3272
Letter
Vol. 40, No. 14 / July 15 2015 / Optics Letters
Dark-field full-field optical coherence tomography EGIDIJUS AUKSORIUS*
AND
A. CLAUDE BOCCARA
Institut Langevin, ESPCI ParisTech, PSL Research University, CNRS UMR 7587, 1 rue Jussieu, 75005 Paris, France *Corresponding author: [email protected] Received 12 May 2015; accepted 14 June 2015; posted 19 June 2015 (Doc. ID 240800); published 6 July 2015
Full-field optical coherence tomography (FF-OCT) provides en face images from deep in the tissue with high spatial resolution. Specular reflections, however, may reduce image contrast as it can be much stronger than the backscattered signal from a specimen. To this end, we demonstrate dark-field FF-OCT (d -FF-OCT) that can block specular reflections by the help of an opaque disk in the pupil-conjugated plane. The reference mirror is replaced by a blazed grating, which eliminates a walk-off between the sample and the reference beams on a camera that otherwise limits the imaging field-of-view (FOV). We show that d -FF-OCT can suppress specular reflections efficiently from the glass–specimen interface by at least two orders of magnitude and yield higher contrast images compared to the conventional FF-OCT. © 2015 Optical Society of America OCIS codes: (110.4500) Optical coherence tomography; (180.3170) Interference microscopy; (170.6900) Three-dimensional microscopy; (170.3880) Medical and biological imaging. http://dx.doi.org/10.1364/OL.40.003272
Full-field optical coherence tomography (FF-OCT) is a lowcoherence interferometric imaging technique [1,2] that is capable of recording highly resolved, optically sectioned images in thick specimens [3,4]. FF-OCT differs from other OCT techniques in its ability to use an inexpensive incoherent light source, a two-dimensional detector for an en face imaging and high numerical aperture (NA) objectives that enable achieving isotropic resolution of less than 1 μm [3]. FF-OCT is a noninvasive imaging technique since it does not require staining or labeling of a specimen. However, it is less specific compared to fluorescence imaging [4,5], where light specularly reflected by a flat surface in a specimen or by an optical element can be blocked easily by an appropriate emission filter. In cell imaging, a coverslip produces specular reflections in the range of 10−2 –10−3 , whereas reflectivity of cells can be as low as 10−6 –10−8 . If the reflections from the coverslip take a substantial part of the camera’s dynamic range, it may reduce image contrast significantly. An image from a glass–specimen interface usually suffers the most since a glass slide produces the strongest signal at this plane that can completely mask the specimen 0146-9592/15/143272-04$15/0$15.00 © 2015 Optical Society of America
signal. Some medical devices, for example, circulating tumor cell chips [6] also have a glass layer that strongly reflects light and can challenge the use of OCT. Moreover, to reduce surface topography-induced optical aberrations and to ensure that the entire field-of-view (FOV) of an en face image contains a signal, a cover glass is often used to flatten out a specimen. Antireflection-coated coverslip or a liquid-immersion objective can be used to minimize specular reflections coming from a glass surface that is closer to an objective, but the glass– specimen interface still might produce a strong specular reflection because of the refractive index mismatch between the glass (or coating) and the specimen. In addition, coated coverslips are not a practical solution and do not work well with spectrally broadband light and with high NA objectives. In addition, spectrally uneven coating modulates the sample beam compromising the optical sectioning. Furthermore, a dry objective might be preferred in high-speed scanning applications if the drag created between liquid and a specimen limits the imaging speed. Even though the main application of OCT is in-depth imaging, in many cases, surface can contain important information. For example, if a cover glass is used to flatten out skin, the signal from the skin surface (stratum corneum layer, which could be as thin as ∼10 μm [7]) might be dominated by the OCT signal coming from the cover glass–skin interface. Strong specular reflections at surfaces also can be an issue for in vivo imaging applications. For example, reflections at a tooth surface preclude imaging of shallow demineralization [8]. The air–tear and cornea–aqueous interfaces in the eye also are known to produce strong specular reflections. Strongly reflecting planes in layered structures [9] can mask a signal coming from between the planes. Finally, specular reflections can limit imaging depth because of compromised detection sensitivity that is defined as the smallest detectable reflectivity, R m [2] [which can be approximated as R m ≈ α∕ξ, where α is the reflectivity of the reference arm reflector and ξ—full well capacity (FWC) of an individual pixel in a camera]. Sensitivity is optimized when α is approximately equal to the incoherent light [2] that includes light specularly reflected from flat surfaces and light coming from other planes of tissue, all of which do not interfere on a camera. Specular reflections increase the amount of incoherent light and, in turn, compromise detection sensitivity and imaging depth. One can tilt a specimen to direct the specular reflections away from the acceptance angle of the objective lens
