as shown in Fig. 4(c). For blood, we targeted its absorption signature between 530 and 600 nm at 5-nm interval, where two peaks and a trough are presented, as shown in Fig. 4(d). To make a fair comparison, we convolved the measured absorption spectrum with that of the source considered of 15 nm bandwidth. Consistency was observed between convolved spectrum and multispectral PAM measurement.

In summary, we developed an all-fiber-based nanosecond SC source for integrated multispectral PAM and OCT applications. We built a dual-modality system and split the SC spectrum for PAM and OCT imaging. We tested imaging capability of the dual system by conducting in vivo experiments on a mouse ear. Preliminary multispectral tests on ex vivo RPE and blood samples also successfully reconstructed appropriate absorption spectra. Although in vivo multispectral imaging of vasculature and melanin was not performed due to the power limitation within the 15 nm narrow bands below 600 nm, by future optimization of the SC source, this may become possible.

We demonstrated that pulsed SC sources might be a promising solution to future development of a combined PAM and OCT system. SC sources render access to a broadband smooth spectrum with extended flexibility, giving rise to multispectral PAM covering 500–2000 nm and ultrahigh resolution OCT. In addition, generating PAM and OCT images with a single probing beam is also achievable which reduces laser exposure time for dual-modality imaging and facilitates clinical translation. The all-fiber-based design further makes this SC source attractive to applications that require compactness and portability. Our future work will focus on optimizing the current SC source design to access higher power, especially at shorter wavelengths.