**1. Introduction**

#### **1.1. Basic concept and working principle**

Various techniques are being currently used for imaging of cells and tissues. Individually, each technique is able to address some aspects of the system under study. For example, optical microscopy is very often used for cell and tissue analysis; it is a cost‐effective method which gives morphological information, but is unable to provide molecular and structural

information. Electron microscopy and atomic force microscopy are high spatial resolution techniques, able to image subcellular compartments; however, they lack chemical specificity. In most cases, fluorescence microscopy, for example, confocal scanning fluorescence micro‐ scopy, is used for cellular visualization. Fluorescence microscopy requires fluorescent labels specifically bound to the substrate under study. Subcellular structures can be visualized, but since each fluorescent label is excited by a different wavelength, the number of structures that can be visualized is limited. The need to introduce fluorophores and their limited sta‐ bility and photo bleaching are drawbacks of using confocal microscopy. Nonetheless, the technique is largely used for cell imaging and imaging of cellular uptake of micro and nanoparticles. There is great need for techniques that provide chemically specific informa‐ tion coupled to spatial information for the visualization of, for example, cellular uptake and localization of biologically active molecules, cellular transport pathways, molecular changes in cancer vs healthy tissues, etc.

Raman mapping (Raman imaging, Raman scanning or Raman micro‐spectroscopy) has recent‐ ly become an emerging imaging technique in biological and biomedical research and applica‐ tions. The Raman effect is based on inelastic scattering of photons when electromagnetic waves interact with atoms or molecules. The small fraction of incident photons scattered inelastically have different frequencies compared to the incident photons. The phenomenon is called Ram‐ an scattering, and the difference in frequency between the incident photons and scattered pho‐ tons is the so‐called Raman shift (cm−1). The Raman shift is related to the vibrational levels of each specific molecule, being used as a fingerprint for molecular identification [1].

In contrast to Raman spectroscopy, which provides discrete chemical information at distinct positions within the sample, Raman mapping provides chemical information coupled with spatial information [2]. Raman mapping is a noninvasive, label‐free technique, with high chemical specificity. In Raman mapping, the laser spot scans the investigated sample area with a preset step size and acquires Raman spectra at every set point. The Raman spectra are then discriminated from each other by chemometric analysis, and the end result is an image of the sample that contains highly precise structural and chemical information. Excitation wave‐ lengths in the visible and near‐infrared range give high spatial resolution (<1 μm), making Raman spectroscopy combined with microscopy an ideal tool for biological samples imaging, and especially for cell and tissue imaging. In this latter case, Raman mapping has important advantages over conventional biological assays: it is a rapid, noninvasive, label‐free technique, which does not damage the cells if using suitable laser wavelengths and power.

#### **1.2. Instrumentation and data analysis**

The most important parameters to ensure the success of a Raman imaging measurement on biological samples are the wavelength and power of the laser, the resolution of the images, and the sample preparation and fixation. The intensity of the scattered radiation is proportional to the wavelength at the power of −4 (~*λ*− 4), meaning that shorter (blue) wavelengths are scattered more strongly than longer (red) wavelengths. Thus, shorter wavelengths generate more photons scattered inelastically, giving thus higher Raman intensities. However, shorter wavelengths typically lead to stronger auto‐fluorescence from the samples, which can mask the Raman signal arising from the molecules of interest. Therefore, a compromise is needed. Hamada et al. studied the influence of 488, 514.5, 532 and 632.8 nm laser excitation wavelengths on the Raman signal yield and background signal for the imaging of living cells [3]. The authors found that the 532 nm excitation is a good compromise between Raman signal intensity and auto‐fluorescence background because it generates strong Raman scattering signals and suppresses auto‐fluorescence. Photodamage caused by light absorption of the biological samples is another important parameter to be considered for choosing the appropriate laser excitation. Puppels et al. [4] found that a 660 nm laser induces no photodamage to cells and chromosomes compared to the 514.5 nm (visible) laser. Even though Raman scattering efficiency decreases with increasing wavelength, recent advances in the design of Raman spectrometers with high optical throughput and highly sensitive CCD (charge‐coupled device) detectors allow measuring spectra and obtaining reasonably high signal strength. Notingher et al. used a 785 nm laser for their measurements on live cells and tissues [5–7]. In one study, they compared the 488, 514 and 785 nm lasers with respect to photodamage of cells and found that the 488 and 514 nm lasers induce photodegradation and reduce the number of living cells; with the 785 laser, cell degradation and auto‐fluorescence were low and the signal intensity was reasonably high [8]. Even though going higher than 785 nm (e.g., 1064 nm) in the laser wavelength would decrease the photodamage of the cells, it would also dramatically decrease the Raman scattering efficiency. The recent literature mentions mostly the use of 785 and 532  nm lasers for cellular mapping [6, 9]. However, when using near‐infrared (NIR) lasers for Raman excitation, cooled deep depletion back‐illuminated CCD detectors are preferred [6] instead of standard back‐illuminated, visible‐optimized CCDs, because of their higher quantum efficiencies (QE) in the near‐infrared (NIR) spectral region (up to 95% with the new Low Dark Current Deep‐Depletion (LDC‐DD) Technology.