Letter [9], which is inconvenient, especially in high-NA imaging cases. Moreover, it often is required to image exactly the same focal plane (or a plane parallel to it) that generates specular reflections. Dark-field detection often is employed to block specular reflections by putting an opaque stop (a block) in the pupil plane of the objective. It is employed in a wide range of applications spanning from medical devices [10] to semiconductor defect inspection systems [11]. Dark-field detection has been demonstrated previously with the coherent light source in a point scanning Fourier domain optical coherence microscopy (OCM) [12] and in parallel OCT configuration with a cumbersome Mach–Zehnder interferometer [13]. Point scanning OCM has to acquire a three-dimensional data volume first to reconstruct an en face image from a single focal plane. FF-OCT, in contrast, can acquire an image from a single focal plane directly without the need to scan a specimen and, therefore, could be used in applications where fast en face imaging is essential. A block put in the center of the pupil plane also can block the reference beam, as it is normally aligned along the optical axis. In principle, the reference beam can be guided around the block by simply tilting the reference mirror, provided that its size is small enough to pass between the edges of the block and the pupil. However, the coherence gate plane of the reference beam also tilts with respect to the coherence gate plane of the incoming beam (see upper inset in Fig. 1) and that of the sample beam. In the case of imaging a flat object, the coherence gate tilt would make the beams interfere in a narrower region on a camera, limiting the imaging FOV. In an off-axis interferometry, the problem is known as a beam walk-off, which can be solved by inserting prisms in the reference and sample arms [14], or a transmission grating in the reference beam [15]. In this Letter, we demonstrate dark-field FF-OCT (d -FFOCT) by implementing dark-field detection in a conventional high-resolution FF-OCT system that consists of a Linnik interferometer and spatially incoherent, broadband light source. d FF-OCT is realized by putting a block in the pupil-conjugated plane and by replacing the reference mirror with a blazed reflective diffraction grating. A spatially incoherent and broadband (Δλ 100 nm) light emitting diode (M565L3, Thorlabs), radiating at the central wavelength of λc 565 nm, is relayed onto the pupil planes of two identical objectives (×10, NA 0.25, Olympus) with the help of two (f 10 cm) lenses, arranged in 4f configuration (Fig. 1). The lenses, together with two irises—the aperture and the field diaphragms—form the Kohler illumination. The illumination is split into the reference and sample arms by a 50∶50 beamsplitter. Since the objectives used here have their pupil planes inside their objective bodies, we reimage the planes by two (f 20 cm) lenses in a 4f configuration to a physically accessible location that is pupil conjugated (pupil’ in Fig. 1). Because of the use of a spatially extended (incoherent) source, the block size has to match the aperture diaphragm in the Kohler configuration, to ensure that the illumination and the detection paths are separated fully to block all the reflected light. In the case of monochromatic illumination, the size of the block has to be 1/3 of the pupil diameter to make it the same size as the thickness of the annulus, for the reference beam to go through
Vol. 40, No. 14 / July 15 2015 / Optics Letters
3273
Fig. 1. Schematic diagram of d -FF-OCT. Scattered light from a specimen (gray) interferes on a camera with the reference beam (light green). Specular reflections from a specimen (and the zeroth-order from the blazed grating) are obstructed by a block in the pupilconjugate plane (pupil’). The reference beam is dispersed spectrally on the pupil-conjugated plane, as shown in the lower inset. The aperture diaphragm is matched to the size of the block of 2 mm. Note that in the diagram, for the sake of simplicity, the beam path is shown for the spatially and temporally coherent illumination case—only one ray, originating at the center of the LED, is drawn, which also does not experience dispersion in the pupil’ plane. Thus, the beam is depicted to focus to a spot on a pupil plane, rather than to a 2 mm disk, as is shown in the lower inset. Upper inset: coherence gate tilt with a mirror and a grating.
