Fibered confocal spectroscopy and multicolor imaging system for in vivo fluorescence analysis

Transcription

1 Fibered confocal spectroscopy and multicolor imaging system for in vivo fluorescence analysis Florence Jean, Genevieve Bourg-Heckly Universite Pierre et Marie Curie, Laboratoire de Biophysique Moleculaire Cellulaire et Tissulaire (BioMoCeTi), CNRS UMR 7033, Genopole Campus 1, 5 Rue Henri Desbrueres, Evry Cedex, France g.bourg-heckly@ccr.jussieu.fr Bertrand Viellerobe Mauna Kea Technologies, 9 Rue d Enghien, Paris, France Abstract: We report the design and implementation of spectroscopic and multicolor imaging capabilities into a fibered confocal fluorescence microscope (FCFM) already capable of in vivo imaging. The real time imaging device and the high resolution fiber probe make this system the first reported capable of performing multi color detection in the field of FCFM. The advantages of the system will allow in vivo morphological and functional imaging. Preliminary experiments were carried out in tissue samples to demonstrate the potential of the technique. The quality of the axial sectioning achieved in the confocal fluorescence spectroscopy mode is demonstrated experimentally, and applications to multicolor imaging are shown Optical Society of America OCIS codes: ( ) Confocal microscopy; ( ) Scanning microscopy; ( ) Endoscopic imaging; ( ) Fluorescence microscopy; ( ) Spectroscopy, fluorescence and luminescence; ( ) Fiber optics imaging; ( ) Multiple imaging. References and links 1. E. Laemmel, M. Genet, G. Le Goualher, A. Perchant, J.F. Le Gargasson and E. Vicaut, Fibered Confocal Fluorescence Microscopy (Cell-viZio TM ) Facilitates Extended Imaging in the Field of Microcirculation: A Comparison with Intravital Microscopy, J. Vasc. Res. 41, (2004). 2. A.R. Rouse, A. Kano, J.A. Udovich, S.M. Kroto and A.F. Gmitro, Design and demonstration of a miniature catheter for a confocal microendoscope, Appl. Opt. 43, (2004). 3. K. Carlson, M. Chidley, K.B. Sung, M. Descour, A. Gillenwater, M. Follen and R. Richards-Kortum, In vivo fiber-optic confocal reflectance microscope with an injection-molded plastic miniature objective lens, Appl. Opt. 44, (2005). 4. P.M. Lane, A. Dlugan, R. Richards-Kortum and C.E. MacCaulay, Fiber-optic confocal microscopy using a spatial light modulator, Opt. Lett. 25, (2000). 5. T.D. Wang, M.J.Mandella, C.H. Contag and G.S. Kino, Dual-axis confocal microscope for high-resolution in vivo imaging, Opt. Lett. 28, (2003). 6. P.M. Delaney, M.R. Harris and R.G. King, Fibre-optic laser scanning confocal microscope suitable for fluorescence imaging, Appl. Opt. 33, (1994). 7. D.L. Dickensheets and G.S. Kino, Micromachined scanning confocal optical microscope, Opt. Lett. 21, (1996). (C) 2007 OSA 2 April 2007 / Vol. 15, No. 7 / OPTICS EXPRESS 4008

