Skip to main content

Innovative contrast imaging by non-linear optics (NLO) for the observation of biological tissues in vivo and in real time, at cellular and molecular levels

Final Report Summary - CARS EXPLORER (Innovative contrast imaging by non-linear optics (NLO) for the observation of biological tissues in vivo and in real time, at cellular and molecular levels)

The CARS EXPLORER primary goal was to deliver innovative, minimally invasive light-based technologies and solutions to researchers in life sciences and clinicians in order to bring their practice to a higher level of reliability, precision and cost-effectiveness.

The CARS EXPLORER project aimed at developing imaging tools based on applying nonlinear optics (NLO). Laser pulse interactions with tissues provide unique possibilities in biomedical research such as an absence of sample preparation, a direct multiparametric visualisation with molecular specificity and cellular resolution, and a deep sample penetration.

However, to fully take advantage of the NLO technology, we had to solve major technological problems related to the propagation of ultra-short pulses in tissues, to the sensitivity improvement of signal detection and to the nature and the extraction of information coming from the biological samples.

As such, the CARS EXPLORER project:

- was split into five research and technological development (RTD) work-packages (WPs): three were intended to overcome specific technological problems with:
(1) the development of dedicated methodology using phase shaping for NLO imaging of deep biological tissues;
(2) the development of specific photonic crystal fibre optics for the excitation, delivery and collection of NLO signals; and
(3) the extraction of relevant information from NLO signals generated in biological tissues.

The two other WPs had to determine the assets and constraints in NLO imaging through appropriate experimental biological models. These models were also assessing each technological improvement made on the microscope and endoscope prototypes. Last, to bring the concept to the diagnostic level, we have explored the molecular and morphological NLO signatures associated with tumour development in skin cancer.

- was implemented throughout a qualified and dedicated interdisciplinary consortium. CARS EXPLORER bring together a highly interdisciplinary team including physicists, biologists and clinicians as well as a private company with complementary expertises necessary for the delineation, implementation, and proof-of-concept programme. The academic participants were strongly benefited from state of the art equipment, intellectual property (IP) and know-how of a small and medium-sized enterprise (SME). The CARS EXPLORER management structure was defined to optimise concerted actions and to guarantee high flexibility and responsiveness to any unforeseen circumstances requiring immediate adjustments of the work plan.

- was ambitioned to provide scientific outcomes social impacts at European Union (EU) level. The development of pulse shaped NLO-based microscopy and PCF endoscope was novel and challenging: it addressed fundamental physical points as well as scientific outcomes in biomedical sciences. In addition, the CARS EXPLORER project had strategic and economic impact by providing informative methods for cancer diagnosis.

Although the CARS EXPLORER project was challenging, we have made outstanding progress, solving technical, methodological and biological issues to bring us far closer to reaching our main goal. Appropriate photonic crystal fibres have been designed and fabricated, fibre-based scanning microscope setups have been demonstrated for the production of sensitive NLO contrasting images and innovative analytical methods have permitted us to extract pertinent observable NLO images for human melanoma diagnostic purpose.

Project context and objectives:

After the development of prevention policies by public health agencies, combating diseases is strongly dependent on tools for their early diagnosis and prognosis. This is exemplified in cancer, where the earlier a tumour is detected the better are the chances for the patient's cure. The current procedures involve biopsies and the diagnosis of a malignant lesion based on histo-pathological observations in conjunction with increasingly sophisticated biological methods, which take into account genetic, chemical and immunological features. However, tumourigenesis is a multistep process driving progressive morphological transformation of normal tissues in tumours. Consequently, biologists, clinicians and surgeons need to explore the body with appropriate non-invasive tools to enable visualisation of the tissue morphology at a cellular level.

A variety of non-invasive imaging methods is now available for the visualisation of whole organisms. They are based on different forms of energy interacting with tissues such as magnetic resonance imaging, computed tomography or positron emission tomography. These scanning methods can identify relatively small tumours, but a tumour must still have a certain mass to be detected, even by PET. One of the most inexpensive and rapid ways to image a particular molecule or cell in an organism is through induction of bioluminescence using genetic engineering; but this technique has strong restrictions: the tracked objects need to use reporter probes that directly (i.e. luminescence) or indirectly (i.e. fluorescence) emit photons. Even if they prove to be effective in animal tumour models, they have significant restrictions for translation into the clinic.

Non-linear optics principles provide promising ways of overcoming this restriction: the intense coherent radiation available from lasers allows non-linear optical responses from biomaterials on ultra-fast time scales. These systems allow photons to be added, subtracted, mixed and manipulated by exploiting molecular resonances inherent to the sample. Developments of NLO imaging technologies will exhibit incomparable advantages such as a direct multi-parametric visualisation at a molecular level with cellular resolution, real-time assistance to patients and a limited technological environment. NLO has recently experienced rapid technical development, opening innovative imaging strategies as illustrated by two-photon excitation microscopy in cell biology. This augurs well for substantial possibilities for the other NLO modes.