the annulus unobstructed. For spectrally broadband illumination, the annulus is made bigger and, correspondingly, the block is made smaller to accommodate the expansion of the reference beam in the pupil plane because of the spectral dispersion (lower inset of Fig. 1) that the illumination beam experiences when diffracting off the grating. The diameters of the block and the annulus are chosen to be of 2 and 3.6 mm, respectively. The aperture diaphragm has to be reduced to 2 mm; however, this can lower the throughput of the illumination light, which can be an issue in low light budget cases. A blazed grating with the periodicity of 300 lines∕mm and the blaze wavelength of 500 nm is used as a reflector in the reference arm, which diffracts the reference beam to the first diffraction order at 9.8° (at 565 nm). The beam is focused 3.1 mm away from the center in the pupil plane. The spectral
3274
Vol. 40, No. 14 / July 15 2015 / Optics Letters
broadening in the pupil because of the use of the diffraction grating does not limit the sectioning strength since the FWHM value of the spectral bandwidth that can pass through the annulus unobstructed by the edges of the block or the pupil is estimated to be of 0.29 μm, corresponding to the optical sectioning strength of 0.48 μm. The reference beam and the scattered light from a specimen that was not suppressed by the block in the pupil-conjugated plane is imaged by a (f 20 cm) lens onto a fast (138 fps) CMOS camera (MV-D1024E-160-CL-12, PhotonFocus) with large FWC (ξ 0.2 Me− ). A fast piezo actuator is used to move the grating along the optical axis to introduce a varying optical path length difference (OPLD) between the reference and the sample beams. The beams form an interference image on the camera if the OPLD is within the temporal coherence length of the light source. If the grating is moved to another position, the interference image is changed but the background signal, coming outside the coherence gate, stays the same. Therefore, subtracting the two images gets rid of the background light and subsequently produces an optically sectioned image. To minimize reconstructing artifacts in an FF-OCT image, generally four images are acquired, which takes 29 ms (34 Hz), with λc ∕4 OPLD in between them [1]. Many FF-OCT images are averaged to improve signal-to-noise ratio (S/N), typically resulting in the acquisition times of around 1 s. Images are virtually free of grating modulations because light diffracts mostly to the 1 diffraction order from the grating, whereas the zeroth-orders coming from the grating and the specimen are eliminated by the block. FF-OCT images at different depths of a specimen are recorded using a motor to move it along the z axis. The reference arm is attenuated by a factor of 10 with the neutral density filter to optimize the detection sensitivity [2], resulting in the reference arm reflectivity of α 0.07. A blank slide of the same thickness and material is inserted in the sample arm to compensate for the chromatic dispersion. Since dark-field minimizes incoherent light, the smallest detectable reflectivity, R m can be estimated from R m α∕8 × ξ. In the case of 50 averaged d -FF-OCT images, the estimated sensitivity is 10−9 or 90 db. In practice, R m is twice as high since the LED used here saturates only half of the FWC. Limited light budget also precludes us from demonstrating sensitivity increase in d -FF-OCT. First, to illustrate d -FF-OCT imaging on a known specimen, a USAF resolution target was imaged and compared to the conventional FF-OCT and dark-field images. For the conventional FF-OCT imaging the grating was replaced with the uncoated neutral density filter that provided ∼4% reflection. The reference and the sample beams were aligned collinearly so that the whole FOV can be captured in a single en face image [Fig. 2(a)]. For the conventional dark-field imaging, the reference beam was blocked. The image [Fig. 2(b)] confirms that the dominating contrast mechanism is light scattering since the signal is coming from the edges of the resolution bars. Very little signal comes from the flat surface, except that which is generated by random scatterers present on the surface. An image acquired with d -FF-OCT [Fig. 2(c)] looks very similar to the conventional dark-field image [Fig. 2(b)] because of the same contrast mechanism. The difference, however, is that d -FF-OCT has the optical
Letter
Fig. 2. USAF resolution target images recorded in (a) FF-OCT, (b) dark-field, and (c) d -FF-OCT. The curves in figure (d) show optical sectioning for the (b) conventional dark-field and (c) d -FF-OCT. The curve for (c) was derived by taking a line profile along the optical axis over a small scatterer that was present on the flat surface of the target. A conventional dark-field image (b) was backgroundsubtracted, as explained in the text. FOV 954 μm.