information. Electron microscopy and atomic force microscopy are high spatial resolution techniques, able to image subcellular compartments; however, they lack chemical specificity. In most cases, fluorescence microscopy, for example, confocal scanning fluorescence micro‐ scopy, is used for cellular visualization. Fluorescence microscopy requires fluorescent labels specifically bound to the substrate under study. Subcellular structures can be visualized, but since each fluorescent label is excited by a different wavelength, the number of structures that can be visualized is limited. The need to introduce fluorophores and their limited sta‐ bility and photo bleaching are drawbacks of using confocal microscopy. Nonetheless, the technique is largely used for cell imaging and imaging of cellular uptake of micro and nanoparticles. There is great need for techniques that provide chemically specific informa‐ tion coupled to spatial information for the visualization of, for example, cellular uptake and localization of biologically active molecules, cellular transport pathways, molecular changes

Raman mapping (Raman imaging, Raman scanning or Raman micro‐spectroscopy) has recent‐ ly become an emerging imaging technique in biological and biomedical research and applica‐ tions. The Raman effect is based on inelastic scattering of photons when electromagnetic waves interact with atoms or molecules. The small fraction of incident photons scattered inelastically have different frequencies compared to the incident photons. The phenomenon is called Ram‐ an scattering, and the difference in frequency between the incident photons and scattered pho‐ tons is the so‐called Raman shift (cm−1). The Raman shift is related to the vibrational levels of

In contrast to Raman spectroscopy, which provides discrete chemical information at distinct positions within the sample, Raman mapping provides chemical information coupled with spatial information [2]. Raman mapping is a noninvasive, label‐free technique, with high chemical specificity. In Raman mapping, the laser spot scans the investigated sample area with a preset step size and acquires Raman spectra at every set point. The Raman spectra are then discriminated from each other by chemometric analysis, and the end result is an image of the sample that contains highly precise structural and chemical information. Excitation wave‐ lengths in the visible and near‐infrared range give high spatial resolution (<1 μm), making Raman spectroscopy combined with microscopy an ideal tool for biological samples imaging, and especially for cell and tissue imaging. In this latter case, Raman mapping has important advantages over conventional biological assays: it is a rapid, noninvasive, label‐free technique,

each specific molecule, being used as a fingerprint for molecular identification [1].

which does not damage the cells if using suitable laser wavelengths and power.

The most important parameters to ensure the success of a Raman imaging measurement on biological samples are the wavelength and power of the laser, the resolution of the images, and the sample preparation and fixation. The intensity of the scattered radiation is proportional to the wavelength at the power of −4 (~*λ*− 4), meaning that shorter (blue) wavelengths are scattered more strongly than longer (red) wavelengths. Thus, shorter wavelengths generate more photons scattered inelastically, giving thus higher Raman intensities. However, shorter wavelengths typically lead to stronger auto‐fluorescence from the samples, which can mask

in cancer vs healthy tissues, etc.

60 Raman Spectroscopy and Applications

**1.2. Instrumentation and data analysis**

In some cases, for samples that cannot be detected by regular Raman scattering, signal enhancement can be induced. In some situations, it is possible to obtain resonance Raman effects. Such effects take place when the laser excitation wavelength overlaps with the absorp‐ tion band of the molecules due to electronic transitions, and this can lead to increase the Raman intensity by a factor of 103–105 [1]. The phenomenon is called resonant Raman scattering. Consequently, Raman imaging of a resonant molecule can be significantly improved by choosing an excitation laser wavelength in the absorption band region of the molecule. For example, cytochrome C absorbs light at around 520 nm and shows a strong resonance Raman effect when analyzed using a 532 nm laser. This property can be used to image its intracellular distribution. Other examples of molecules that can benefit from strong Raman resonance effects are the carotenoids, chlorophylls, vitamin B12 and heme proteins [3, 10].

Another way to obtain signal enhancement in Raman spectroscopy is to use surface‐enhanced Raman spectroscopy (SERS) or coherent anti‐Stokes Raman scattering (CARS). In SERS, it is possible to reach high enhancements of the Raman intensity (by a factor of 107 or more) when the molecule of interest is adsorbed onto or in the very close vicinity of plasmonic metallic nanostructures such as silver and gold [11]. This effect significantly lowers the detection limit of molecules. CARS is a nonlinear optical effect in which two lasers, a pump laser and a Stoke laser, are overlapped and strongly focused onto the sample to generate the CARS signal. When

the difference in frequency between the pump and the Stoke lasers is tuned to the exact value of a vibrational frequency within the sample, strong enhancements of the CARS signal can occur [1].