2 8. H. Miyajima, K. Murakami and M. Katashiro, MEMS Optical scanners for microscopes, IEEE J. Quantum Electron. 10, (2004). 9. K. Sokolov, J. Aaron, B. Hsu, D. Nida, A. Gillenwater, M. Follen, C. MacAulay, K. Adler-Storthz, B. Korgel, M. Descour, R. Pasqualini, W. Arap, W. Lam and R. Richards-Kortum, Optical systems for in vivo molecular imaging of cancer., Technol. Cancer Res. Treat. 2, (2003). 10. A. Perchant, G. Le Goualher, M. Genet, B. Viellerobe and F.Berier, An integrated fibered confocal microscopy system for in vivo and in situ fluorescence imaging - applications to endoscopy in small animal imaging, in Proceedings of the IEEE International Symposium on Biomedical Imaging: From Nano to Macro, A.R. Rouse and A.F. Gmitro, Multispectral imaging with a confocal microendoscope, Opt. Lett. 25, (2000). 12. G. Le Goualher, A. Perchant, M. Genet, C. Cave, B. Viellerobe, F. Berier, B. Abrat and N. Ayache, Towards optical biopsies with an integrated fibered confocal fluorescence microscope, Lecture Notes in Computer Science 3217(II): , Springer (Medical Image Computing and Computer Assisted Intervention), P. Vincent, I. Charvet, L. Bourgeais, L. Stoppini, N. Leresche, J-P. Changeux, R. Lambert, P. Meda, D. Paupardin- Tritsch, Live imaging of neural structure and function by fibered fluorescence microscopy, EMBO Reports 7, 11, , L. Thiberville, S. Moreno-Swirc, T. Vercauteren, E. Peltier, C. Cave, G. Bourg-Heckly, In vivo imaging of the bronchial wall microstructure using fibered confocal fluorescence microscopy, Am. J. Respir. Crit. Care Med. 1, 175, 22-31, Introduction Laser scanning fiber-optic confocal microscopy provides the ability to perform non or minimally invasive subsurface in vivo and in situ imaging at cellular level. Different technical approaches, based on fiber bundles [1, 2, 3, 4, 10] or single fibers [5, 6, 7, 8] have been developed by various groups [9]. Confocal imaging can be reflectance-based [3, 7, 8, 9], using backscattered light from within the tissue, or fluorescence-based, using fluorescent light generated by chemical probes labeling specific tissue microstructures [1, 2, 4, 5, 6, 9, 10, 11]. The latter technique has emerged as the leading modality in the field of biological imaging, due to the high sensitivity and specificity of fluorescence probes. The availability of a growing number of target-specific dyes and fluorescence proteins spanning the visible and near infrared spectral range has led to a new generation of benchtop microscopes featuring multiple excitation wavelengths and detection channels, capable of measuring simultaneously the distribution of several fluorescent probes, such enabling multilabeling studies. Besides, in addition to imaging the integrated fluorescence emission in the detection channel spectral windows, the knowledge of the complete emission spectrum from a localized region brings additional capabilities; spectral information is relevant for detecting changes in the microenvironment associated to spectral shifts, helping unmix signals from overlapping fluorophores, assessing the contribution of tissular autofluorescence... Applicative results can be found in the ref [14] where direct identification of the autofluorescence components of the bronchial mucosa was made possible by the combination of imaging and spectroscopy in a single color mode. Then, the incorporation of multicolor imaging and spectroscopic analysis capabilities in fibered confocal fluorescence microscopy systems would considerably extend the potentialities of in vivo, in situ imaging. To our knowledge, all current fibered systems are single wavelength excitation devices, only one of them exhibiting multispectral imaging and 2 excitation lasers [12], but with a slit-scanning solution and only one laser used at a time. The goal of this study was to demonstrate the concept of such a point scanning system, featuring two excitation wavelengths and two independent detection channels, with an additional spectroscopic channel, capable of real time operation. In order to allow the use of a large number of fluorophores, two prototypes were developed, each featuring two excitation wavelengths: 405 nm/488 nm and 488 nm/638 nm. The principle of operation is based on the fibered confocal fluorescence microscopy technology developed by Mauna Kea Technologies (Paris, France) which is already implemented in several commercialized instruments with single wavelength illumination mode at 488 nm or 660 nm for small (C) 2007 OSA 2 April 2007 / Vol. 15, No. 7 / OPTICS EXPRESS 4009

