Multispectral imaging with a confocal ... - Semantic Scholar
1708
OPTICS LETTERS / Vol. 25, No. 23 / December 1, 2000
Multispectral imaging with a confocal microendoscope Andrew R. Rouse and Arthur F. Gmitro Department of Radiology and the Optical Sciences Center, University of Arizona, Tucson, Arizona 85724 Received May 16, 2000 The concept of a multispectral confocal microscope for in vivo imaging is introduced. To demonstrate the concept we modified a slit-scan f luorescence confocal microendoscope incorporating a fiber-optic catheter for in vivo imaging to record multispectral images. The system was designed to examine cellular structures during optical biopsy and to exploit the diagnostic information contained within the spectral domain. Preliminary experiments were carried out in phantoms and cell cultures to demonstrate the potential of the technique. © 2000 Optical Society of America OCIS codes: 170.1790, 170.2150, 170.3880.
Benchtop confocal microscopes are widely used to create high-quality optical images of biological samples. The fundamental characteristic of the confocal microscope is the ability to reject light from out-of-focus planes and provide a clear in-focus image of a thin section within the sample. This optical sectioning property is what makes the confocal microscope ideal for imaging thick biological samples. Confocal microscopes have been adapted for in vivo imaging of skin,1 – 3 cornea,4 teeth,5 and cervix.6 For imaging deeper within the body, specialized endoscopic systems based on vertical-cavity surface-emitting laser arrays,7 micromachine scan mirrors,8 single f ibers with dispersive elements,9 and coherent optical fiber bundles have been developed.10 – 12 Confocal microscopes typically operate as epiillumination systems in f luorescence or ref lectance. However, f luorescence has emerged as the primary imaging technique because of the sensitivity and targeting specif icity of f luorescent probes. The emission characteristics of many f luorescent probes used in microscopy are affected by the local environment in the sample. Therefore, multispectral f luorescence imaging provides the ability to determine properties of the local environment in a spatially resolved manner. Confocal microspectrof luorometers have been developed to measure spatially resolved f luorescence spectra13 and have been used to study drug– target interactions,13 ion concentrations,14 and pH levels15 in living cells. However, these devices are essentially benchtop systems with slow mechanical scanning and offer no solution for evaluating remote in vivo locations. In this Letter we present and demonstrate the concept of a multispectral confocal microendoscope (MCME) for remote in vivo imaging. The MCME combines the features of a catheterbased confocal microscope with those of a microspectrof luorometer. It has three main components: the illumination system, the f iber-optic catheter, and the detection system. The illumination consists of a light source, which is focused and scanned across the proximal end of a f iber-optic imaging bundle. The f iberoptic catheter transfers the scanned illumination prof ile to the distal, in vivo end of the catheter. A miniature objective lens images the distal end of the fiber bundle into the tissue, and a focusing mechanism controls the depth of the image plane. Induced 0146-9592/00/231708-03$15.00/0
sample f luorescence is imaged back through the catheter to the detection system and descanned onto the confocal aperture. Light transmitted by the aperture is dispersed (by a prism or a grating) and imaged onto an array detector. As the system scans, a three-dimensional data set containing two spatial dimensions and one spectral dimension is obtained. A four-dimensional data set (three-dimensional spatial and one-dimensional spectral) may be acquired by collection of information from successive depths. To demonstrate the concept of the MCME we modified a slit-scan confocal system developed previously.12 The system was configured to collect spectral information across a 286-nm wavelength range centered at 600 nm with a spectral resolution of approximately 11 nm. However, the design is f lexible and can be adapted to vary these spectral parameters. Figure 1 shows a layout of the MCME. An argon-ion laser operating at 488 nm and an anamorphic optical system produce a line of light that is scanned in one dimension across the proximal end of the f iber-optic catheter. The catheter consists a fiber-optic imaging bundle, a miniature objective, and a hydraulic focusing mechanism. The fiber bundle contains 30,000 optical elements with 3-mm center-to-center spacing. The overall diameter of the fiber is 1 mm, with an active image diameter of 720 mm. The miniature F 兾1 achromatic objective has a nominal magnification of 1.67 from tissue to fiber. A hydraulic focusing mechanism moves the distal face of the fiber with respect to the objective, which is in contact with the tissue. This mechanism allows focus control to 200 mm below the surface of the tissue.