Despite the advances in NLO imaging technologies, significant challenges remain to be overcome. These include a number of engineering and biological obstacles, around which the CARS EXPLORER consortium was built. We have delineated the following major goals:

- Making dedicated methodology for NLO imaging of biological samples
There are different NLO contrast mechanisms possible within biomaterials, which need to be explored and improved to optimise the signal to noise ratio, the spatial resolution and the depth of penetration in scattering tissues. Keeping all of the requirements at the optimal level necessitates the development of innovative strategies at each step of the image formation. Moreover, spatial and temporal pulse shaping will be extensively used to obtain the optimal NLO contrast.

- Making specific fibres available for the excitation, delivery and collection of NLO signals
To go further into the exploration on living samples, we aim to develop endoscopes that use NLO contrasts. There is a real challenge to develop a fibre-based microscopy system that could provide a local sub-surface and microscopic vision of the tissue and consequently serve as a guiding tool for biopsies. Specific photonic crystal fibres (PCF) will be designed, manufactured and characterised for NLO applications by the CARS EXPLORER consortium (WP3). We will investigate the use of hollow core (HC) photonic crystal fibres (PCF) whose transmission, dispersion properties and non-linear effects can be controlled.

- Generating useful biological information from NLO signal
Another challenge is to extract the useful information from the NLO signals generated in biological tissues and to use them to create images. By sharing the theoretical and experimental knowledge required to perform the computer programs for data acquisition and processing, the CARS EXPLORER aims at implementing analytical tools allowing optimised NLO applications for life science.

- Making NLO endoscopy work for tumour screening
Although molecular markers will still provide advances in diagnosis and prognosis, there is now more and more evidence that reliable diagnosis can be established on the basis of these sole changes during tumourigenesis. In contrast to a two-photon excitation mode, NLO techniques are by themselves highly demanding and have not been systematically explored even at a basic level in a biomedical context. These contrasts will be generated from biological samples and in various physiological conditions.

Project results:

During the project, each partner has been mainly implicated in the development of dedicated and specific tools and methodologies in order to provide appropriate techniques and expertises to assemble together during the last period. The work performed and major results are the following:


Based on requirements identified by the biologist partners, different microscope prototypes have been developed by physicists. A preliminary request was to get enough spatial resolution to interpret the images of biological samples. This has satisfactorily been achieved to observe fixed biological samples. The other request was to gain enough sensitivity in order to generate fast images of living samples without long temporal signal integration. These prototypes are now available to report biological observation of different samples. In the meantime, biologists have set and prepared biological samples to image them with these different techniques.


In order to improve the NLO capabilities, phase and polarisation control have set on the incoming laser pulses. Two directions have been investigated:

(1) a polarisation control to determine the orientation of the NLO reporter and

(2) a phase control to realise single pulse CARS microscopy.

These innovative schemes are now possible on the new experimental setup, namely the 'pulse shaped NLO microscope'.


Hollow-core fibres according to the conventional design (i.e. 7-cell core) for the two wavelength ranges of interest for the CARS work - 800 nm and 1060 nm have been fabricated. Moreover, an accurate measurement of the fibre core-guided-mode dispersion has also been developed for the first time. We also designed and fabricated fibres with smaller cores (three-cell core) to investigate whether this would offer advantages for the CARS project, and performed the dispersion measurements on these fibres as well.

Moreover, we have developed optimised nonlinear photonic crystal fibres for generating supercontinua useful for the Stokes pulse in CARS. This work has proceeded rapidly, and we believe that we have made significant progress which will be of wide interest to the photonics community.


The radiometric model is strongly dependent on CARS signal analysis through fibres. The preliminary results have revealed a strong background. Further studies are in progress to analyse the complete properties of such signal, its variability among fibres and its spectral signature.

An alternative workplan has been adopted which relies on implementation of a software allowing experimentalists to develop and validate routine during the tests in the optics labs. This open-software has been setup during this first period and will be implemented by partners.


The specific objectives have been to explore the Raman signature and the potential of NLO microscopy to detect tumour development. The bank of frozen tumour tissues has been extended. A standard protocol has been established for the preparations of tissue samples. Raman and multiplex CARS spectral analyses have been performed on different biological samples. As of general interest, the establishment of a quantifiable index of proliferation has been explored. Raman microscopy on mouse melanoma and skin showed promising results allowing us to define clusters distinguishing tumour area from the surrounding tissue. We were also able to correlate Raman spectrum with Ki67 and H&E staining, which report proliferation cell activity and nucleic acid distributions, respectively.

