14 Clinical multimodal CARS imaging

: Nonlinear contrast methods such as two-photon excited autoﬂuorescence (in combination with ﬂuorescence lifetime imaging) and second-harmonic generation have been combined with a further multiphoton contrast mechanism called coherent anti-Stokes Raman scattering. We describe the principle and the instrumentation for its implementation in tomographs for dermatological applications and present multi-modal optical skin biopsies


CARS
Raman scattering, named after C. V. Raman, is an inelastic scattering process that originates from light's interaction with molecular vibrations.The interaction leads to a characteristic energy shift of the photon energy which contains information about the vibrational modes of a system.A shift to higher energies leads to so-called anti-Stokes Raman scattering whereas a shift to lower energy results in the Stokes signal (Fig. 14.1).Raman spectroscopy was realized already in the late 1920s [1] to identify molecules and became, in particular with the advent of the laser, a fundamental method commonly used in chemistry, solid state physics, and biomedicine [2][3][4].A typical Raman spectrum of complex tissue like human skin in vivo is illustrated in Fig. 14.1 [5].It represents a unique signature of the molecular tissue composition.Linear Raman scattering however, is a very inefficient "low-signal process".Therefore, several Raman spectroscopy methods have been developed to increase the sensitivity [6], among them nonlinear Raman techniques like coherent anti-Stokes Raman scattering (CARS).In contrast to spontaneous Raman, CARS is based on a coherently driven transition which involves three photons with different frequencies: pump, probe, and Stokes photons.Technically, these photons can be provided by (i) one (spectrally broadband or broadened) laser, (ii) two lasers, or (iii) three lasers with different frequencies (ω pump , ω probe , ω Stokes ).
The experimental setup must ensure the spatial and temporal overlapping of the pulses and focal volumes, respectively, in the Raman medium.To reduce the complexity of the setup, pump and probe photons from one excitation source are often used (ω pump = ω probe ), and a second source is employed to generate the Stokes photons.Resonant CARS signals are generated in the case where the difference frequency Ω = ω pump − ω Stokes of pump and Stokes photons with frequencies ω pump and ω Stokes , respectively, matches the frequency of a Raman-active molecular vibration [7].The CARS signal itself is generated at the anti-Stokes Raman frequency ω CARS = ω pump + ω probe − ω Stokes (Fig. 14.1) [8].
In the case of high molecular concentrations, the CARS signal is stronger than the spontaneous Raman signal because the CARS signal arises from the coherent addition of all signals from multiple molecules.In contrast, the total spontaneous Raman signal is the sum of signals from individual molecules without coherent addition.The coherent addition of the individual CARS signals is additive when phase matching is fulfilled, i.e., when the condition ∆k = k pump + k probe − k Stokes − k CARS = 0 is satisfied (Fig. 14.1).The CARS signal propagates in a beam along the direction k CARS while the total spontaneous Raman signal is emitted in all directions.In practice, this leads to the possibility to collect CARS signals up to 10 6 times faster than spontaneous Raman scattering.This allows faster (µs beam dwell time per pixel instead of seconds per pixel) and more detailed imaging (more image pixels) [9].The total CARS signal, however, also contains an inherent nonresonant background [10].This nonresonant background arises from off-resonance transitions which also generate signals that coherently add up, but are molecule unspecific (Fig. 14.1).Thus, the sensitivity is reduced by the ratio of resonant to nonresonant background (R/NRB).The nonresonant back-Fig.14.1: Top: energy schemes of (i) spontaneous Raman, (ii) resonant CARS and (iii) nonresonant CARS background, (iv) phase-matching condition for the CARS process; bottom: Raman spectrum of human skin (with kind permission adapted from [5]).
ground causes the CARS spectrum to differ from spontaneous Raman spectra.This is often regarded as a problem and a drawback for the application of CARS.Therefore, several methods have been developed to minimize and suppress the nonresonant background in the CARS signal or correct the recorded signals for nonresonant contributions [11][12][13][14][15].