Fig. 1. Layout of the MCME. © 2000 Optical Society of America
December 1, 2000 / Vol. 25, No. 23 / OPTICS LETTERS
We modified the detection arm of the microendoscope to collect multispectral data by placing a collimating lens and a dispersing prism after the confocal slit. The dispersed light is imaged onto a cooled 512 3 512 CCD. At a fixed position of the scan mirror the two-dimensional light distribution on the CCD represents one (vertical) dimension of spatial information and one (horizontal) dimension of spectral information. A zoom lens mounted on the camera is adjusted such that the spatial extent of the image covers 256 pixels in the vertical dimension. At this magnification, the image of the slit aperture on the CCD is approximately 1 pixel wide. An 18± wedge prism is used to produce a smear in the spectral dimension of 26 pixels over a 286-nm range centered at 600 nm. The 256 3 26 pixel region of interest is read out of the CCD in 16 ms using a 1-MHz 12-bit digitizer. The scan rate of the mirror is adjusted so that the illumination moves one spatial resolution element during the 16-ms integration period. As the mirror is scanned, 256 frames of data are read out in 4.1 s to produce the full three-dimensional data set 共256 3 256 spatial 3 26 spectral兲. We carried out imaging experiments to validate the technique. Figure 2 shows images of a phantom made of a mixture of 6-mm green, 15-mm yellow, and 15-mm red f luorescent microspheres. The spectral dimension of the data set was weighted and summed using three weight vectors that were chosen empirically to sort the microspheres based on color. Figures 2a, 2b, and 2c show the resultant green, yellow, and red images, respectively. The spectrum of a representative microsphere from each group is plotted in Fig. 2d. We performed experiments with cell cultures to illustrate the potential of multispectral confocal imaging for examination of cell morphology. Acridine Orange (AO) was used as the f luorophore for these experiments. At low concentration, the f luorescence of this dye is green, with a peak emission at 520 nm. However, at high concentrations, the emission spectrum shifts to the red, with a peak around 650 nm. AO is a nucleic acid stain and is perferentially attracted to low-pH environments. The highly acidic lysosomes in the cytoplasm result in preferential accumulation of AO and, with the proper input concentration, yield green emission in the nucleus and red emission in the cytoplasm.15 Figure 3 shows images obtained from a sample of human lung cancer cells in culture. We processed the spectral information at each spatial location with two spectral weight vectors to isolate the nuclei from the rest of the cell structures based on the difference in f luorescent emission spectra. Figures 3a and 3b show the normalized intensity images of the green (nuclei) and red (cytoplasm) channels, respectively. The gray-scale image in Fig. 3c was created by summing of the data along the spectral dimension and is therefore similar to an image obtained with a standard confocal microendoscope. The color reproduction shown in Fig. 3d was created by combination of the two spectral images. Figure 4 shows an image of AO-stained rat sinusoidal endothelial cells, along with spectra from the nucleus and the cytoplasm of one of the cells.
1709
In terms of imaging performance, the spatial resolution of the MCME is essentially the same as that of the confocal microendoscope, since the components that determine the spatial resolution are unchanged. The lateral resolution of the MCME is fundamentally limited by the pixel spacing in the fiber bundle. The center-to-center spacing of the f iber pixels is 3 mm. The magnification of the distal objective is 1.67, yielding a theoretical lateral resolution limit of 1.8 mm in the tissue, which is the measured value within experimental error. The axial (or depth) resolution of the slit-scan confocal system, as measured by the response to a planar object scanned through focus, is 25 mm. The spectral resolution in the MCME is determined by the dispersing power of the prism and the imaging optics in the detection arm. For the results presented here, a spectral range of 286 nm centered at 600 nm was spread over 26 pixels on the CCD. Figure 5 shows a plot of the spectrum obtained with light from a He – Ne laser fed back into the system. The FWHM of the distribution is 1.0 CCD pixel, showing that the spectral resolution of the system is indeed 11 nm.
Fig. 2. Phantom of 6-mm green, 15-mm yellow, and 15-mm red f luorescent microspheres. The microspheres were spectrally sorted into a, green; b, yellow; and c, red channels. d, Spectra of a representative microsphere from each channel.
Fig. 3. Human lung cancer cells stained with AO. a, Green channel showing primarily nucleus; b, red channel showing primarily cytoplasm; c, gray-scale image; d, RGB color reconstruction.
1710
OPTICS LETTERS / Vol. 25, No. 23 / December 1, 2000
Fig. 4. Rat sinusoidal endothelial cells stained with AO. a, RGB color reconstruction; b, spectra from the nucleus and cytoplasm of the highlighted cell.
Fig. 5.
Measured spectral response of a He–Ne laser.