Moreover, the development of an inducible melanoma in the mouse with a fluorescent read out has been achieved.

WP1 - Assets and constraints in imaging tissues by NLO

WP1 coordinator: Didier Marguet (Inserm 1A)
Main partners involved: Anne-Marie Schmitt-Verhulst (Inserm 1B), Hervé Rigneault (CNRS), Andreas Volkmer (USTUTT), François Lacombe (MKT)

General objectives

The goal of WP1 was to explore and optimise the experimental conditions as well as to delineate the limits to generate fast and contrasted images in biological samples by using NLO imaging techniques. WP1 was focused on NLO microscopy (CARS, SHG, THG) and aims at providing inputs to WP2 and WP3 for building an endoscope prototype and extract the biological relevant information from the NLO signals (WP4). WP objectives were:

- to evaluate and to define the physical parameters allowing proper NLO imaging in live tissues (toxicity versus incident light power, nature of the contrast, reachable depth and signal recovery;
- to provide WP3 and WP4 the inputs for a sound radiometric model of the complete NLO scanned endoscope, and guidelines for system calibration;
- to identify NLO molecular and morphological signatures in microscopy associated with different physio-pathological situations in skin samples and lymphoid organs.

Progress towards objectives

During the first period, the work achieved was mostly focused:

(1) on setting protocols to evaluate cellular damages induced by light during NLO observations. Both the CARS imaging setups and the spectral analysis procedures have been successfully implemented and tested;
(2) on implementation of different experimental setups with appropriate spatiotemporal resolution to explore NLO signals in biological samples;
(3) on imaging cells with the different experimental setups.

However, we have faced technical, methodological and biological difficulties with the implementation of the nonlinear imaging techniques which combine scanning methods of imaging with the physical constraints to keep efficient the generation of NLO contrasts using different combinations of pulsed laser sources. Moreover, the optimisation of the procedures to prepare biological samples and the identification of pertinent analytical parameters for Raman and NLO imaging required the development of specific imaging methods.

Along the project, we had extensive discussions about the relative advantages of different nonlinear imaging modalities. Originally the project was mainly focused on CARS. Therefore, we have investigated different experimental modalities to generate the contrast and adjusted the modality as function of different parameters such as the SNR ratio or the signal generated. As a result of these discussions we focused our efforts to record image based on:

(1) CARS imaging modality;
(2) SRS imaging modality;
(3) mixing TPEF, SHG and CARS-related imaging modalities.

Altogether, the progress made by the consortium allowed us to conduct an in-depth evaluation and systematic comparison between CARS, SRS, and spontaneous Raman spectra on biological samples and to achieve four main objectives:

(1) the setting of prototype for 'real-time' non-linear microscope for CARS, SHG and THG in epi direction;
(2) the development of differential CARS system with two optical parametric oscillators to subtract the non-resonant background;
(3) the first successful demonstration of SRS imaging providing quantitative image contrast;
(4) the comparison of normal and cancerous tissue samples;
(5) the evaluation of hyper-spectral CARS imaging on biological samples.


Overall, we have established the following major results:

- By setting up multimodal imaging, we have developed new methodologies to get chemical and morphological information with minimal preparation of the biological samples.
- The contrast generated from different modalities of imaging depends of the wavenumber range and the nature of the sample. In the lipid regions, the strong emitted signal allows the resonances detection with CARS and SRS. In the fingerprint region, due to the presence of a strong non-resonant contribution in CARS, we need statistical tools to reveal relevant signatures. As a complementary technique, we have developed and successfully demonstrated SRS imaging at USTUTT, which provides a quantitative image contrast.
- Finally, using hyper-spectral CARS imaging (within a selected region of interest in the sample) and clustering analysis, we can access the full wealth of Raman spectral information.

WP2 - Optimal NLO contrats and propagation using pulse shaping

WP2 coordinator: Hervé Rigneault (CNRS)
Main partners involved: Andreas Volkmer (USTUTT), Jonathan Knigh (Bath), François Lacombe (MKT)

General objectives

- to optimise focusing in scattering tissues for deep NLO imaging thanks to spatial wavefront control (in microscope);
- to develop temporal pulse shaping to optimise the ultra-short pulse delivery in tissues (chirp pre-compensation for single point NLO in CARS, SHG and THG);
- to obtain wide field NLO images with z sectioning capabilities (in SHG and THG microscopy);
- to use temporal pulse shaping for optimal pulse delivery at the fibre end (endoscope).

Progress towards objectives

For optimal light matter interaction, it is important to minimise the pulse width at the sample. This requirement is particularly important when addressing nonlinear contrast mechanisms such as second harmonic generation (SHG) that is optimally obtained with fs pulses.