sectioning strength of 1.5 μm, which is in good agreement with the theoretical estimate of 1.4 μm, whereas conventional darkfield has no sectioning capability, as illustrated in the graph in Fig. 2(d). d -FF-OCT provides isotropical resolution of 1.5 μm since the estimated lateral resolution is also 1.5 μm. If the illumination light is converging when it goes through the beamsplitter, a surface of which is not close to the pupil plane of the objective and the antireflection coating is not perfect, some of the light may be reflected toward the camera. Such reflection does not focus precisely on the block in the pupil-conjugate plane and some of it leaks through, contributing to the background. For example, dark-field image in Fig. 2(b) is background-subtracted because the reflections coming from the beamsplitter contributed almost 50% of the total signal and were nonuniform. However, the background does not have to be subtracted in the d -FF-OCT image as it gets eliminated by the coherence gating. Generally, conventional dark-field images suffer from scattering occurring anywhere in the optical path—if a dust or some other imperfection in an optical element scatters light, it might surpass the block and be detected by a camera. d -FF-OCT enables acquisition of optically sectioned dark-field images, which helps to get rid of the unwanted reflected and scattered light by either physically blocking it or gating it through interferometric detection. The d -FF-OCT in vivo image of human skin, shown in Fig. 3(b), also clearly illustrates that light is scattered at the
Letter
Fig. 3. Images of human skin recorded in vivo at the coverslip–skin interface with (a) FF-OCT and (b) d -FF-OCT. Acquisition time: 0.6 s (20 FF-OCT images averaged).
Vol. 40, No. 14 / July 15 2015 / Optics Letters
3275
specular reflections are no longer affecting the image contrast. This is because lateral resolution for both techniques is practically the same and the main contrast generating mechanism deep in the tissue is scattering. In conclusion, we demonstrated d -FF-OCT with isotropical resolution of 1.5 μm, which consisted of a dark-field detection implemented in a Linnik-type FF-OCT system. The principle of d -FF-OCT was realized by putting a block in the central part of the pupil-conjugated plane and by replacing the reference mirror with a blazed reflective diffraction grating. d -FF-OCT enabled capturing high contrast OCT images at a coverslip–tissue interface by suppressing specular reflections by at least two orders of magnitude, which otherwise severely reduce image contrast. Funding. European Programme (312792).
Union’s
Seventh
Framework
Acknowledgment. The authors wish to thank Yaron Bromberg for fruitful discussions and valuable suggestions. REFERENCES
Fig. 4. Images of a fresh lamb brain slice recorded at the coverslip– brain interface with (a) FF-OCT and (b) d -FF-OCT. Individual myelinated axons are clearly visible in image (b) because of the increased image contrast achieved by rejecting specular reflections. Acquisition time: 1.5 s (50 FF-OCT images averaged).
edges which, in this particular case, is created around the skin wrinkles or other voids in contact with a coverslip. The bright parts in the FF-OCT image [Fig. 3(a)] are generated by the specular reflections coming from the coverslip–air interface. Some other places produce less specularly reflected light because of the smaller coverslip–skin refractive index mismatch. Specular reflections are blocked in the d -FF-OCT image [Fig. 3(b)], with the suppression factor measured to be more than 20 db. Furthermore, we demonstrate that d -FF-OCT can discern clearly myelinated axons in a fresh lamb brain slice [Fig. 4(b)] that cannot be seen in the conventional FF-OCT image [Fig. 4(a)] because of the strong specular reflections completely dominating the signal. This example is a clear demonstration that d -FF-OCT can be used to acquire OCT images close to a highly reflective surface, which would otherwise severely reduce the image contrast. d -FF-OCT images visually look very similar to FF-OCT images below the coverslip, where
1. E. Beaurepaire, A. C. Boccara, M. Lebec, L. Blanchot, and H. SaintJalmes, Opt. Lett. 23, 244 (1998). 2. A. Dubois, L. Vabre, A. C. Boccara, and E. Beaurepaire, Appl. Opt. 41, 805 (2002). 3. A. Dubois, G. Moneron, K. Grieve, and A. C. Boccara, Phys. Med. Biol. 49, 1227 (2004). 4. E. Coron, E. Auksorius, A. Pieretti, M. M. Mahé, L. Liu, C. Steiger, Y. Bromberg, B. Bouma, G. Tearney, M. Neunlist, and A. M. Goldstein, Neurogastroenterol. Motil. 24, e611 (2012). 5. E. Auksorius, Y. Bromberg, R. Motiejuenaite, A. Pieretti, L. Liu, E. Coron, J. Aranda, A. M. Goldstein, B. E. Bouma, A. Kazlauskas, and G. J. Tearney, Biomed. Opt. Express 3, 661 (2012). 6. S. L. Stott, C.-H. Hsu, D. I. Tsukrov, M. Yu, D. T. Miyamoto, B. A. Waltman, S. M. Rothenberg, A. M. Shah, M. E. Smas, G. K. Korir, F. P. Floyd, A. J. Gilman, J. B. Lord, D. Winokur, S. Springer, D. Irimia, S. Nagrath, L. V. Sequist, R. J. Lee, K. J. Isselbacher, S. Maheswaran, D. A. Haber, and M. Toner, Proc. Natl. Acad. Sci. 107, 18392 (2010). 7. E. Dalimier and D. Salomon, Dermatology 224, 84 (2012). 8. T. Louie, C. Lee, D. Hsu, K. Hirasuna, S. Manesh, M. Staninec, C. L. Darling, and D. Fried, Lasers Surg. Med. 42, 898 (2010). 9. L. Vabre, V. Loriette, A. Dubois, J. Moreau, and A. C. Boccara, Opt. Lett. 27, 1899 (2002). 10. P. T. Goedhart, M. Khalilzada, R. Bezemer, J. Merza, and C. Ince, Opt. Express 15, 15101 (2007). 11. M. Vaez-Iravani, “Single laser bright field and dark field system for detecting anomalies of a sample,” U.S. patent 5,798,829 (August 25, 1998). 12. M. Villiger, C. Pache, and T. Lasser, Opt. Lett. 35, 3489 (2010). 13. S. E. Schausberger, B. Heise, C. Maurer, S. Bernet, M. Ritsch-Marte, and D. Stifter, Opt. Lett. 35, 4154 (2010). 14. Z. Ansari, Y. Gu, M. Tziraki, R. Jones, P. M. W. French, D. D. Nolte, and M. R. Melloch, Opt. Lett. 26, 334 (2001). 15. Y. Choi, T. D. Yang, K. J. Lee, and W. Choi, Opt. Lett. 36, 2465 (2011).
Letter
Vol. 40, No. 14 / July 15 2015 / Optics Letters
Dark-field full-field optical coherence tomography EGIDIJUS AUKSORIUS*
AND
A. CLAUDE BOCCARA
Institut Langevin, ESPCI ParisTech, PSL Research University, CNRS UMR 7587, 1 rue Jussieu, 75005 Paris, France *Corresponding author: [email protected] Received 12 May 2015; accepted 14 June 2015; posted 19 June 2015 (Doc. ID 240800); published 6 July 2015
Full-field optical coherence tomography (FF-OCT) provides en face images from deep in the tissue with high spatial resolution. Specular reflections, however, may reduce image contrast as it can be much stronger than the backscattered signal from a specimen. To this end, we demonstrate dark-field FF-OCT (d -FF-OCT) that can block specular reflections by the help of an opaque disk in the pupil-conjugated plane. The reference mirror is replaced by a blazed grating, which eliminates a walk-off between the sample and the reference beams on a camera that otherwise limits the imaging field-of-view (FOV). We show that d -FF-OCT can suppress specular reflections efficiently from the glass–specimen interface by at least two orders of magnitude and yield higher contrast images compared to the conventional FF-OCT. © 2015 Optical Society of America OCIS codes: (110.4500) Optical coherence tomography; (180.3170) Interference microscopy; (170.6900) Three-dimensional microscopy; (170.3880) Medical and biological imaging. http://dx.doi.org/10.1364/OL.40.003272
Full-field optical coherence tomography (FF-OCT) is a lowcoherence interferometric imaging technique [1,2] that is capable of recording highly resolved, optically sectioned images in thick specimens [3,4]. FF-OCT differs from other OCT techniques in its ability to use an inexpensive incoherent light source, a two-dimensional detector for an en face imaging and high numerical aperture (NA) objectives that enable achieving isotropic resolution of less than 1 μm [3]. FF-OCT is a noninvasive imaging technique since it does not require staining or labeling of a specimen. However, it is less specific compared to fluorescence imaging [4,5], where light specularly reflected by a flat surface in a specimen or by an optical element can be blocked easily by an appropriate emission filter. In cell imaging, a coverslip produces specular reflections in the range of 10−2 –10−3 , whereas reflectivity of cells can be as low as 10−6 –10−8 . If the reflections from the coverslip take a substantial part of the camera’s dynamic range, it may reduce image contrast significantly. An image from a glass–specimen interface usually suffers the most since a glass slide produces the strongest signal at this plane that can completely mask the specimen 0146-9592/15/143272-04$15/0$15.00 © 2015 Optical Society of America