We obtained the fine sampling of this distribution by tilting the CCD camera and collapsing the intensity distribution along the direction of the line image. The spectral resolution in the system can be adjusted by a change in the prism wedge angle or by variation of the tilt angle of the prism. Alternatively, the system can be conf igured with a diffraction grating if significantly higher spectral resolution is desired. The spectral range is also easily varied by altering the number of pixels read out of the CCD along the spectral dimension. The frame rate of the system is currently limited by the readout speed of the CCD. A 256 3 256 3 26 multispectral image takes approximately 4 s to acquire. Higher frame rates can be achieved with a faster CCD. However, there is a trade-off among spectral sampling rate, data-acquisition speed, and signal-to-noise ratio. The data presented here represent a reasonable compromise for f luorescence imaging of cells in culture with AO as the f luorescent dye. However, specific applications may require coarse spectral sampling, which can be achieved with faster frame rates and (or) an improved signal-to-noise ratio. The optimum trade-off between these parameters will, of course, depend on the specif ic application. Clearly, there is information stored in the spectral domain. Multispectral imaging systems, which usually employ switched spectral filters, have already shown promise for in situ diagnosis. The MCME discussed here provides high spatial resolution, the depth discrimination of a confocal system, and the potential for fine and optically efficient sampling of the spectral domain. Exactly how much spectral information is needed and how it is used are areas of
continuing research. However, the exciting results being obtained with confocal microspectrof luorometers point the way toward applications that can now be extended to in vivo imaging of remote locations inside animals and humans. We believe that multispectral imaging can be used to improve the diagnostic utility of the in vivo confocal microendoscope. Color reconstructions appear to enhance image contrast and may be valuable for real-time in vivo visualization. Spectral data might be processed to yield functional information, such as pH maps, or to provide automated disease diagnosis. In conclusion, we have introduced the concept of a multispectral confocal microendoscope for in vivo imaging. To demonstrate the principle we modified a slit-scan confocal microendoscope incorporating a fiber-optic catheter to collect multispectral data. Preliminary experiments demonstrated the ability of the system to provide high-quality multispectral images of live cells. Demonstration of the clinical utility of the instrument will require further work. This research was supported by grants from the National Institutes of Health (CA73095) and the Arizona Disease Control Research Commission (9711). We thank Ben Kriederman for providing the cell cultures and Yash Sabharwal for help with the technical development of the instrumentation. A. F. Gmitro’s e-mail address is [email protected]. References 1. K. Prashanth and A. P. Dhawan, Comput. Med. Imag. Graph. 16, 153 (1992). 2. M. Rajadhyaksha, R. R. Anderson, and R. H. Webb, Appl. Opt. 38, 2105 (1999). 3. B. R. Masters, D. J. Aziz, A. F. Gmitro, J. H. Kerr, T. C. O’Grady, and L. Goldman, J. Biomed. Opt. 2, 437 (1997). 4. H. Ichijima, W. M. Petroll, J. V. Jester, P. A. Barry, P. M. Andrews, M. Dai, and H. D. Cavanagh, Cornea 12, 369 (1993). 5. T. F. Watson, Scanning 16, 168 (1994). 6. T. Collier, P. Shen, B. de Pradier, K. Sung, A. Malpica, M. Follen, and R. Richards-Kortum, Opt. Express 6, 40 (2000), http://epubs.osa.org/opticsexpress. 7. R. H. Webb and F. J. Rogomentich, Opt. Lett. 20, 533 (1995). 8. D. L. Dickensheets and G. S. Kino, Opt. Lett. 21, 764 (1996). 9. G. J. Tearney, R. H. Webb, and B. E. Bouma, Opt. Lett. 23, 1152 (1998). 10. A. F. Gmitro and D. Aziz, Opt. Lett. 18, 565 (1993). 11. R. Juskatitis, T. Wilson, and T. F. Watson, Scanning 19, 15 (1997). 12. Y. S. Sabharwal, A. R. Rouse, L. Donaldson, M. F. Hopkins, and A. F. Gmitro, Appl. Opt. 38, 7133 (1999). 13. J. M. Millot, S. Sharonov, and M. Manfait, Cytometry 17, 50 (1994). 14. S. Sebille, M. Pereira, J. M. Millot, J. Jacquot, A. M. Delabrosie, M. Arnaud, and M. Manfait, Biochem. Biophys. Res. Commun. 246, 111 (1998). 15. C. Millot, J. M. Millot, H. Morjani, A. Desplaces, and M. Manfait, J. Histochem. Cytochem. 45, 1255 (1997).