The aim of this work has been to provide a versatile microscope that can control the temporal pulse profile at the focal point by implementing a pulse shaper on a microscope setup. The shaper is aimed at controlling the amplitude, the phase and the polarisation of the incoming light. Such control is intended to optimise the light matter interaction at the microscope objective lens focal point. The pulse shaper implemented was able to correct from most of the phase distortion that was induced by the setup and was able to shape the desired temporal pulse profile.

For applications in medicine there is a requirement to perform nonlinear imaging of buried organs. Unfortunately, these buried organs are not accessible from the outside because of the scattering properties of tissues. We choose to develop nonlinear imaging in an endoscope scheme in which the same fibre is used to deliver the excitation beams and to collect the CARS emitted signal in the backward detection. We have used an original custom made hollow core photonic crystal fibre design and drawn within the framework of the CARS EXPLORER project. We have demonstrated that double clad HC-PCFs are suitable to perform CRS in an 'endoscope like' scheme. The key features of these nanostructures fibres being:

(1) to generate no wave mixing for temporally overlapping pulses travelling down to the hollow core fibre;
(2) to extend the spectral range and improve the back coupling efficiency thanks to the double clad.

We believe such fibres and their forthcoming development to be suitable for spectroscopy and imaging in coherent Raman spectroscopy endoscopes.

Coherent anti-Stokes Raman scattering (CARS) microscopy is shown to be a powerful method in molecule specific imaging where molecular vibrations are used as a contrast mechanism. In this method, the specimen is illuminated by two light fields, called pump and Stokes signal. Their frequency difference is tuned to a Raman-active transition of the chemical compound of interest. This results in the emission of a so-called resonant anti-Stokes (AS) signal. However, this signal is affected by an electronic nonresonant background which decreases the contrast and distorts the spectrum as compared to Raman. There are different strategies to perform wide field z-slicing, one of which is using the temporal focusing scheme. Originally we have retained this approach as being the one to be implemented within the CARS EXPLORER project. Nevertheless progress in phase imaging has brought us to use an alternative, yet easier, way to perform z-slicing and background rejection. We present here an alternative that relies on the 'phase detection in wide field CARS imaging'. We have demonstrated that wide field CARS imaging resolved in phase can perform background free vibrational image and revealed the phase shift associated to the molecular resonance.

Such a scheme can show interest to achieve fast background free CARS image for biological application.


Two key results have been achieved within the CARS EXPLORER second working period:

- We have demonstrated for the first time a CARS and SRS endoscope scheme that permits to (1) deliver the pump and Stokes beams and (2) collect the CARS and SRS signals.
- We have demonstrated wide field background free CARS and phase retrieval in an original scheme that is compatible with living samples such as cells and tissues.

WP3 - Photonic crystal fibres

WP3 coordinator: Jonathan Knight (Bath)
Main partners involved: Hervé Rigneault (CNRS), Andreas Volkmer (USTUTT), François Lacombe (MKT)

General objectives

The goal of WP3 was to use the assets of photonics crystal fibres (PCF) to deliver ultra-short pulses for NLO endoscope use. The two major advantages of PCF for this purpose are:

(1) hollow core (HC) PCF which minimises the NLO interaction of launched pulses inside the fibre and enables high-power femtosecond pulses to be delivered as solitons, and
(2) the dispersion control achievable with PCF which will be used to maintain femtosecond pulses at the fibre end.

Based on the fundamentals on NLO light generation and scattering of WP1, WP3 is strongly connected to WP4 (image processing) to provide biologically relevant images that will be used in WP5.

The main objectives were to:

1. develop a fibre incorporating a pathway for the return signal;
2. demonstrate a prototype multichannel fibre imager;
3. demonstrate a NLO endoscope utilising CARS, SHG and THG as contrast mechanisms.

We initially had drawn a first-attempt fibre incorporating pathway for return signal and had made the first-attempt multiple-hollow-core fibre. Of these two, the first was developed more highly through the period - efforts to scale up the second were reduced as we investigated alternative directions (including multiple solid core fibres, coupled hollow-core fibres, and stacked conventional fibres.)

Progress towards objectives

We have explored three possible methods by which return signal paths could be incorporated into HC-PCF for signals whose wavelengths lie outside of the limited transmission window of the fibre. Of these, the most successful was construction of an all-silica fibre in which an outer cladding layer formed of air was introduced. We have now demonstrated this for hollow-core fibre for a range of different pump wavelengths. The design was optimised through the use of extensive ray-tracing calculations. Furthermore, we have demonstrated the use of this fibre for experimental imaging utilising second harmonic generation, SRS, CARS and TPEF.

