HTML
-
In this work, an ultra-compact endomicroscopic set-up for multimodal non-linear endoscopy combining CARS, SHG, and TPEF imaging has been designed, using piezo-based resonant spiral scanning of an FoV of 180 μm at 1 fps for 1000 × 1000 pixels per frame. It provides sub-micron and hence sub-cellular spatial resolution at an exceptionally high overall laser transmission of 65% from the laser output into the sample plane in a compact housing of 2.4 mm in diameter and 39 mm in length, which is comparable in size or even smaller than endomicroscopes for two-photon excited fluorescence14, 37. In combination with signal collection efficiency above 80 % for CARS, SHG, and TPEF signals within the collection NA of 0.55, this endoscopic setup has to the best of our knowledge the highest throughput so far demonstrated for flexible endoscopic systems for multimodal CARS endoscopy. Key elements of the novel concept are (i) the DCDC fiber for the FWM background free single-mode delivery of pump and Stokes lasers, (ii) a special all-fiber laser source for the delivery of long ps pulses of high peak power providing alignment-free spatial and temporal pulse overlap, (iii) specially designed optics for coupling the laser into the separate cores of the DCDC fiber and recombination of both excitation lasers in the endomicroscopic objective, and (iv) a special, highly corrected endomicroscopic objective employing a resonant fiber scanner. Key parameters of the setup are summarized in Table 1.
Specification Value FoV 180 μm WD 30 μm Diameter head 2.4 mm Length of head 39 mm Frame rate 1 fps Pixel resolution 1000 × 1000 Lateral resolution ≤900 nm NA 0.55 Core NA > 0.1 Laser attenuation / 1 km fiber < 10 dB Signal attenuation / 1 m fiber < 0.2 dB Power @ sample 42 mW Laser transmission 65% Table 1. List of key specifications of the endomicroscopic set-up
The set-up is guiding pump and Stokes laser pulses in separate cores of a specially designed DCDC fiber for eliminating the non-phase matched FWM generation at the anti-Stokes wavelength in the delivery fiber of 1 m in length, which is typical for solid core fiber-delivery15, 16, 18, enabling small endoscopic heads without requiring excitation filtering7. Still, the solid DCDC fiber is efficiently guiding high-power pulses even at a small bending radius. Bending loss measurements were performed at diameters down to 12.7 mm with fundamental mode loss of less than 0.1 dB turn-1 at 1030 nm for the Stokes core and less than 0.2 dB turn-1 at 795 nm for the pump core as described in detail in the supplementary information (Supplementary Information: 1. Bending loss measurements and Figs. S1 and S2). It is highly stable and can be manufactured at a low cost with high reproducibility, in comparison to alternative hollow-core fibers, which are significantly more difficult to manufacture and have higher attenuation11, 24. The fiber operates in single-mode for two core sizes of 4.8 and 6.3 µm, matching the CARS excitation wavelengths of 795 nm and 1030 nm. The small core size results in diffraction-limited spot sizes in the focal plane of the probe. To avoid both the loss of temporal pulse overlap and pulse broadening, and the loss of peak intensity by self-phase modulation (SPM) in the delivery fiber, an all-fiber laser source of 20 and 60 ps pulse duration at the pump and Stokes wavelengths has been employed for the fiber delivery of synchronized pump and Stokes pulse trains. The dispersion within 1 m of the DCDC fiber results in a time delay significantly less than the pulse length of the 60 ps Stokes pulse so that the temporal overlap is maintained. The long ps-pulses at a repetition rate of 1 MHz are customized not only to virtually eliminate SPM but also to provide pulse peak powers equivalent to an 80 MHz pulse train of the same average power, but of 500 fs duration for the efficient TPEF and SHG excitation. Still, the spectral resolution is about 30 cm-1, providing sufficient chemical contrast for CARS microscopy at 2850 cm-1. The low repetition rate requires averaging of 10-50 frames to obtain a high-quality image by illuminating all pixels in the outer region of the spiral scan. However, this problem can be solved by using a laser with a higher repetition rate, which, however, has not yet been available in the experiments. The endomicroscopic objective design enables a sub-micron spatial resolution and a sufficient colour correction in the spectral range from 1000-3000 cm-1, except for the diffractive element which allows the coupling of pump wavelengths in the range from 781-868 nm and of Stokes wavelengths from 1030-1050 nm. The whole experimental set-up provides an exceptionally high transmission of 65% of the excitation lasers, which is even well comparable with high-end laser scanning microscopes, especially designed for the NIR excitation in non-linear imaging and outperforms alternative endoscopic concepts11.
The endoscopic images of 1 μm polystyrene beads (Fig. 4) and the biological specimens (Fig. 5) demonstrate the efficient SHG, TPEF, and CARS signal generation as well as the signal collection in epi-direction over the whole FoV at diffraction-limited resolution. Images from a human head and neck tissue specimen, a dura mater section of ovis aries and of the model organism Galleria mellonella larvae (Fig. 5) clearly visualize the tissue morphology. The image of human epithelial tissue (Fig. 5e) clearly demonstrates the resolution of individual cells and cell nuclei. The image quality of the endomicroscope is comparable to images acquired with a laser scanning microscope using a 20× apochromatic objective of a NA of 0.8 proving the high spatial resolution and signal collection efficiency of the probe. While the working distance of 30 µm in water is rather small, it is not limited by the optics and can be increased to 200 µm in water. However, for focusing deeper into the tissue, the excitation laser intensity at the focus is reduced by scattering and an increase of the point spread function is caused by aberrations. In addition, the tissue surface area emitting the signal photons is increasing, which results in a significant reduction of the collection efficiency, because signal photons are only collected within an area of about 30 µm diameter by the DCDC fiber taking into account the magnification of the optics of 5.5.
In summary, the reported all-fiber-based endoscopic set-up for multimodal non-linear endoscopy represents a promising design for routine clinical imaging applications such as surgical guidance and in vivo diagnostics. The probe is compliant with ethanol and ETO sterilization, but for clinical use approval according to the Medical Device Regulation is required.
-
The experimental set-up for multimodal non-linear endomicroscopy consists of a compact all-fiber laser for CARS microscopy (CARS-1032, AFS, Germany) of a pulse duration of 20/60 ps, an average power of 20/100 mW at 795/1030 nm and a repetition rate of 1 MHz. The laser delivers synchronized pump and Stokes pulses in a PCF fiber connected to the custom coupling unit by a standard FC/PC connector. The laser beam is collimated by an achromatic lens (f = 3 mm, #84-127, Edmund Optics, USA). In the collimated beam path, a 750 nm long-pass filter F1 (FELH0750, Thorlabs, USA) filters FWM from the fiber. The Stokes laser power is adjusted by a short-pass dichroic mirror (#86-696, Edmund Optics, USA), and a dichroic mirror (SEM-FF757-Di01-25×36, Semrock, USA) is used to reflect CARS/SHG and TPEF signals. The signals are filtered by bandpass filters and focused onto the PMT detector (H10723-20MOD, Hamamatsu, Japan) by an achromatic lens. The individual filters are summarized in Table 2.
Signal Bandpass filter CARS FF02-675/67-25 TPEF1 FF01-578/105-25 TPEF2 FF01-458/64-25 SHG FF01-514/3-25 Table 2. list of bandbass filters for selection of the nonlinear signals, all filters are from Semrock, USA
The fibers and the connecting wires are protected by a medical endoscopic tube, and the endoscopic head is sealed to enable the clinical application. The whole length of the DCDC fiber from the endoscope head to the coupling unit is about 1 m. The endoscopic head consists of the fiber scanner and the endomicroscopic objective, which are mounted in a stainless steel tube with an outer diameter of 2.4 mm and a length of 38.94 mm. The scanning procedure is realized using resonant piezo scanning techniques14, 18, 19.
-
The linear blazed gratings which are used to couple both laser wavelengths in the separated cores of the double-core double-clad fiber in the coupling unit and, also, to overlap the collimated, but slightly deflected beams again in the probe head were fabricated by a grey tone lithography on thin glass wafers (Fraunhofer IOF, Germany). The required grating periods g could be derived from the diffraction law in combination with the applied focal lengths f of the collimating lenses between the fiber and the grating,
$$ g = \frac{{f\Delta \lambda }}{a} $$ ∆λ is the wavelength difference between pump and Stokes wavelengths, and a corresponds to the distance between the centers of the two cores. With the wavelength difference of 235 nm, a core distance of 24 µm, and the respective focal lengths of the incoupling lens L3 of 4 mm and of the GRIN lens of 4.17 mm, the required grating periods were 39.4 and 40.8 µm. The design height of the blazed structure was chosen to provide optimum diffraction efficiency in the first order for a wavelength of 927 nm. The diffraction efficiencies were measured using collimated Gaussian beams of 0.5 mm diameter from single-mode fiber-coupled diode lasers at 1065, 780, 635, and 532 nm. The laser beams transmitted the gratings, and the optical power levels were measured for the different diffraction orders with a photodetector in reference to an area on the glass wafer without a diffractive structure. Diffraction efficiencies of more than 80% were measured in the first diffraction order for the wavelength of 780 and 1065 nm, and in the first two diffraction orders for 532 and 635 nm.
-
The polystyrene bead sample was prepared by depositing a high-density bead solution onto a standard cover glass.
Biological sample measurements were performed using a fixed assembly. The probe head was placed on a microscope stage and held by a mechanical arm for precise positioning. The biological samples investigated in this study were prepared according to the regulations and disposed of properly after measurements. The Galleria mellonella larvae were embedded in distilled water for cryo-sectioning. Afterwards, the tissue sections were placed on the microscopy glass substrate.
-
Due to the 1 MHz repetition rate of the laser, the number of sampling points is limited in the outer part of the spiral scan range. For high-quality images, 50 frames are averaged. All images are sampled at the resolution of 1000 by 1000 pixels and an average of 50 sampling points per pixel, for two different FoV (70 µm and 180 µm). For further details, see the Supplementary Information, Fig. S4.
-
The LSM images in Fig. 5d-LSM and 5e-LSM additionally provided were taken from similar areas of the same sample, using a commercial Laser Scanning Microscope (LSM 510, Zeiss, Germany) equipped with a ps pulse laser system, which consists of a Ti: sapphire laser (MiraHP, Coherent, USA), pumping an optical parametric oscillator (APE, Germany). For CARS microscopy at 2850 cm-1 and for TPEF imaging (emission filter FF01-458/64-25, Semrock, USA), the following laser parameters have been used: pump 672.5 nm, 35 mW at the sample; Stokes 832 nm, 40 mW at the sample, 1 to 2 ps pulse duration, 76 MHz repetition rate. A 20×/0.8 NA plan apochromatic objective lens has been used. CARS signals have been collected in forward direction. The LSM set-up has been described in detail previously38.