CARS microscopy
Historically, it was assumed the above mentioned phase-matching condition ∆k = 0 can only be fulfilled in a noncollinear beam geometry.In such a beam geometry with the excitation beams entering the back-focal plane of the focusing optics at different positions and angles (which required a nonfilled back-focal plane) the effective numerical aperture (NA) is lowered.Such a CARS microscope setup was first demonstrated by Duncan et al. in 1982 [16].However, this first CARS microscope was complicated and the tuning of the excitation wavelengths was challenging.Additionally, the configuration-immanent low NA limited the spatial resolution [16].
Interestingly, almost two decades later, Sunney Xie's group at Harvard showed that the phase-matching condition can also be fulfilled in a collinear excitation geometry using high-NA focusing optics [17].The collinear excitation geometry not just simplifies the setup but also allows for diffraction limited focusing.The limited interaction length of the (CARS pump and CARS Stokes ) excitation radiation and the nonlinear intensity dependence of the CARS signal ensure that signals are generated only inside the focal volume.This intrinsic three-dimensional imaging capability (optical sectioning capability) is the basis for high-resolution CARS microscopy [17,18].By scanning the focal volume across the sample with a laser scanning setup (or moving the sample across the fixed focal region) and collecting CARS signals spatially resolved in forward and/or backward direction, images based on the CARS-signal intensity per pixel can be generated.
Different CARS microscope setups have been developed using mostly (i) ps pulses (ps-CARS) or (ii) fs pulses (fs-CARS), (iii) a combination of both or (iv) single broadband pulses which contain CARS pump and CARS Stokes photons [7,18,19].
While ps-CARS provides in general a higher spectral selectivity than fs-CARS, the use of fs pulses is favorable for a multimodal approach where CARS and additional nonlinear optical image modalities like two-photon excited autofluorescence (2P-AF) and second-harmonic generation (SHG) are intended to be implemented in one common system [20].Furthermore, the low spectral resolution of fs-CARS has been countered by the development of methods to increase and even tune the spectral resolution with fs pulses in the range of 10-100 cm −1 [21].This increased spectral resolution is possible because the spectrum of the vibrational excitation is not determined by the whole width of the fs pulse spectra but only by the effective spectrum of their temporal overlap.This temporal overlap can be altered by individually chirping the pump and Stokes pulses.By inducing equal linear chirps on pump and Stokes pulses a constant instantaneous frequency difference exists between them.This "spectral focusing" results in a vibrational excitation centered at the instantaneous frequency difference with an instantaneous bandwidth of the chirped pulses which is only determined by the stretched pulse duration but not by the whole bandwidth of the broadband pulses [22].Multimodal optical imaging enables simultaneous recording of fluorescent and nonfluorescent (but Raman-or SHG-active) molecules to provide corresponding and complementing information on a sample.In biological samples, the CH 2 transition in particular has been used as a contrast mechanism for CARS which allows detecting nonfluorescent lipids [7,23].CARS microscopy was recently introduced to clinical imaging and has been evaluated as a new in vivo diagnostic tool for skin diseases [24][25][26][27][28].