Although this fibre has been demonstrated to be useful and effective in a single-core configuration, and indeed enables measurements which were almost certainly not previously possible, extending the concept to the highly multi-core designs required for proximal-scanning endoscopy will be demanding and may not be feasible. We have not investigated the use of the return channel in multicore configurations during the course of the project.

We have demonstrated a range of prototype multichannel imagers using different concepts. In doing so, we have chosen various configurations either because of their obvious advantages for nonlinear endoscopy (e.g. hollow-core fibre arrays), or because of what we could learn from studying them (e.g. double-core hollow core fibre, polarisation-maintaining solid-core fibre array), or because they are most readily scalable (solid-core array). All of these designs are potentially useful for imaging, but none are easily scalable to the thousands or tens-of-thousands of cores array sizes required for endoscopic imaging. Some of the main points learned during this investigation are:

- It is possible to make multicore arrays (imaging fibres) out of solid or hollow (bandgap) fibres.
- The arrays have had up to 31 cores (solid core fibres) and 7 cores (hollow cores). Further scaling of core size is more likely for solid core fibres.
- We can easily observe coupling between cores in both of these types of fibre, under selected conditions.
- In any array using air holes as a part of the fibre structure, there is unintentional variation in the air hole size across the array.
- In solid core fibres, there is a trade-off between confinement and cross-talk / polarisation. Closely-spaced small cores separated by air holes have increased cross-talk and depolarisation due to surface scattering at the silica./.air interface.
- In hollow-core fibres, there is no prospect for small and closely-spaced /.decoupled cores. Fabricating large arrays of hollow cores looks difficult for any separation.

Therefore, there are some real and quantifiable advantages in using some of the air-silica fibre designs developed during the course of CARS EXPLORER for nonlinear microscopy. They also offer advantages for endoscopy. However, the difficulty of scaling the fabrication to the large number of closely-spaced cores which would be required for high image quality using proximal scanning endoscopy makes it unlikely that they will be used in this configuration.

Throughout the CARS EXPLORER project, we discussed the merits of various nonlinear imaging modalities. In addition to CARS, we have investigated SRS, TPEF, and SHG, all using hollow-core fibre. In most cases, we also detected the signal by coupling back through the same fibre using the return path.

These measurements were performed using either pico-pico, pico-femto or femtosecond pulse excitation.

Strengths of the schemes demonstrated were:

- We recorded high-quality low-noise images using fibre delivery which were uncorrupted by the effects of nonlinear fibre response.
- We controlled the dispersion of the fibre in the experiments so that the effects of dispersion were low.
- We collected the signal through the same fibre in many cases, so that the system would be directly useful for distal-scanning endoscopy.

Weaknesses of our demonstrations were the following:

- The signal collected by the multimode high-NA collection channel was relatively weak. In some cases, this restricted the range of biological samples which could be studied to samples offering a strong signal. We improved this substantially through optical engineering of the distal end, but this has implications for the probe size. This could be further improved.
- The systems demonstrated used single-core fibres and as such were not useful for proximal-scanning endoscopy.


We made outstanding progress towards the objectives up to the point at which we needed to scale the fibre design to a large number of cores. Despite our being aware from early in the project that this was a potential barrier, we were unable to find a solution which would have enabled us to incorporate the strengths of the earlier progress into a fibre with a large number of cores. The only way we could see to mitigate this risk would have been to sacrifice the several and very real advantages which we had developed and demonstrated during the project, and to perform the very incremental task of fabricating a conventional-fibre-technology multicore array. Such an array could possibly be engineered so as to give a modest improvement over the rather poor performance of currently-available fibres for CARS endoscopy, but could not possibly have offered the orders-of-magnitude advantage which we have demonstrated using our single-core and few-core fibres. Such incremental work would most likely not lead to a commercial advantage and would have been outside of the spirit of the CARS EXPLORER project

Future work should include areas in both conventional and unconventional fibres, but concepts proposing the use of air-hole structures in a highly multi-core configuration should be considered very carefully before proceeding. An exception to that may be the incorporation of a single hollow waveguide core into a conventional imaging bundle, enabling the possibility of performing nonlinear spectral analysis on a single spot whilst simultaneously imaging the greater area. Conventional fibre bundles using either silica / doped silica or silicate glasses remain an area of research which promises advances. Recent interest from the telecommunications industry in multi-core fibres is leading to highly specified fibre bundles for that application - typically with just a few (e.g. seven) cores, but the technology will lend itself to applications in imaging bundles.