3 APD 100µm slit Pinhole Spectrometer Optical fiber Dichroic Filter D3 APD Fiber bundle Beamsplitter Rejection Filters R1,R2 Injection System LIM Optical head Laser λ2 2 Dichroic Filter D2 Sample Laser λ 1 Dichroic Filter D1 Afocal G=1 Depth of visualisation ( µm) Resonant Mirror M1 Galvanometric Mirror M2 Afocal G=1 Fig. 1. Lay-out of the fibered confocal fluorescence spectroscopy and multicolor imaging system. animal imaging [13] and for microendoscopy in the gastrointestinal tract and in the lungs [14]. 2. System description The scheme of the experimental set-up is shown in figure 1. A solid-state laser diode at 488 nm (Laser Sapphire from Coherent) and a second laser at 638 nm (Laser Cube from Coherent), are scanned by two mirrors on the proximal face of a fiber bundle. Horizontal line scanning is performed using a 4 khz oscillating mirror M1 (GSI Lumonics) while a galvanometric mirror M2 (GSI Lumonics) performs frame scanning. The resulting frame rate, limited by the number of scanned lines, reaches 12 images/s for a pixels image size. The afocal systems (G = 1) ensure the conjugation of the scanning mirrors with the entrance pupil plane of the injection system. One of the main challenges was to obtain a high coupling efficiency of the laser beam with the micron-sized optical fibers. For that purpose, a specific opto-mechanical connector was designed to connect the fiber bundle to the laser scanning unit. The connector gives a high repeatability in the focus position, as well as the optimal injection, along the whole image field. The fiber bundle is blocked in rotation and translation within a few microns. Finally, the optical beam was optimized to ensure that only one fiber is injected at a time, each fiber sequentially acting as a point source and a point detector. Thus, the system is working in a flying spot mode (point to point reconstructed image). The light injection module (LIM) is a custom made optical system composed of several doublets and triplets of lenses acting as a microscope objective. It was specifically designed to work on a large field of view (FOV typically 600 µm) with minimized aberrations (WFE < λ/5), particularly in terms of spherical (< 2 µm) and chromatic aberrations (< 9 µm). Thus, this optical system is compatible with a (C) 2007 OSA 2 April 2007 / Vol. 15, No. 7 / OPTICS EXPRESS 4010

4 large range of excitation wavelengths, from 400 nm to 700 nm. Its numerical aperture NA = 0.5 and point spread function (PSF) were calculated to get the best injection rate into the fibers, around 70 %, the PSF fitting at best the diameter of the fundamental mode of the optical fibers (around 1.8 µm). Effectively, the best coupling efficiency between the excitation laser beam and the fibers, at proximal and distal ends, is obtained when the NA of the beam approximately matches the NA of the best-excited propagation mode in the fiber (0.22 for the TEM00 mode). Typically, the PSF is ranging from 1.2 µm to 1.5 µm depending on the wavelength. The microprobes are composed of a connector (mentioned above), a fiber bundle and an optical head. The fiber bundle (Fujikura, FIGH S) is made of 30,000 fiber cores. The lateral resolution of the microprobe is limited by the fiber bundle inter-core distance of 3.3 µm and the fiber core diameter of 1.9 µm. At the distal end of the fiber bundle, the light beam is focused in a plane located at a given depth in the biological tissue. Two types of fiber bundle probes were developed: Probe I: miniaturized (overall diameter 1.4 mm), the fibers being directly in contact with the tissue. This probe provides images of a 15 µm thick tissue layer located immediately below the surface, with a lateral resolution of 3.5 µm [13]. The field of view is adjustable from several hundred microns up to the useful entire diameter (600 µm). The scanned surface is adjustable, with a nominal size of 280 µm 400 µm. In general, the field of view is equal to the scanned surface divided by the magnification of the distal optics. Then, in the case of the Probe I, the field of view is simply equal to the scanned surface. Probe II: not miniaturized yet, based on an achromatic optical head (figure 2) composed of a 25 objective O1 (Melles Griot, NA = 0.32, f = 160mm), a field lens (Melles Griot, F = 80mm) and a 100 objective O2 (Melles Griot, NA = 1.3, f = 160mm, oil immersion, 0.17 cover slip). The resulting magnification (tissue to fiber) is 4. Fluorescence light collected by the distal optics propagates through the bundle using all available modes. The best NA for the optics which collects this light must thus be larger than the geometric NA of the fibers (0.42) on the proximal side. On the other hand, the geometric NA around 0.32 on the distal side clearly insures a maximum coupling efficiency for the fluorescence coming back from the tissue. This probe provides images with a field of view of 80 µm 80 µm of a 2 µm thick tissue layer. The depth is adjustable from the surface of the tissue down to approximately 100 µm. This z scanning capability is realized by translating the specimen. This feature is not included inside the distal head, meaning that adjustable depth of observation will not be implemented in vivo. Fiber bundle Tissue O1 (x25) O2 (x100) Field Lens Fig. 2. Scheme of probe II. A miniaturized version of the probe II is currently being assembled and tested (overall diameter = 4.2mm). The standard optical design of this probe is based on 4 lenses, one doublet and (C) 2007 OSA 2 April 2007 / Vol. 15, No. 7 / OPTICS EXPRESS 4011

5 Table 1. Specifications of the probe II s miniaturized version. Specification Value Magnification 2.5 Numerical aperture 0.8 (tissue) Transmission 95% Field of view (FOV) 240 µm Lateral resolution (LR) 1.3 µm Axial resolution (AR) 3-5 µm Axial chromatism 2 µm ( nm) Lateral chromatism < 0.5 µm ( nm) Overall diameter 4.2mm Total length 26mm Fig. 3. Typical images obtained in vivo with the miniaturized version of probe II. Healthy colonic crypts stained with fluorescein can be seen. FOV is 240 µm. The working distance is 30 µm. (C) 2007 OSA 2 April 2007 / Vol. 15, No. 7 / OPTICS EXPRESS 4012

6 3 singlets. One lens is divergent in order to lower the chromatic aberrations. The high NA is obtained via one half-ball lens at the distal end of the micro-objective. The main specifications of this fibered microprobe are listed in table 1. Image quality is diffraction limited within a wide spectral range from 488 nm to 700 nm. Several working distances ranging from 20 µm and 100 µm can be set depending on the application. Early in vivo imaging results have been obtained, demonstrating the confocal capability of this miniaturized probe, as shown in figure 3. The detection optical path is divided into two channels by a beamsplitter enabling simultaneous imaging and spectroscopy: 80 % of the fluorescent signal being used for imaging and 20 % for spectral analysis. The imaging signal is detected by two avalanche photodiodes (Hamamatsu), each one being associated with a pinhole, in two distinct spectral ranges (λ 1 < λ < λ 2 and λ 2 < λ < 800nm) via a dichroic filter D3 (see table 2). The pinhole diameter, 20 µm, was calculated to reject the fluorescent light coming from adjacent fibers (encircling the excitation fiber). The detected optical signal consists of a complex mixing of multiple radiations that can be splitted into two majors contributions: the fluorescence signal emitted by the tissue and the background signal generated by the interaction of the laser with the optical fibers of the bundle (Rayleigh scattering, Raman scattering and Autofluorescence). Their levels are depending on the laser power injected into the fibers of the bundle. Typically, the incident power on the specimen is in the range of 3-6 mw depending on the fiber probe type. Table 2. Dichroic and rejecting filters properties in the case of 488/635 nm architecture. Type of filter Cut-off wavelength Bandwidth D1 High-pass Dichroic 500 nm 250 nm D2 Dual-band Dichroic 500 nm 110 nm 645 nm 105 nm D3 High-pass Dichroic 645 nm 105 nm R1 High-pass rejection 500 nm 250 nm R2 Band-pass rejection 625 nm 20 nm A customized hardware is responsible for mirrors control and synchronous signal digitization. Images are rebuilt starting from this signal by an in-house built software described in the MICCAI conference paper [12] which is a reference for medical image computing. While scanning the fiber bundle, the input light is modulated in amplitude by the injection rate in each individual fiber of the bundle. This modulation creates a honeycomb effect on the raw images. The task of the processing module is therefore to restore the images by removing the fiber bundle modulation, the scanning distortions and any unwanted signal such as the fiber autofluorescence. The spectroscopic channel consists of an optical fiber (100 µm core) coupled to a grating spectrometer (USB2000, Ocean Optics) with an entrance slit width of 100 µm achieving a spectral resolution better than 4 nm. The spectrometer is controlled through the Spectra Suite software (Ocean Optics). After the beamsplitter, the fluorescence signal is focused into the optical fiber which serves as a confocal limiting aperture. On one hand, the axial resolution is obviously much better for the imaging channel compared to the spectroscopic one, due to the smaller pinhole size. On the other hand, the lateral resolution of the spectroscopic channel depending on different parameters (pinhole size, scanned area, integration time of the spectrometer) cannot be compared directly to the imaging channel. The fluorescence spectrum is recorded between 500 nm and 750 nm. As excitation light generates autofluorescence and Raman scattering signal within the fiber, a spectrum of this background was acquired prior to each experiment and subtracted from the tissular emission spectrum. Then the spectrum is corrected for the spectral (C) 2007 OSA 2 April 2007 / Vol. 15, No. 7 / OPTICS EXPRESS 4013