Setup of CARS tomographs
The general setup of a CARS imaging system consists of two parts: the excitation source(s) and the imaging module (Fig. 14.2).The excitation source(s) for CARS can be realized by "single beam CARS" [19] using a single broadband fs oscillator which provides the CARS pump and CARS Stokes photons.More common is a two-beam setup.Technically, this can be a combination of an fs oscillator (e.g., an 80 MHz fs Ti:sapphire laser) with (i) an optical parametric oscillator (OPO) [17], (ii) a photonic crystal fiber (PCF) [29], (iii) an additional fs laser or by a single dual-output laser [30].
In vivo CARS imaging of human skin has been realized by two-beam setups with OPO and PCF technology [24][25][26][27] as described below.
OPO technology: The OPO contains a periodically poled potassium titanyl phosphate (KTP) crystal which is used to generate so-called signal and idler pulses at longer wavelengths.Two temporally synchronized outputs are provided by the OPO, the residual fundamental oscillator pulses (which can be employed as CARS pump pulses) and the newly generated signal pulses which can be employed as CARS Stokes pulses.The advantages of the OPO are: (i) high output power, in particular a high spectral power density, (ii) the possibility for computer-controlled tunability and (iii) a high pulse-to-pulse stability.However, the large footprint of the OPO is nonoptimal for a compact device.
PCF technology: CARS Stokes photons can also be provided from PCFs.In particular, coupling the pulse energy of an fs oscillator into the PCF core can result in a strong spectral broadening of the transmitted pulses.Due to the small core diameter of about 1-2 µm and the small group velocity dispersion of these fibers close to their zero dispersion wavelength(s), the high optical intensities are mostly maintained over the length of the fiber [31].Over this long interaction length nonlinear optical effects such as self-phase modulation, stimulated Raman scattering and four-wave mixing [32] are induced.This leads to strong broadening of the output spectrum -also called white-light continuum or supercontinuum, which is temporally synchronized with the fs oscillator.Advantageously, already moderate mean powers of the fs oscillator are sufficient for a strong spectral broadening which, in combination with the original pulses, can be used for CARS excitation in the spectral range from the socalled fingerprint (< 1500 cm −1 ) up to the high wavenumber region around 3000 cm −1 .The employed PCF fibers are commercially available and integrated in easy to handle and cost-efficient modules (e.g., FemtoWHITE CARS, NKT Photonics, Denmark) with small footprint.However, their integration and stable long-term operation is challenging due to the high requirements regarding the beam pointing stability of the incoming fs laser beam.Even in a stable breadboard design, minor temperature differences can affect this beam pointing stability (thermal drift) and thus the PCF coupling efficiency.Additionally, limited output powers (limited by material damage threshold) and a low spectral power density -superimposed by a reduced spectral amplitude stability (spectral amplitude noise) -challenge the use for imaging.
The imaging module in Fig. 14.2 is based on beam scanning.Briefly described, the incoming laser beam experiences an angular deflection by galvanometer scanners which is translated through the scan-and tube-lens system into an angular deflection inside the back-focal plane of the focusing optics.The focusing optics itself focuses the incoming collimated laser beam into a diffraction limited spot in the focal plane or more specifically in the object plane.Inside the focal spot the interaction of the radiation with the molecules can induce (e.g., CARS, 2P-AF and SHG) signals.These can be collected in epi direction by the same focusing optics, separated from the excitation radiation by dichroic beam splitters and detected by PMTs.Since the position of the focal spot depends on the angle of incidence of the excitation beam inside the back-focal plane, raster-scanning can be performed point-by-point (and line-by-line) for two-dimensional imaging.Moving the axial position of the focusing optics allows for a movement of the object plane inside the tissue.

Requirements for CARS imaging
For CARS imaging the interaction of CARS pump and CARS Stokes photons with the molecules in focus is necessary.When CARS pump and CARS Stokes photons are generated by OPO or PCF technology they are temporally synchronized but not overlapped in time and space.The temporal overlapping can be realized by an optical delay line in one of the beams which compensates the optical path difference.The required spatial overlap of CARS pump and CARS Stokes pulses in the focal volume inside the tissue can be realized by the collinear superposition of both beam paths by a dichroic beam splitter (in this context called beam combiner, see Fig. 14.3).

First clinical CARS tomographs
The first CARS tomographs for in vivo imaging of human skin were realized by extending the imaging capabilities of the clinical tomograph DermaInspect (JenLab, Germany).Two different implementations with this rigid but stable imaging platform are put into practice by the introduction of the DermaInspect CARS-MPT -one based on OPO and one on PCF technology (see Fig. 14.4).
The OPO setup employs a Chameleon OPO (APE GmbH, Germany) which is pumped by a high power fs oscillator (MaiTai HP, repetition rate 80 MHz, pulse duration about 100 fs, Newport Spectra Physics, USA).The signal photons of the OPO are used as CARS Stokes photons which are combined (collinear alignment) with the residual oscillator pulses (CARS pump photons) by a beam combiner (Fig. 14.  ensured by tuning the length of a temporal delay line.By tuning both the center wavelength of the OPO signal pulses and the pump pulses, CARS excitation of molecular vibrations in the range between about 2500 cm −1 and 4000 cm −1 can be achieved.For CARS skin imaging, the symmetric CH 2 stretch vibration (at 2845 cm −1 ) and the OH-water resonance (at about 3250 cm −1 ) are probed.The maximum output power of the OPO is about 120 mW, which is attenuated by the internal optical power attenuator of the scan detector unit of the DermaInspect to about 17 mW.The maximum optical power applied to the skin is limited to 50 mW (P pump ± P Stokes ≤ 50 mW), which is below harmful power levels [33].