Furthermore, we have demonstrated the design and fabrication of low-noise normal dispersion supercontinuum photonic crystal fibres, which have been successfully implemented in ultrabroadband fs-supercontinuum hyperspectral CARS imaging with improved performance

WP4 - Signal recovery and imaging processing

WP4 coordinator: François Lacombe (MKT)
Partners involved: Didier Marguet (Inserm 1A), Hervé Rigneault (CNRS), Nicolas van Baren (ICP)

General objectives

- extract useful information from single point NLO;
- extract instrumental data which enable further relevant signal extraction;
- develop specific processing of NLO microscopy image;
- infer the original signal coming from single fibre;
- build image reconstruction from a multi-fiber endoscope.

Progress towards objectives

At the end of the previous reporting period, we were mentioning the difficulty for building a relevant radiometric model of NLO microscopy signal through fibre optics, and consequently the difficulty of developing a software capable of extracting relevant NLO signatures from a raw signal. We then proposed an alternative strategy, consisting in providing our partner with a software package which should permit an easy implementation of test algorithms, to validate them directly on the prototype multichannel imager to be delivered during this period. We delivered the multichannel imager in February 2010, with the software which makes it possible to control this imager and to acquire NLO images through fibre bundles.

To solve the problem of system calibration with NLO endoscopes, we developed of a new method for calibrating bundle based endomicroscopes without the need of any calibration fluorescent reference solution. As opposite to standard techniques, with include several reference measurements performed before the experimental procedure, this new technique takes advantage of signal fluctuations, and more specifically of their statistical correlations between adjacent fibres in the bundle, to evaluate the proper transmission and background of each fiber.

This novel approach offers many advantages:

- It does not require any specific calibration medium, or solution, which, in the case of CARS emission would be very difficult to fabricate, particularly if one thinks of using the system at variable wavelength.
- It takes implicitly into account light propagation effects within the medium, which may affect the system actual PSF, and which may be poorly reproduced in a given calibration solution.
- It can be performed in real time, or offline, after the data are acquired.
- It can be continuously updated, during the acquisition procedure, meaning it is adaptive, and automatically compensates for tissue characteristics changes.

The technique is applicable to any situation where a multichannel system suffers from relative heterogeneities in sensitivity and /or offset. A provisional US patent application is currently being prepared and should be filed by MKT in the next few weeks.

This new approach permits using MKT standard image acquisition and reconstruction algorithms with any king of data, obtained with NLO of not.

This technique was intensively tested and validated in a standard fluorescence configuration by MKT. It is now being validated at Fresnel Institute with CARS data, first offline, at later in real time.

The algorithms were ready in October 2010 (month 31), and were patented by MKT in May 2011 under US 61/486551 (month 38).


The original plan was to develop specific softwares for NLO images processing, in the context of fibre based endomicroscopy. We changed this plan to overcome the intrinsic difficulty of establishing a convenient radiative model that we initially had seen as a pre-requisite of such a development. This led us to develop a completely new approach. This change allowed us not only to solve the initial problem but to extend our knowledge and our expertise in fibre bundle image processing. This in particular led to a patent filing.

The innovative solution developed for D4.2 in the course of this work-package has allowed MKT to increase its expertise in the field of fibre bundle image processing and to improve its existing products. This solution was shown to be a much more powerful tool, since it applies to many more imaging techniques, provided that they make use of multichannel detection, or, to be more specific, fibre bundles.

The efficiency of this new technique does not overcome all the limitations of some fibre bundles, in particular the level of background signal.

WP5 - Molecular morphological signature associated with tumour development

WP5 coordinator: Nicolas van Baren (ICP)

Partners involved: Didier Marguet (Inserm 1A), Anne-Marie Schmitt-Verhulst (Partner 1A), Hervé Rigneault (CNRS), Andreas Volkmer (USTUTT)

General objectives

The CARS EXPLORER final goal is the development of a diagnostic method for the detection of early modifications in the skin and in sentinel lymph nodes as a signature for metastatic melanoma. WP5 has explored the signatures of normal and cancer cells by Raman and CARS spectroscopy. This has been started in parallel to the establishment of the basic characteristics of NLO on skin and lymphoid tissue in WP1.

During the first period, the focus has been laid on setting up the material and procedures to permit a systematic analysis by NLO imaging of tumour samples, tumour cells in culture, and their non-cancerous counterparts, and on developing the mouse model to allow NLO analysis of early steps of tumourigenesis. These initial objectives have been achieved. A large tissue, tumour and cell line bank has been maintained. The procedures to prepare, stain, store and ship microscopic material have been optimised. The recombinant mouse strain with inducible, fluorescent melanoma has been established.