7 sensitivity of the spectrometer detector and optical setup. The spectrometer integration time is set to 80 ms, identical to the image scanning time, the spectra acquisition being synchronized with the frame trigger. (a) (b) (c) (d) (e) Fig. 4. Images and spectra of a fixed human cervix sample stained with a 100 µm solution of DiA (Invitrogen) (a,d) and with a 100 µm solution of POPO-1 (Invitrogen) excited respectively at 488 nm and 405 nm (c,e). (b) is the fusion of these two images. (Probe II, FOV: 80 µm 100 µm). Signal losses visible on the spectra are due to filters spectral response in the case of 405/488nm architecture system. For each detection channel, image and spectrum are simultaneously acquired and displayed on the computer monitor in real-time. Then images are superimposed to get a multi-colored fusion image (figure 4). False colors were selected to provide the best contrast. 3. Results Both channels were validated separately and measurements of the axial and lateral resolutions at both wavelengths were performed using 2 µm-diameter fluorescent beads (Estapor, FXC 200). A drop of solution of those fluorescent beads was deposited on a plate, diluted in ethanol and evaporated. The maximum of emission fluorescence intensity was measured by bidirectional micrometric translation. Results are given in table 3 and figure 5. Table 3. Lateral and axial resolution (LR, AR) measured at 488 and 638 nm, with probes I and II, for the spectroscopic and imaging channels (micrometers). 488 nm 638 nm AR LR AR LR Probe Imaging channel I Spectroscopic channel Probe Imaging channel < 2 < 1 < 2 < 1 II Spectroscopic channel < (C) 2007 OSA 2 April 2007 / Vol. 15, No. 7 / OPTICS EXPRESS 4014