OPTICS LETTERS / Vol. 25, No. 23 / December 1, 2000
Multispectral imaging with a confocal microendoscope Andrew R. Rouse and Arthur F. Gmitro Department of Radiology and the Optical Sciences Center, University of Arizona, Tucson, Arizona 85724 Received May 16, 2000 The concept of a multispectral confocal microscope for in vivo imaging is introduced. To demonstrate the concept we modified a slit-scan f luorescence confocal microendoscope incorporating a fiber-optic catheter for in vivo imaging to record multispectral images. The system was designed to examine cellular structures during optical biopsy and to exploit the diagnostic information contained within the spectral domain. Preliminary experiments were carried out in phantoms and cell cultures to demonstrate the potential of the technique. © 2000 Optical Society of America OCIS codes: 170.1790, 170.2150, 170.3880.
Benchtop confocal microscopes are widely used to create high-quality optical images of biological samples. The fundamental characteristic of the confocal microscope is the ability to reject light from out-of-focus planes and provide a clear in-focus image of a thin section within the sample. This optical sectioning property is what makes the confocal microscope ideal for imaging thick biological samples. Confocal microscopes have been adapted for in vivo imaging of skin,1 – 3 cornea,4 teeth,5 and cervix.6 For imaging deeper within the body, specialized endoscopic systems based on vertical-cavity surface-emitting laser arrays,7 micromachine scan mirrors,8 single f ibers with dispersive elements,9 and coherent optical fiber bundles have been developed.10 – 12 Confocal microscopes typically operate as epiillumination systems in f luorescence or ref lectance. However, f luorescence has emerged as the primary imaging technique because of the sensitivity and targeting specif icity of f luorescent probes. The emission characteristics of many f luorescent probes used in microscopy are affected by the local environment in the sample. Therefore, multispectral f luorescence imaging provides the ability to determine properties of the local environment in a spatially resolved manner. Confocal microspectrof luorometers have been developed to measure spatially resolved f luorescence spectra13 and have been used to study drug– target interactions,13 ion concentrations,14 and pH levels15 in living cells. However, these devices are essentially benchtop systems with slow mechanical scanning and offer no solution for evaluating remote in vivo locations. In this Letter we present and demonstrate the concept of a multispectral confocal microendoscope (MCME) for remote in vivo imaging. The MCME combines the features of a catheterbased confocal microscope with those of a microspectrof luorometer. It has three main components: the illumination system, the f iber-optic catheter, and the detection system. The illumination consists of a light source, which is focused and scanned across the proximal end of a f iber-optic imaging bundle. The f iberoptic catheter transfers the scanned illumination prof ile to the distal, in vivo end of the catheter. A miniature objective lens images the distal end of the fiber bundle into the tissue, and a focusing mechanism controls the depth of the image plane. Induced 0146-9592/00/231708-03$15.00/0
sample f luorescence is imaged back through the catheter to the detection system and descanned onto the confocal aperture. Light transmitted by the aperture is dispersed (by a prism or a grating) and imaged onto an array detector. As the system scans, a three-dimensional data set containing two spatial dimensions and one spectral dimension is obtained. A four-dimensional data set (three-dimensional spatial and one-dimensional spectral) may be acquired by collection of information from successive depths. To demonstrate the concept of the MCME we modified a slit-scan confocal system developed previously.12 The system was configured to collect spectral information across a 286-nm wavelength range centered at 600 nm with a spectral resolution of approximately 11 nm. However, the design is f lexible and can be adapted to vary these spectral parameters. Figure 1 shows a layout of the MCME. An argon-ion laser operating at 488 nm and an anamorphic optical system produce a line of light that is scanned in one dimension across the proximal end of the f iber-optic catheter. The catheter consists a fiber-optic imaging bundle, a miniature objective, and a hydraulic focusing mechanism. The fiber bundle contains 30,000 optical elements with 3-mm center-to-center spacing. The overall diameter of the fiber is 1 mm, with an active image diameter of 720 mm. The miniature F 兾1 achromatic objective has a nominal magnification of 1.67 from tissue to fiber. A hydraulic focusing mechanism moves the distal face of the fiber with respect to the objective, which is in contact with the tissue. This mechanism allows focus control to 200 mm below the surface of the tissue.