signal. Some medical devices, for example, circulating tumor cell chips [6] also have a glass layer that strongly reflects light and can challenge the use of OCT. Moreover, to reduce surface topography-induced optical aberrations and to ensure that the entire field-of-view (FOV) of an en face image contains a signal, a cover glass is often used to flatten out a specimen. Antireflection-coated coverslip or a liquid-immersion objective can be used to minimize specular reflections coming from a glass surface that is closer to an objective, but the glass– specimen interface still might produce a strong specular reflection because of the refractive index mismatch between the glass (or coating) and the specimen. In addition, coated coverslips are not a practical solution and do not work well with spectrally broadband light and with high NA objectives. In addition, spectrally uneven coating modulates the sample beam compromising the optical sectioning. Furthermore, a dry objective might be preferred in high-speed scanning applications if the drag created between liquid and a specimen limits the imaging speed. Even though the main application of OCT is in-depth imaging, in many cases, surface can contain important information. For example, if a cover glass is used to flatten out skin, the signal from the skin surface (stratum corneum layer, which could be as thin as ∼10 μm [7]) might be dominated by the OCT signal coming from the cover glass–skin interface. Strong specular reflections at surfaces also can be an issue for in vivo imaging applications. For example, reflections at a tooth surface preclude imaging of shallow demineralization [8]. The air–tear and cornea–aqueous interfaces in the eye also are known to produce strong specular reflections. Strongly reflecting planes in layered structures [9] can mask a signal coming from between the planes. Finally, specular reflections can limit imaging depth because of compromised detection sensitivity that is defined as the smallest detectable reflectivity, R m [2] [which can be approximated as R m ≈ α∕ξ, where α is the reflectivity of the reference arm reflector and ξ—full well capacity (FWC) of an individual pixel in a camera]. Sensitivity is optimized when α is approximately equal to the incoherent light [2] that includes light specularly reflected from flat surfaces and light coming from other planes of tissue, all of which do not interfere on a camera. Specular reflections increase the amount of incoherent light and, in turn, compromise detection sensitivity and imaging depth. One can tilt a specimen to direct the specular reflections away from the acceptance angle of the objective lens
Letter [9], which is inconvenient, especially in high-NA imaging cases. Moreover, it often is required to image exactly the same focal plane (or a plane parallel to it) that generates specular reflections. Dark-field detection often is employed to block specular reflections by putting an opaque stop (a block) in the pupil plane of the objective. It is employed in a wide range of applications spanning from medical devices [10] to semiconductor defect inspection systems [11]. Dark-field detection has been demonstrated previously with the coherent light source in a point scanning Fourier domain optical coherence microscopy (OCM) [12] and in parallel OCT configuration with a cumbersome Mach–Zehnder interferometer [13]. Point scanning OCM has to acquire a three-dimensional data volume first to reconstruct an en face image from a single focal plane. FF-OCT, in contrast, can acquire an image from a single focal plane directly without the need to scan a specimen and, therefore, could be used in applications where fast en face imaging is essential. A block put in the center of the pupil plane also can block the reference beam, as it is normally aligned along the optical axis. In principle, the reference beam can be guided around the block by simply tilting the reference mirror, provided that its size is small enough to pass between the edges of the block and the pupil. However, the coherence gate plane of the reference beam also tilts with respect to the coherence gate plane of the incoming beam (see upper inset in Fig. 1) and that of the sample beam. In the case of imaging a flat object, the coherence gate tilt would make the beams interfere in a narrower region on a camera, limiting the imaging FOV. In an off-axis interferometry, the problem is known as a beam walk-off, which can be solved by inserting prisms in the reference and sample arms [14], or a transmission grating in the reference beam [15]. In this Letter, we demonstrate dark-field FF-OCT (d -FFOCT) by implementing dark-field detection in a conventional high-resolution FF-OCT system that consists of a Linnik interferometer and spatially incoherent, broadband light