8 Micrometric translation Fig. 5. Axial intensity profiles of a fluorescent bead showing the resolution of the spectroscopic channel for Probe I and Probe II at 488 nm. Axial resolution is determined by the FWHM of the curves. The spectral interval between the two excitation wavelengths has been chosen to be wide enough to prevent the respective fluorophores absorption and emission spectra from overlapping and the corresponding imaging channels from perturbating each other, the illumination being simultaneous or sequential. The quality of axial and lateral achromatism was demonstrated by a negligible shift of the images allowing easy superimposition. Images and confocal spectra from various samples of fixed human tissue (oesophagus, cervix, colon, thyroid), stained with different fluorophores are shown in figures 4, 6 and 7. The results obtained in the figures 4, 6 and 7 clearly indicate the cellular imaging capability of the system, giving the possibility to a direct comparison with standard histology. For example, images from figure 4 show cell nuclei (in blue) and cellular/nuclear membranes (in red). The regular nuclei sizes and shapes seem to be compatible with an healthy tissue, histological analysis giving the same conclusion. Furthermore, images a and b from figure 6 exhibit specific cellular architecture, squamous epithelium for the first one (cervix tissue) and glandular epithelium for the second one (thyroid tissue), in agreement with standard histological patterns. In the same way, parakeratosis pathology, characterized by high nuclei density with small sizes and elongated shapes, is illustrated in the images c, d and e from figure 6, histology giving the same diagnosis. The high signal to noise ratio (SNR > 30dB) obtained ex vivo clearly opens new in vivo applications which requires spectral analysis such as Forster Resonance Energy Transfer (FRET). The spectral analysis will be more relevant in the future by adding Region Of Interest illumination capabilities. In this case, spectral informations as changes in the microenvironment, dyes colocalization, contribution of tissular autofluorescence will definitely be helpful. This present work presents only early results and proof of concept of spectral acquisition capability within a multicolor system. Recent in vivo results associating spectral information within a single color system were demonstrated by Thiberville et Al [14]. (C) 2007 OSA 2 April 2007 / Vol. 15, No. 7 / OPTICS EXPRESS 4015

9 (a) (b) (c) (d) (e) Fig. 6. Multilabeled images from fixed squamous and glandular human tissues with probe II. Cervix (a) and thyroid (b) samples are stained with 100 µm solutions of DiA (red) and To-Pro-1 (Invitrogen) (blue), respectively excited at 488 nm and 638 nm. Multilabeled images (c,d,e) from fixed human cervix samples exhibiting different parakeratosis grades, confirmed by the histological analysis (FOV: 80 µm 100 µm). (a) (b) Fig. 7. Multilabeled images with probe I from fixed human cervix samples stained with DiA (blue) excited at 488 nm and To-Pro-1 (red) excited at 638 nm (FOV: 430 µm 300 µm). The density of cell nuclei (in red) indicates a high-grade parakeratosis. The lower resolution of the Probe I cannot give a detailed of view of the cellular membranes (in blue). The squamous epithelium pattern is clearly visible. (C) 2007 OSA 2 April 2007 / Vol. 15, No. 7 / OPTICS EXPRESS 4016

10 4. Conclusion This study demonstrates the feasibility of a point scanning FCFM system with spectroscopic and muticolor imaging capabilities. Preliminary experiments clearly demonstrate the ability of the system to perform real-time imaging. Miniprobes, such as probe I and miniaturized version of probe II, can be used directly or inserted in the working channel of an endoscope, allowing in vivo, in situ morphological and functional imaging. This should be of particular interest for small animal imaging, where the wide range of available dyes allows a large number of combinations for multiple fluorescence staining. Due to toxicity issues, the number of dyes approved for human use is obviously more limited; however fluorescent probes such as fluorescein, cresyl violet or indocyanine green are of common use in clinical practice and in vivo single color mode microendoscopy is a fast growing technique. The development of new fluorescent contrast agents is a rapidly progressing field which should also benefit to molecular imaging in human beings. In conclusion, the development of such a FCFM system, featuring two excitation wavelengths with spectroscopic analysis capability, paves the way to multilabeling studies in live animals and clinical trials as well. Acknowledgments We thank A. Perchant from Mauna Kea Technologies for image processing. Human tissues were provided through a collaboration with Dr E. Peltier. The support of OSEO-ANVAR (n A Q) is gratefully acknowledged. (C) 2007 OSA 2 April 2007 / Vol. 15, No. 7 / OPTICS EXPRESS 4017