Next, the main objective was to perform systematic NLO analysis of biological material, and provide a proof of concept that these imaging techniques can help identifying tumour cells in melanoma tissue samples, and can help distinguishing them from neighbouring non-cancerous cells. A secondary objective was to determine, using these NLO techniques, whether mitotic cells could be differentiated from non-mitotic cells, as the former are frequently

Progress towards objectives

Bank of human samples

We have established and maintained, throughout the whole project, a biobank of frozen human and mouse tissues and cell lines that have allowed us to develop the NLO imaging techniques and apply them to the study of melanoma in situ.

At the end of the project, the bank comprises approximately 1450 melanoma metastases, 150 biopsies of primary melanomas, 26 benign naevi, and 15 biopsies of normal skin.

In addition, we have established a number of melanoma cell lines, immortalised in cultures starting from fresh melanoma samples. Most of these cell lines were already available at the start of the project, and new ones have been established since then. Recently, we have also obtained a primary melanocyte cell culture (i.e. with non-cancerous pigment cells), which is intended at allowing comparative analyses with their malignant counterparts.

Develop an inducible melanoma with a fluorescent read out

A mouse model has been established in which tumour suppressor genes are deleted and the oncogene H-rasG12V is expressed upon induction of the CRE recombinase. This leads to progressive melanoma development in the skin and their migration to the sentinel lymph node (Huijbers et al., 2006., Cancer Res. 66: 3278). The tumour microenvironment has been further characterised in this model (Soudja et al., 2010. Cancer Res. 70: 3515; Grange et al., 2011. Cancer Res. in press).

Furthermore, the ability to follow the early steps of melanoma development has been validated during the second part of the project. To this end, an inducible melanoma with a fluorescent read out has been generated in the mouse. The appropriate mice crosses have been established in order to introduce the ROSA-tdRFPlox/lox reporter gene in the previously characterised TiRP/INK4Alox/lox 'inducible melanoma' mouse strain (Huijbers et al., 2006. Cancer Res. 66: 3278). After 5 sequential crosses, we established a breeding that should provide 50 % of mice in which melanoma will develop with an incidence of about 40 % upon treatment with tamoxifen. We validated the possibility to monitor melanoma development by biofluorescence measurements on those mice

Analysis of human and mouse melanoma samples by Raman microscopy

Selected areas on frozen tissue sections were scanned with a Raman microscope, and the spectral measurements were recorded. Two dimensional images were reconstructed using either of two different approaches, and compared with an adjacent tissue section immunostained for melanoma cells and counterstained with hematoxylin. The first approach used K-means clustering to distinguish a defined number of areas in the tissue section. These areas were roughly similar to tissue areas, including tumour tissue, that were distinguishable on conventional histological examination. The second approach defined a so-called tumour index, calculated from the Raman intensity obtained at wavenumbers corresponding to C-C and P-O molecular bonds abundant in collagen and DNA, respectively. Images were reconstituted as heat maps. Hot areas showed, both in human and mouse melanoma skin samples, a good correlation with tumoural areas in the samples. These encouraging results deserve to be extended to additional samples, and to the new stimulated Raman scattering (SRS) method.

Analysis of human and mouse melanoma samples by combined NLO techniques

Selected tissue samples were submitted to spontaneous Raman spectroscopy, hyperspectral CARS, SRS, sum frequency generation (SFG) and two photon excited fluorescence (TPEF). Complementary and contrasted images were obtained, with clear visualisation of epidermis, tumour area and collagen. Further refinement of this innovative multi-modality approach is warranted, particularly with the emerging SRS technique, which appears to have much less background than CARS.

Analysis of human melanoma samples by Raman microscopy

Hyperspectral Raman imaging in the fingerprint region was performed on selected regions of interest inside melanoma skin tissue sections of 10 µm thickness. The spatial heterogeneities and relative concentrations of nucleic acids, proteins and lipids are clearly visualised. Because of the sub-cellular spatial resolution of these Raman maps, we are now able to directly compare their cell morphological information content with that obtained by conventional tissue labelling.

To compare the above label-free Raman images with a well-established reference method, samples have been post-stained with hematoxylin. Since hematoxylin counterstains all nuclei in blue, a correlation of hematoxylin contrast in the post-stained image with the Raman image contrast for nucleic acids is expected. This concept was experimentally demonstrated in both skin cell cultures and tissue sections.

These analyses have to be refined to determine whether mitotic cells, which contain dense chromatin, can be distinguished from resting cells. The move to SRS, by accelerating data acquisition, is expected to allow imaging of much larger areas of tissue sections at high resolution.


