Coherent anti-Stokes Raman Spectroscopy (CARS) microscopy; Stain-free histopathology. — Basal cell carcinoma (BCC) is the most common malignancy in Caucasians. Non-linear microscopy has been previously utilized for the imaging of BCC, but the captured images do not correlate with standard hematoxylin and eosin (H&E) staining. This year we have developed a novel algorithm to post-process images obtained from dual vibration resonance frequency (DVRF) CARS measurements to acquire high-quality pseudo H&E images of BCC samples (Fig. 1). We adapted our CARS setup to utilize the distinct vibrational properties of CH3 (mainly in proteins) and CH2 bonds (primarily in lipids). In a narrow-band setup, the central wavelength of the pump laser is set to 791 nm and 796 nm to obtain optimal excitation. Due to the partial overlap of the excitation spectra and the 5-10 nm FWHM spectral bandwidth of our lasers, we set the wavelengths to 790 nm (proteins) and 800 nm (lipids). Non-resonant background from water molecules also reduces the chemical selectivity which can be significantly improved if we subtract the DVRF images from each other. As a result, we acquired two images: one for “lipids” and one for ”proteins” when we properly set a multiplication factor to minimize the non-specific background. By merging these images, we obtained high contrast H&E “stained” images of BBCs. Non-linear microscope systems upgraded for real time DVRF CARS measurements, providing pseudo H&E images can be suitable for in vivo assessment of BCC in the future.
Figure 1. CARS images of mouse epidermis (A-C) and human basal cell carcinoma (D-F). A, D: “CH2“ images; B, E: “CH3“ images; C, F: merged color DVRF-CARS images (red: CH2-CH3, blue: CH3-CH2). Scale bar: 50 µm.
Multimodal stain-free mosaic imaging of malignant tumour in the skin – BCCs often have poorly defined borders, the clinical assessment of the tumor margins can be challenging. Therefore, there is an emerging demand for efficient in vivo imaging techniques for the evaluation of the tumor borders of BCC prior to and during surgeries. This demand might be met in the near future by non-linear microscopy techniques (such as auto-fluorescence (AF) and second harmonic generation (SHG) mosaic imaging) utilizing our novel, fibre-laser based, hand-held 3D nonlinear microscope system (FiberScope).
Figure 2. Multimodal (AF+SHG) mosaic image of human basal cell carcinoma comprising 6x10 microscope images with an overall area of 2,5x 4.2 mm2. Yellow: detected AF signal, purple: SHG signal of collagen.
This year, we have measured AF and SHG signal of collagen on 10 ex vivo healthy control and BCC skin samples and compared the images by different quantitative image analysis methods. These included integrated optical density (IOD) measurements on AF and SHG images and application of fast Fourier transform (FFT), CT-FIRE and CurveAlign algorithms on SHG images to evaluate collagen structure. In the BCC samples, we found significantly lower IOD of both the AF and SHG signals and higher collagen orientation index utilizing FFT. CT-FIRE algorithm revealed increased collagen fiber length and decreased fiber angle while CurveAlign detected significantly higher fiber alignment of collagen fibers in BCC. These results are in line with previous findings which describe pronounced changes in the collagen structure of BCC. In the future, these novel image analysis methods could be integrated in our FiberScope imaging system for sensitive and specific identification of BCC.
Figure 3. Normal human skin (A, C) and basal cell carcinoma (B, D) collagen SHG images processed by different numerical algorithms. Image size: 420x420 μm2.
Femtosecond fiber lasers for nonlinear microscopy. — During the last couple of years, we have been developing a pulsed Yb-fiber oscillator and amplifier system with a variable repetition rate in the 1 to 36 MHz range in collaboration with our industrial partner, R&D Ultrafast Laser Ltd. First we used this laser in our novel, hand-held 3D nonlinear microscope system (FiberScope) for applications in dermatology and nanomedicine. The laser delivers ~0.4 ps pulses at around ~1030 nm, which makes it an excellent candidate for two-photon imaging of GCaMP in neuroscience as well. GCaMP is a genetically encoded calcium indicator, and its main advantage is that it can be genetically specified for studies in living organisms. In the absence of calcium, this protein is in a poorly fluorescent state, but after Ca2+-binding, it is brightly fluorescent. In Figure 1, we show 3D reconstructions of 2PE auto-fluorescence images of brain slices comprising GCaMP in its certain neurons in the “poorly fluorescent state”. The high quality of the images results from the fact that our fiber laser was operated at a relatively low, ~6 MHz repetition rate.
Figure 1. 2PE auto-fluorescence images of GCaMP expressing neurons excited by our mode-locked Yb-fiber laser oscillator - amplifier system operating at ~6 MHz repetition rate (central wavelength ~1030 nm, pulse duration: ~0.4 ps, average power on sample: ~14 mW). Left: 3D reconstruction of z-stack images with dimensions of 600 µm x 600 µm. Right: 3D reconstruction of z-stack images with dimensions of 50 µm x 50 µm.
Coherent anti-Stokes Raman spectroscopy (CARS) microscopy. — Last year, CARS imaging of the CH2 bonds of lipids was successfully used to perform a quantitative analysis of the myelin loss in a cuprizone model of sclerosis multiplex (SM) depending on the drug treatment. This year, we have aimed at label-free imaging of proteins as well as NO in living tissues, such as in murine skin or in the brain. To this end, we optimized our CARS imaging system for protein and NO detection, which required several modifications and improvement in our optics and the laser system. Note that the concentration (and hence the CARS signal) of NO is orders of magnitude smaller than that of the lipids in myelin, that is why NO imaging is a challenging task. For demonstrative purposes, we show two CARS images of murine skin recorded for “lipid” and “protein” settings (Fig. 2, left) and one CARS image of an artificial sample containing NO in a gaseous state (Fig. 2, right). In the latter experiment, NO was generated from an aqueous solvent of sodium nitroprusside (SNP) after exposing it to white light for a few tens of seconds before the imaging.
Figure 2. Left: CARS images of the stratum corneum of murine skin for “lipid” (green) and “protein” (blue) settings. Right: CARS image of NO distribution in a SNP solvent containing artificial tissue.
Currently, we are working on the development of more sophisticated, non-resonant background-free imaging techniques for 3D distribution detection of low-concentration molecules (e.g. NO), such as FM-CARS or our newly developed TD-CARS method.
Broadband beam steering mirrors in nonlinear microscopy. — Broadband dielectric beam steering mirrors are key components for nonlinear microscopes comprising tunable Ti:sapphire lasers. This year we characterized dispersion of a few mirror samples by spectral interferometry and showed how they affected our imaging quality. As an example, we show the measured group delay of a BB1-E03 mirror (product Thorlabs Inc.) for s- and p-polarized light (Fig. 3, up), and how it affects the pulse shape after four reflections at around a resonance wavelength (Fig. 3, down). In practice, one obtains considerably lower signal in a nonlinear microscope when the peak intensity of the optical pulses is reduced by such unexpected dispersive effects (see Fig. 3, down).
Figure 3. Up: Measured group-delay vs. wavelength functions of an ultra-broadband dielectric mirror (two reflection, AOI: 45 degrees, Thorlabs Inc. BB1-E03). Down: Calculated intensity vs. time functions of transform limited 130 fs pulses being reflected on the same dielectric mirrors for s-polarized light at 45 degrees of AOI at different central wavelenghts.