Fig. 1. Layout of the MCME. © 2000 Optical Society of America
December 1, 2000 / Vol. 25, No. 23 / OPTICS LETTERS
We modified the detection arm of the microendoscope to collect multispectral data by placing a collimating lens and a dispersing prism after the confocal slit. The dispersed light is imaged onto a cooled 512 3 512 CCD. At a fixed position of the scan mirror the two-dimensional light distribution on the CCD represents one (vertical) dimension of spatial information and one (horizontal) dimension of spectral information. A zoom lens mounted on the camera is adjusted such that the spatial extent of the image covers 256 pixels in the vertical dimension. At this magnification, the image of the slit aperture on the CCD is approximately 1 pixel wide. An 18± wedge prism is used to produce a smear in the spectral dimension of 26 pixels over a 286-nm range centered at 600 nm. The 256 3 26 pixel region of interest is read out of the CCD in 16 ms using a 1-MHz 12-bit digitizer. The scan rate of the mirror is adjusted so that the illumination moves one spatial resolution element during the 16-ms integration period. As the mirror is scanned, 256 frames of data are read out in 4.1 s to produce the full three-dimensional data set 共256 3 256 spatial 3 26 spectral兲. We carried out imaging experiments to validate the technique. Figure 2 shows images of a phantom made of a mixture of 6-mm green, 15-mm yellow, and 15-mm red f luorescent microspheres. The spectral dimension of the data set was weighted and summed using three weight vectors that were chosen empirically to sort the microspheres based on color. Figures 2a, 2b, and 2c show the resultant green, yellow, and red images, respectively. The spectrum of a representative microsphere from each group is plotted in Fig. 2d. We performed experiments with cell cultures to illustrate the potential of multispectral confocal imaging for examination of cell morphology. Acridine Orange (AO) was used as the f luorophore for these experiments. At low concentration, the f luorescence of this dye is green, with a peak emission at 520 nm. However, at high concentrations, the emission spectrum shifts to the red, with a peak around 650 nm. AO is a nucleic acid stain and is perferentially attracted to low-pH environments. The highly acidic lysosomes in the cytoplasm result in preferential accumulation of AO and, with the proper input concentration, yield green emission in the nucleus and red emission in the cytoplasm.15 Figure 3 shows images obtained from a sample of human lung cancer cells in culture. We processed the spectral information at each spatial location with two spectral weight vectors to isolate the nuclei from the rest of the cell structures based on the difference in f luorescent emission spectra. Figures 3a and 3b show the normalized intensity images of the green (nuclei) and red (cytoplasm) channels, respectively. The gray-scale image in Fig. 3c was created by summing of the data along the spectral dimension and is therefore similar to an image obtained with a standard confocal microendoscope. The color reproduction shown in Fig. 3d was created by combination of the two spectral images. Figure 4 shows an image of AO-stained rat sinusoidal endothelial cells, along with spectra from the nucleus and the cytoplasm of one of the cells.
1709
In terms of imaging performance, the spatial resolution of the MCME is essentially the same as that of the confocal microendoscope, since the components that determine the spatial resolution are unchanged. The lateral resolution of the MCME is fundamentally limited by the pixel spacing in the fiber bundle. The center-to-center spacing of the f iber pixels is 3 mm. The magnification of the distal objective is 1.67, yielding a theoretical lateral resolution limit of 1.8 mm in the tissue, which is the measured value within experimental error. The axial (or depth) resolution of the slit-scan confocal system, as measured by the response to a planar object scanned through focus, is 25 mm. The spectral resolution in the MCME is determined by the dispersing power of the prism and the imaging optics in the detection arm. For the results presented here, a spectral range of 286 nm centered at 600 nm was spread over 26 pixels on the CCD. Figure 5 shows a plot of the spectrum obtained with light from a He – Ne laser fed back into the system. The FWHM of the distribution is 1.0 CCD pixel, showing that the spectral resolution of the system is indeed 11 nm.
Fig. 2. Phantom of 6-mm green, 15-mm yellow, and 15-mm red f luorescent microspheres. The microspheres were spectrally sorted into a, green; b, yellow; and c, red channels. d, Spectra of a representative microsphere from each channel.
Fig. 3. Human lung cancer cells stained with AO. a, Green channel showing primarily nucleus; b, red channel showing primarily cytoplasm; c, gray-scale image; d, RGB color reconstruction.
1710
OPTICS LETTERS / Vol. 25, No. 23 / December 1, 2000
Fig. 4. Rat sinusoidal endothelial cells stained with AO. a, RGB color reconstruction; b, spectra from the nucleus and cytoplasm of the highlighted cell.
Fig. 5.
Measured spectral response of a He–Ne laser.