The main objective of this WP to provide proof-of-principle that NLO techniques can provide useful imaging information on tissue samples, has been achieved. Tissue image reconstitution without chemical or fluorescence staining was successful with several approaches. In the samples tested, tumoural and non-tumoural areas could be distinguished from each other in the same preparations. Importantly, complementary visual information were provided by different approaches. The results obtained are preliminary, but encouraging, and should be followed up. It is necessary at this stage to analyse a larger number of samples, to diversify the type of samples to test, to implement the much innovative SRS technique, and to optimise the combination of NLO techniques so as to retrieve the highest level of imaging information. NLO imaging of tissues is only at its beginning phase. Our observations are a strong encouragement to continue their development.

Potential impact:

Although the CARS EXPLORER project was challenging, our interdisciplinary team including physicists, biologists and clinicians has provided significant outcomes in the following fields:

- Fibre optics

Photonic crystal fibres (PCF) have been designed and manufactured with specific requirements to deliver and collect ultra-short pulses suitable for NLO endoscopy observations (fibres with flat dispersion curve for proper pulse propagation; first realisation of a double clad hollow core fibre for CARS applications, demonstrations of both multi-hollow-core fibres and polarisation-maintaining multi-solid-core fibres). We have also demonstrated low-noise supercontinuum fibers suitable for use in CARS microscopy.

- Nonlinear microscopy

Our project has raised nonlinear contrast (TPEF, SHG and CARS) to a new state of the art by setting up point scanning microscopes (using compact Ti:Sapph lasers and optical parametric oscillators) and developing new software to fully control the systems and process the data.

Furthermore, seminal work demonstrating a novel image contrast mechanism (SRS) in nonlinear microscopy has been carried out that was covered as a News Feature in Nature (Vol. 459 (4 June 2009), p. 636).

- Signal recovery and imaging processing

A prototype of CARS endo-microscope has been built for exploring the very specificities of in-vivo and in-situ CARS observations. In particular, original calibration techniques, which enable a rigorous evaluation of the instrumental and sample contributions, have been developed (and are being patented).

- Tools and methodologies

Different lines of researches have been explored extending the interest to pursue the CARS EXPLORER project:

(1) the ones based on light polarisation to provide key information at the molecular level;
(2) the elimination of non-resonant background in CARS mode;
(3) the hyper-spectral CARS imaging to reveal the vibrational signature of molecules in complex environments such as a biological tissue.

- Biological models and human cancer diagnostic

Human melanoma and mouse models have been chosen as biological samples to assess the value of NLO imaging techniques in cancer diagnostics. By developing a ratiometric method based on Raman images, we have been able to distinguish tumours from normal tissues and to specify the wavelengths of interest to generate informative CARS contrast. Currently, tumour areas present inside samples of cutaneous melanoma metastases can be distinguished from surrounding non-cancerous tissues in digital images. Furthermore, we have successfully demonstrated label-free mapping of nucleic acids, proteins and lipids in melanoma skin tissue sections with sub-cellular resolution based on their characteristic Raman signatures in the fingerprint region. We showed that the label-free Raman image contrast for nucleic acids directly correlates to the conventional hematoxylin-stained image contrast. In the meantime, animal models have been developed that should allow us to explore now the potential of such contrasting methods on living animals.

- Nonlinear endoscopy

The end-product of the CARS EXPLORER project was to deliver a prototypic endoscope system. Based on hollow core photonic crystal fibre, a CARS endoscope-like system has been achieved allowing us to record image contrasts generated either in CARS, two photons fluorescence (2PEF), second harmonic generation (SHG), or sum frequency generation (SFG), using a nonlinear crystal as a sample target.

In conclusion, the main results established through the CARS EXPLORER project have significant potential impact, at scientific and technological levels:

- By raising non-linear contrast to a new state-of-the-art by setting up NLO scanning microscopes, such as the first successful demonstration of SRS imaging at USTUTT, and providing the proof of principle of bioimaging multimodalities (WP1).
- By demonstrating for the first time a CARS and SRS endoscope scheme that permits to (1) deliver the pump and Stokes beams and (2) collect the CARS and SRS signals (WP2).
- By recording wide field background free CARS and phase retrieval in an original scheme that is compatible with living samples such as cells and tissues (WP2).
- By making outstanding progress towards the objectives of WP3 up to the point at which we needed to scale the fibre design to a large number of cores.
- By providing innovative solution in the field of fibre bundle image processing which significant improvement of existing products. The solution was shown to be a much more powerful tool, since it applies to many mode of imaging techniques, provided that they make use of multichannel detection, or, to be more specific, fibre bundles. The efficiency of this new technique does not overcome all the limitations of some fibre bundles, in particular the level of background signal.
- By reaching the proof-of-principle that NLO techniques can provide useful imaging information on tissue samples without chemical or fluorescence staining (WP1 and WP5). Moreover, tumoural and non-tumoural areas were efficiently distinguished from each other. And finally, the different contrasting approaches provide complementary visual information both at the chemical and morphological level.

Project website: