Researchers have developed a label-free and non-invasive Raman spectro­scopy approach that can acquire microscopic images of biological samples and identify a wide range of biomolecules with unpre­cedented speed and sensitivity. “Our work could lead to a non-invasive, label-free and user-friendly device for clinical use,” said research team leader Dario Polli from Politecnico di Milano in Italy. “This innovative micro­scope, coupled with deep learning-based algorithms, could eventually make it easier and faster to diagnose cancer by allowing the visuali­zation of the chemical constituents of human tissues and cells.”

The new technique is based on coherent anti-stokes Raman scattering (CARS) microscopy. CARS microscopy produces images based on the vibra­tional signatures of molecules by exploiting the interaction between ultrashort laser pulses and biological samples. The new approach provides access to the hard-to-detect region of the vibrational spectrum known as the finger­print region, which spans from 400 to 1800 cm⁻¹. Although many individual compounds can be identified using their vibrational finger­prints in this region, it tends to produce weak signals that are difficult to detect.

“Commonly used techniques in biomedical sciences often require staining, which is not only cumbersome but can also introduce structural and chemical altera­tions that can lead to artifacts, or errors, in imaging and data processing,” said Polli. “Because our system can distinguish between many different chemical species in biological tissues without labels, it could be useful for live cell imaging and analyzing tissue biopsies.” This new work is part of the Crimson project funded by the European Commis­sion, which aims to develop a turnkey imaging device that uses vibrational spectro­scopy for fast cell and tissue classi­fication. The project’s goal is to transform the study of the cellular origin of diseases to enable new approaches that could advance personalized therapy.

As a key step toward this goal, the researchers developed a CARS microscope based on a commercial laser that produces ultrashort pulses with durations of approxi­mately 270 femtoseconds in the near-infrared wavelength range. They designed the micro­scopy system to use laser pulses with a repetition rate of 2 Megahertz (MHz), which is much lower than the 40 or 80 MHz used by most other CARS systems. This lower repe­tition rate reduces photothermal damage to the sample because it creates a delay of 0.5 microseconds between two conse­cutive pulses. It also produces a higher pulse energy and peak intensity at the focal point, which generates a stronger CARS signal and allows a faster acquisition speed.

“The most important advantage of the lower repetition rate is that it allowed us to generate broadband, red-shifted Stokes pulses that cover the whole fingerprint vibrational region by using white-light super­continuum generation in a bulk crystal,” said Federico Vernuccio, doctoral student at Politecnico di Milano. “Compared to other methods, this approach is technically simpler, more compact and robust.” Using a spectral region that is red-shifted compared to standard setups means that higher laser intensities can be used before the onset of photo­damage. The researchers also developed new algorithms that combine standard numerical compu­tational approaches with arti­ficial intelligence. These algorithms retrieve more information from the acquired data and turn it into images that allow different chemical species to be easily distin­guished.

“Thanks to our improvements, the CARS system delivers high-quality images at a state-of-the-art acquisition speed,” said Vernuccio. “Our system has a less than 1 millisecond pixel dwell time without compro­mising sample integrity. This speed is limited by the spectro­meter refresh rate.” To test their system, the researchers used reference samples to compare spectra retrieved with the new microscope with ones acquired using a state-of-the-art, although slower, vibrational spectro­scopy technique. The two methods showed excellent agreement, demons­trating that the new system could deliver spectra at very high speeds with good spectral resolution and chemical specificity.

The researchers then determined the detection limit of their system by acquiring CARS spectra of a set of dimethyl sulfoxide solutions with various concen­trations. The system was able to measure chemical concentration with the unpre­cedented sensi­tivity of 14.1 mmol/liter, around double the sensitivity of other CARS systems working in the fingerprint region. They also showed the system’s ability to distinguish and spatially localize various transparent micron-sized plastic beads based on their vibra­tional signature and took measurements from biological tissues to demonstrate that the technique works on biological samples without inducing damage.

“Our CARS microscope allows label-free imaging with chemical specificity at higher speeds, thus making Raman imaging of living cells more feasible,” said Polli. “This could allow our system to be used to analyze the interactions of cancer cells with immune cells or to characterize how chemo­therapy affects cells, for example.” The researchers are now working to improve their system by creating an even broader wavelength range of Stokes pulses through white-light super­continuum generation. This would improve both the speed of imaging and the number of detectable chemical analytes. They are also working toward commerciali­zation by developing user-friendly software, compact optical sources and designs for a commercial prototype and detection system. (Source: Optica)

Reference: F. Vernuccio et al.: Fingerprint multiplex CARS at high speed based on supercontinuum generation in bulk media and deep learning spectral denoising, Opt. Exp. 30, 30135 (2022); DOI: 10.1364/OE.463032

Link: Dept. of Physics, Politecnico di Milano, Milan, Italy