We obtained the fine sampling of this distribution by tilting the CCD camera and collapsing the intensity distribution along the direction of the line image. The spectral resolution in the system can be adjusted by a change in the prism wedge angle or by variation of the tilt angle of the prism. Alternatively, the system can be conf igured with a diffraction grating if significantly higher spectral resolution is desired. The spectral range is also easily varied by altering the number of pixels read out of the CCD along the spectral dimension. The frame rate of the system is currently limited by the readout speed of the CCD. A 256 3 256 3 26 multispectral image takes approximately 4 s to acquire. Higher frame rates can be achieved with a faster CCD. However, there is a trade-off among spectral sampling rate, data-acquisition speed, and signal-to-noise ratio. The data presented here represent a reasonable compromise for f luorescence imaging of cells in culture with AO as the f luorescent dye. However, specific applications may require coarse spectral sampling, which can be achieved with faster frame rates and (or) an improved signal-to-noise ratio. The optimum trade-off between these parameters will, of course, depend on the specif ic application. Clearly, there is information stored in the spectral domain. Multispectral imaging systems, which usually employ switched spectral filters, have already shown promise for in situ diagnosis. The MCME discussed here provides high spatial resolution, the depth discrimination of a confocal system, and the potential for fine and optically efficient sampling of the spectral domain. Exactly how much spectral information is needed and how it is used are areas of
continuing research. However, the exciting results being obtained with confocal microspectrof luorometers point the way toward applications that can now be extended to in vivo imaging of remote locations inside animals and humans. We believe that multispectral imaging can be used to improve the diagnostic utility of the in vivo confocal microendoscope. Color reconstructions appear to enhance image contrast and may be valuable for real-time in vivo visualization. Spectral data might be processed to yield functional information, such as pH maps, or to provide automated disease diagnosis. In conclusion, we have introduced the concept of a multispectral confocal microendoscope for in vivo imaging. To demonstrate the principle we modified a slit-scan confocal microendoscope incorporating a fiber-optic catheter to collect multispectral data. Preliminary experiments demonstrated the ability of the system to provide high-quality multispectral images of live cells. Demonstration of the clinical utility of the instrument will require further work. This research was supported by grants from the National Institutes of Health (CA73095) and the Arizona Disease Control Research Commission (9711). We thank Ben Kriederman for providing the cell cultures and Yash Sabharwal for help with the technical development of the instrumentation. A. F. Gmitro’s e-mail address is [email protected]. References 1. K. Prashanth and A. P. Dhawan, Comput. Med. Imag. Graph. 16, 153 (1992). 2. M. Rajadhyaksha, R. R. Anderson, and R. H. Webb, Appl. Opt. 38, 2105 (1999). 3. B. R. Masters, D. J. Aziz, A. F. Gmitro, J. H. Kerr, T. C. O’Grady, and L. Goldman, J. Biomed. Opt. 2, 437 (1997). 4. H. Ichijima, W. M. Petroll, J. V. Jester, P. A. Barry, P. M. Andrews, M. Dai, and H. D. Cavanagh, Cornea 12, 369 (1993). 5. T. F. Watson, Scanning 16, 168 (1994). 6. T. Collier, P. Shen, B. de Pradier, K. Sung, A. Malpica, M. Follen, and R. Richards-Kortum, Opt. Express 6, 40 (2000), http://epubs.osa.org/opticsexpress. 7. R. H. Webb and F. J. Rogomentich, Opt. Lett. 20, 533 (1995). 8. D. L. Dickensheets and G. S. Kino, Opt. Lett. 21, 764 (1996). 9. G. J. Tearney, R. H. Webb, and B. E. Bouma, Opt. Lett. 23, 1152 (1998). 10. A. F. Gmitro and D. Aziz, Opt. Lett. 18, 565 (1993). 11. R. Juskatitis, T. Wilson, and T. F. Watson, Scanning 19, 15 (1997). 12. Y. S. Sabharwal, A. R. Rouse, L. Donaldson, M. F. Hopkins, and A. F. Gmitro, Appl. Opt. 38, 7133 (1999). 13. J. M. Millot, S. Sharonov, and M. Manfait, Cytometry 17, 50 (1994). 14. S. Sebille, M. Pereira, J. M. Millot, J. Jacquot, A. M. Delabrosie, M. Arnaud, and M. Manfait, Biochem. Biophys. Res. Commun. 246, 111 (1998). 15. C. Millot, J. M. Millot, H. Morjani, A. Desplaces, and M. Manfait, J. Histochem. Cytochem. 45, 1255 (1997).