3). Temporal overlap of CARS pump and CARS Stokes photons inside the focal volume is
The excitation sources of the second DermaInspect-based CARS tomograph are a "wavelength extension unit" (WEU, Newport-Spectra Physics, USA) containing a PCF module [34] and an external fs oscillator (MaiTai XF-1).In addition to the PCF module, the WEU includes further manually adjustable modules, e.g., an optical delay line and power attenuators.After entering the WEU, the fs oscillator output is split into two parts.Part (I) is used directly as the CARS pump laser passing the power attenuator and the delay line.Part (II) -with a mean optical power of about 500 mW -is coupled into the PCF for supercontinuum generation.Since its core diameter is just 1.4 µm, beam pointing stability is essential for a high coupling efficiency and thus a stable output power level.Technically, this requires a stable vibration-free optical breadboard, which is provided by the design of the tomograph DermaInspect and a temperaturecontrolled environment.The latter is of importance to minimize temperature-induced drifts of the optical components.In accordance with the specifications of the PCF (FemtoWHITE CARS, NKT Photonics, Denmark) the PCF-pump wavelength of the fs oscillator is set to 800 nm.For CARS excitation, only a small part of the supercontinuum is needed.This required part is isolated by (i) a long-pass filter (LP780) and (ii) a band-pass filter (BP1045/30).
The collimated and spectrally shaped output then provides the CARS Stokes beam with maximum spectral power densities at a wavelength of about 1035 nm.Collinear aligned with the CARS pump beam, lipid-CARS contrast at 2845 cm −1 can be generated.

The flexible multimodal CARS tomograph
The requirement for more flexibility, compactness and mobility of CARS tomographs is fulfilled by the second generation imaging device [27] -the MPTflex CARS (see Fig. 14.4).The design is based on the established flexible platform of the multimodal (2P-AF, SHG and FLIM) tomograph MPTflex [35].Briefly described, it consists of a mobile and compact housing (I) and a scan head (II) which is connected to the housing (I) by an articulated mirror arm (III).The mirror arm provides free-space beam delivery and significant flexibility that are necessary for a high accessibility and time-efficient imaging (Fig. 14.4).The CARS excitation sources of the MPTflex CARS are similar to those described above of the DermaInspect CARS-MPT with WEU extension.However, the desired flexibility and compactness of the MPTflex CARS required a stacked optical breadboard design and active beam stabilization modules.More details on the setup are given in Fig. 14.5.WEU: wavelength extension unit; SU: safety unit; BS: beam splitter (with kind permission adapted from [27,28]).
The optical system setup inside the opto-electronic housing includes a lightweight carbon-fiber breadboard with a two-level design.The first level holds the Ti:sapphire fs oscillator.Its output (1100 mW) is guided by a periscope to the second level.A polarizing beam splitter inside the WEU splits the pulse train into two parts where the first part (about 500 mW) is fed into the core of the PCF module.Its collimated spectrally broadened output provides the CARS Stokes photons.As mentioned before, coupling the laser into the small diameter of the PCF core is technically challenging.Minor changes to the beam pointing induced by thermal drift strongly influence the coupling efficiency and thus the stability of the output spectrum.Stable operation of the system is ensured by the integration of an active beam stabilization.Based on a feedback algorithm, the transmission of the PCF is monitored and a motorized kinematic mirror mount (M1), directly behind the fs oscillator, is electronically adjusted to optimize transmission.The PCF spectrum can be additionally monitored by a fiber-coupled spectrometer.
The CARS pump pulses are provided by the residual part of the fundamental beam.These pulses pass a polarization-maintaining optical power attenuator and an optical delay line, which are both computer controlled.The polarization orientation parallel to the CARS Stokes beam is set by a half-wave plate.Before entering the articulated mirror arm the CARS pump beam passes a second beam stabilization (M2), which centers the centroid of the beam profile at the center of the back-focal plane of the focusing optics.This beam stabilization is necessary to compensate displacements due to internal mechanical tolerances, tensions inside the mechanical support structure and alignment tolerances of the articulated mirror arm.The feedback signal is provided by a detector located inside the scan head, close to the back-focal plane of the focusing optics (see Fig. 14.5) [36].The collinear superposition of CARS pump and CARS Stokes beams is realized by the superposition of the centroids of both beams on the surface of a beam combiner (SP850 or alternatively SP950 from Fig. 14.3).It is important to note that angular deviations of the CARS Stokes beam direction can be induced by thermal drift as well.Since the long distance of the following beam path in the articulated mirror arm translates these deviations into a spatial displacement in the back-focal plane of the focusing optics, it results in a noncentered illumination of the image area.Furthermore, due to the induced deviation between the angle of incidence of CARS pump and CARS Stokes beams at the back-focal plane, the required spatial overlap of CARS pump and CARS Stokes in the focal volume at the sample position is disturbed.Thus, a third active beam stabilization is required, which is incorporated into the mechanical mount of the beam combiner (motorized kinematic mirror mount M3).The same sensor as for M2 is used for M3 in a time-multiplex manner.This is the basis for the spatial overlap of the focal volumes in the two-beam setup of the MPTflex CARS.The required temporal overlap for the interaction of CARS pump and CARS Stokes photons in the object position is achieved by adjusting the motorized delay line.In case sole 2P-AF and SHG imaging is required, CARS imaging can be suppressed by blocking the CARS Stokes beam with an internal computer-controlled mechanical shutter.

Multichannel-detection
The detection is typically realized with a two-channel PMT-based detection module, where channel (1) is used for CARS and channel (2) for 2P-AF and SHG signal acquisition [27].The demand for even more detection channels triggered the development of a four-channel PMT-based detection module for simultaneous detection in individual channels.Fig. 14.5 presents the individual detection channels for 2P-AF, SHG pump , and CARS signals.The included PMTs have a small footprint and provide a fast temporal response that is necessary for fluorescence lifetime imaging (FLIM).The signal light inside the detection beam path is discriminated from the residual backscattered excitation light by a short pass filter (SP720).A long-pass dichroic beam splitter (LP565) separates the CARS signal from the 2P-AF/SHG signal.The 2P-AF/SHG beam path is further divided by a long-pass dichroic beam splitter (LP405) into a 2P-AF-detection channel (1) with a short-pass (SP600) filter to block the residual CARS signals.The SHG-detection channel ( 2) is narrowed by a 40 nm (FWHM) band-pass filter centered at 390 nm to further reduce the weak wing of the superimposed AF spectrum.To suppress non-CARS signals a 13 nm (FWHM) band-pass filter centered at 650 nm is used to narrow the CARS-detection channel (3).(Channel (4) is not employed (n.e.) but offers further detection possibilities.)The detector electronics were developed for simultaneous single photon counting (SPC) as well as time-correlated single photon counting (TCSPC) to enable FLIM.Images with 512 × 512 pixels are recorded at a mean dwell time of about 25 µs/pixel.The corresponding lifetime measurements can be accomplished in parallel to SPC operation by internal TCSPC hardware with a standard image reso-lution of 128 × 128 pixels (alternative resolution settings are possible).In this chapter, lifetime imaging is presented for detection channel (1) only.However, signals from all four channels can be recorded using the integrated TCSPC router electronics.The analysis of the FLIM data shown below is done with the following settings: double exponential best-fit, 3-fold binning.
Ex vivo imaging is demonstrated using the inverted scan head position (Fig. 14.5).The animal studies were carried out according to ethical standards.Murine tissue samples were retrieved about 5 h post mortem.Brain, liver, spinal cord, abdominal fat, intestine and muscle tissue from a hairy mouse were placed on a microscope coverslip which was fixed by the sample adapter ring.XY and Z positioning of the tissue relative to the optical axis of the system was realized by the integrated motorized stage with maximum vertical and axial travel ranges of 5 × 5 mm² and 5 mm, and a resolution (step size) of 12.7 µm and 1.25 µm, respectively.Images at sample depths of 5-30 µm were acquired with optical powers (P pump + P stokes ) in the range of 10-40 mW without averaging.For the benefit of optimal print visualization, individual images are contrast enhanced using ImageJ software [37].
Human skin imaging is realized by placing an adapter ring with a fixed coverslip in between skin and focusing optics.Immersion oil (type 518 F; Zeiss, Germany) and water are placed at the interface focusing optics/coverslip and coverslip/skin, respectively.Imaging is performed on healthy human skin (volar side of male forearm) at a skin depth of about 70 µm.The maximum output power (P pump + P Stokes ) is limited by the safety unit (SU) to 50 mW at the skin.