The instrumental resolution is very important for cell and tissue mapping, in order to be able to image cellular and subcellular structures. The lateral resolution is limited by the wavelength of the laser and the numerical aperture of the objective used for the experiment, while the axial resolution is given by the instrument aperture (slit or pinhole) and the refractive index of the immersion medium. Currently available Raman spectrometers can go down to 200 nm for the lateral resolution and 500 nm for the axial resolution [2].

Cells and tissues could be fixed on specific substrates for usage over longer periods of time. The most important requirements for a substrate suitable for Raman imaging are as follows: (a) transparency in the visible and near‐infrared region of the light spectrum; (b) low back‐ ground signal to avoid overlapping with the Raman signals from the sample; and (c) suitability for cell culture growth or tissue fixation. Calcium and magnesium fluoride (CaF2 and MgF2) and quartz are the preferred substrates for Raman imaging. Glass and plastic substrates are not recommended because of high background signals [2, 12]. Zinc selenide (ZnSe) has the disadvantage of weak cell adherence [12]. A variety of cell fixation methods has been so far reported can be used: paraformaldehyde, methanol, methanol:acetone, formalin, air‐drying and cytocentrifugation [2, 6, 13]. For live cell imaging, special instrument setups, in which cells are confined in a sterile chamber and kept at 37°C and a 5% CO2 atmosphere to ensure viability [14, 15], have been reported.

After taking the pixel‐by‐pixel Raman spectra, the raw dataset needs to be processed in order to identify the key molecules in the sample and based on their spectral fingerprint, to generate the false color Raman images. Since no label is used, the pixel‐to‐pixel spectral variations are small and multivariate methods of analysis need to be employed to get the Raman images from the dataset. Several approaches are currently used: principal component analysis (PCA), self‐ modeling curve resolution (SMCR), K‐means cluster analysis (KMCA), hierarchical cluster analysis (HCA), divisive correlation cluster analysis (DCCA), vertex component analysis (VCA), fuzzy C‐means cluster analysis (FCCA) and linear discriminant analysis (LDA) [6, 16– 20].

Here we aim to highlight the recent advances of Raman mapping and provide an overview on its emerging applications, which range from single cell and tissue imaging to medical diag‐ nosis, including cancer detection. Some applications that will be discussed include:


the difference in frequency between the pump and the Stoke lasers is tuned to the exact value of a vibrational frequency within the sample, strong enhancements of the CARS signal can

The instrumental resolution is very important for cell and tissue mapping, in order to be able to image cellular and subcellular structures. The lateral resolution is limited by the wavelength of the laser and the numerical aperture of the objective used for the experiment, while the axial resolution is given by the instrument aperture (slit or pinhole) and the refractive index of the immersion medium. Currently available Raman spectrometers can go down to 200 nm for the

Cells and tissues could be fixed on specific substrates for usage over longer periods of time. The most important requirements for a substrate suitable for Raman imaging are as follows: (a) transparency in the visible and near‐infrared region of the light spectrum; (b) low back‐ ground signal to avoid overlapping with the Raman signals from the sample; and (c) suitability for cell culture growth or tissue fixation. Calcium and magnesium fluoride (CaF2 and MgF2) and quartz are the preferred substrates for Raman imaging. Glass and plastic substrates are not recommended because of high background signals [2, 12]. Zinc selenide (ZnSe) has the disadvantage of weak cell adherence [12]. A variety of cell fixation methods has been so far reported can be used: paraformaldehyde, methanol, methanol:acetone, formalin, air‐drying and cytocentrifugation [2, 6, 13]. For live cell imaging, special instrument setups, in which cells are confined in a sterile chamber and kept at 37°C and a 5% CO2 atmosphere to ensure viability

After taking the pixel‐by‐pixel Raman spectra, the raw dataset needs to be processed in order to identify the key molecules in the sample and based on their spectral fingerprint, to generate the false color Raman images. Since no label is used, the pixel‐to‐pixel spectral variations are small and multivariate methods of analysis need to be employed to get the Raman images from the dataset. Several approaches are currently used: principal component analysis (PCA), self‐ modeling curve resolution (SMCR), K‐means cluster analysis (KMCA), hierarchical cluster analysis (HCA), divisive correlation cluster analysis (DCCA), vertex component analysis (VCA), fuzzy C‐means cluster analysis (FCCA) and linear discriminant analysis (LDA) [6, 16–

Here we aim to highlight the recent advances of Raman mapping and provide an overview on its emerging applications, which range from single cell and tissue imaging to medical diag‐

**•** Single cell and microorganism imaging [2, 18, 24, 25], including evidentiation of subcellular

nosis, including cancer detection. Some applications that will be discussed include:

**•** Stem cell research, especially stem cell differentiation [6, 21–23]

lateral resolution and 500 nm for the axial resolution [2].

occur [1].

62 Raman Spectroscopy and Applications

[14, 15], have been reported.

compartments [23]

**•** Identification of cell cycle phase [26]

**•** Monitoring of cell death [27, 28]

**•** Cellular responses to drugs [9]

20].