Demonstration of the spatio-temporal overlap
In the case where the spatio-temporal overlap is fulfilled, the signal spectrum may include CARS signal(s), 2P-AF signals, SHG signals (of the CARS pump und CARS Stokes fundamentals), and the sum frequency (SFG) from both excitation beams (see Fig. 14.6 [27]).As a test sample for the quality of the spatio-temporal pulse overlap at the sample position, urea crystals (grown from urea solution) can be used.If there is only spatial but no temporal overlap, the urea crystals produce the SHG signals of the CARS pump (from 800 nm) and CARS Stokes (from, e.g., 1035 nm) pulses at 400 nm and 517 nm, respectively.In the case of a good spatio-temporal overlap, an additional SFG signal is produced at 451 nm.The strength of this signal can be used to optimize the spatiotemporal overlap by fine adjustment of the spatial overlap on the beam combiner and subsequent readjustment of the spatial overlap at the focal position.CARS contrast at 2845 cm −1 can then be tested using a sample with a high concentration of lipids such as grease.In this case, the CARS signal is generated at a wavelength of about 651 nm.It is worth noting that during skin imaging the SHG signal of the CARS Stokes as well as  [27]); bottom: in vivo spectrum of human skin at the epidermal-dermal junction (adapted from [26]).
the SFG pump+Stokes signal can be present in the spectral range of 400-600 nm, which is mostly used for 2P-AF detection of the skin.However, in practice, the SFG signal is not visible inside the epidermis (absence of collagen) and just slightly visible inside the spectrum of epidermal-dermal junction of human skin [26], as can be seen in Fig. 14.6.In particular, with animal studies (reduced thickness of skin, e.g., for mice) these collagen fibers and/or muscles can be abundant and can create a significant structure "cross talk" between the detector channels.In this case, an effective suppression of the SHG Stokes and SFG pump+Stokes signals inside the 2P-AF image can be accomplished by enhanced optical filter design such as multiband rejection filters or by time gating [38].The latter can be realized by the inherent TCSPC hardware, which allows the temporal separation of the instantaneous SHG and SFG signals (fast components) from the decayed fluorescence lifetime components.

Ex vivo imaging
Multimodal imaging of fresh murine tissue is shown in Fig. 14 and thus strong CARS signals are additionally present inside the myelin sheaths of the spinal cord, intracellular lipid droplets/vacuoles in liver tissue, and in adipocytes between muscle tissue.Furthermore, the images from muscle tissue show highly periodic structures (possibly from myofilaments) which are almost identical for the SHGand 2P-AF-images.This congruence is most likely induced by nonlinear interaction of both CARS pump and CARS Stokes excitation light at the muscle fibrils producing an SHG Stokes signal (at a wavelength of 517 nm) and possibly also an SFG pump+Stokes signal (at a wavelength of 451 nm).
These experiments present the detection capabilities of the system.Of course these ex vivo images can also be generated in vivo, as has already been demonstrated with living mice with the 2P-AF and SHG imaging capability of the tomograph MPTflex [39].

In vivo human skin imaging
Multimodal imaging of healthy human skin at the dermal-epidermal junction at a depth of 70 µm is demonstrated in Fig. 14.8.Inside the 2P-AF channel, basal cells are arranged in a circular manner surrounding elastin fibers.Further contrast is presented by the associated lifetime image (FLIM), which superimposes the intensity contrast of the 2P-AF image by visualization of a heterogeneity of lifetimes in pseudo-colors.The spatially resolved representation of these lifetimes helps to distinguish between cell types or structures in skin.For instance, the basal cells with a high content of melanin reveal short lifetimes (orange color coded) and thus can be clearly isolated from the elastin fibers (blue-green color coded), which reveal longer lifetimes [40] compared to melanin.The image from the SHG-detection channel visualizes corresponding endogenous collagen fibers by detecting SHG from the nonlinear light-collagen interaction.The image obtained in the CARS-detection channel presents additional morphological information about the skin.Fine structured details of separated cellular details but also the roundish pattern of the dermal papillae are visible.Especially in between the fibers of the dermal-epidermal structures further information can be extracted from the sample.The images in Fig. 14.8 show no significant 2P-AF detector channel cross talk.However, deeper skin regions provide the possibility to produce SHG Stokes and SFG pump+Stokes signals by interaction with collagen structures of the dermis.From the emission spectrum in Fig. 14.6 (b) it can be expected that their signal strengths are small compared to the 2P-AF signal.Still, if required they can be eliminated by the same methods discussed above.

Chemical contrast
The spectral resolution of the fs-CARS system is determined by the spectral width of the CARS pump laser.A typical pulse width of about 100 fs (transform limited) at 800 nm presents a spectral width of about 150 cm −1 which is significantly broader than the line width of many Raman resonances in the fingerprint region [41].Spectral parts of the excitation spectrum which do not overlap with the Raman-line width excite only the nonresonant background and reduce the overall chemical contrast.This challenges the application of fs pulses for imaging in the fingerprint region [7].However, in the high wavenumber region around 3000 cm −1 the situation is different.For instance, the broad Raman resonances of lipids (2845 cm −1 ) and proteins (2950 cm −1 ) [42] have a width which is similar to the spectral width of the fs pulses.Thus, the sensitivity and the ratio R/NRB is improved for these resonances, which has been verified by Chen et al. who compared the lipid contrast for ps-and fs-CARS systems [20].
These results are also confirmed by our experiments.Ex vivo fresh murine myelin sheaths (see Fig. 14.7) reveal a contrast ratio of > 7 : 1, similar to contrast ratios determined by other groups [20].For imaging on human skin the ratio R/NRB is significantly smaller, apparently due to lower concentrations of lipids.However, in some skin regions such as the exit of sebaceous glands a higher lipid concentration is present (sebum) leading to R/NRB ratios in the range between 10 to 20 [28].

Conclusion
This chapter described multiple multimodal imaging systems capable of acquiring optical sections of human skin in vivo as well as animal tissue samples.In a single scan of a few seconds a multitude of modalities (2P-AF/FLIM, SHG, and CARS) generate a multitude of images with subcellular resolution.A variety of applications can benefit from these systems, such as rapid ex vivo imaging of tumor biopsies [43] and the in vivo investigation of skin barrier problems [44].

Fig. 14 . 4 :
Fig. 14.4: CARS tomographs for in vivo imaging of human skin.Left: DermaInspect CARS-MPT with (i) OPO and (ii) PCF technology; middle: MPTflex CARS with PCF technology; right: demonstration of the flexibility of the scan head.

Fig. 14 . 6 :
Fig. 14.6: Top: excitation and emission spectra of in vitro test samples.All signals are normalized to their maximum.The band-pass filters (BP) to isolate the CARS Stokes excitation spectrum and the CARS emission spectrum are schematically indicated by dashed lines (adapted from[27]); bottom: in vivo spectrum of human skin at the epidermal-dermal junction (adapted from[26]).