source. d FF-OCT is realized by putting a block in the pupil-conjugated plane and by replacing the reference mirror with a blazed reflective diffraction grating. A spatially incoherent and broadband (Δλ 100 nm) light emitting diode (M565L3, Thorlabs), radiating at the central wavelength of λc 565 nm, is relayed onto the pupil planes of two identical objectives (×10, NA 0.25, Olympus) with the help of two (f 10 cm) lenses, arranged in 4f configuration (Fig. 1). The lenses, together with two irises—the aperture and the field diaphragms—form the Kohler illumination. The illumination is split into the reference and sample arms by a 50∶50 beamsplitter. Since the objectives used here have their pupil planes inside their objective bodies, we reimage the planes by two (f 20 cm) lenses in a 4f configuration to a physically accessible location that is pupil conjugated (pupil’ in Fig. 1). Because of the use of a spatially extended (incoherent) source, the block size has to match the aperture diaphragm in the Kohler configuration, to ensure that the illumination and the detection paths are separated fully to block all the reflected light. In the case of monochromatic illumination, the size of the block has to be 1/3 of the pupil diameter to make it the same size as the thickness of the annulus, for the reference beam to go through
Vol. 40, No. 14 / July 15 2015 / Optics Letters
3273
Fig. 1. Schematic diagram of d -FF-OCT. Scattered light from a specimen (gray) interferes on a camera with the reference beam (light green). Specular reflections from a specimen (and the zeroth-order from the blazed grating) are obstructed by a block in the pupilconjugate plane (pupil’). The reference beam is dispersed spectrally on the pupil-conjugated plane, as shown in the lower inset. The aperture diaphragm is matched to the size of the block of 2 mm. Note that in the diagram, for the sake of simplicity, the beam path is shown for the spatially and temporally coherent illumination case—only one ray, originating at the center of the LED, is drawn, which also does not experience dispersion in the pupil’ plane. Thus, the beam is depicted to focus to a spot on a pupil plane, rather than to a 2 mm disk, as is shown in the lower inset. Upper inset: coherence gate tilt with a mirror and a grating.
the annulus unobstructed. For spectrally broadband illumination, the annulus is made bigger and, correspondingly, the block is made smaller to accommodate the expansion of the reference beam in the pupil plane because of the spectral dispersion (lower inset of Fig. 1) that the illumination beam experiences when diffracting off the grating. The diameters of the block and the annulus are chosen to be of 2 and 3.6 mm, respectively. The aperture diaphragm has to be reduced to 2 mm; however, this can lower the throughput of the illumination light, which can be an issue in low light budget cases. A blazed grating with the periodicity of 300 lines∕mm and the blaze wavelength of 500 nm is used as a reflector in the reference arm, which diffracts the reference beam to the first diffraction order at 9.8° (at 565 nm). The beam is focused 3.1 mm away from the center in the pupil plane. The spectral
3274
Vol. 40, No. 14 / July 15 2015 / Optics Letters
broadening in the pupil because of the use of the diffraction grating does not limit the sectioning strength since the FWHM value of the spectral bandwidth that can pass through the annulus unobstructed by the edges of the block or the pupil is estimated to be of 0.29 μm, corresponding to the optical sectioning strength of 0.48 μm. The reference beam and the scattered light from a specimen that was not suppressed by the block in the pupil-conjugated plane is imaged by a (f 20 cm) lens onto a fast (138 fps) CMOS camera (MV-D1024E-160-CL-12, PhotonFocus) with large FWC (ξ 0.2 Me− ). A fast piezo actuator is used to move the grating along the optical axis to introduce a varying optical path length difference (OPLD) between the reference and the sample beams. The beams form an interference image on the camera if the OPLD is within the temporal coherence length of the light source. If the grating is moved to another position, the interference image is changed but the background signal, coming outside the coherence gate, stays the same. Therefore, subtracting the two images gets rid of the background light and subsequently produces an optically sectioned image. To minimize reconstructing artifacts in an FF-OCT image, generally four images are acquired, which takes 29 ms (34 Hz), with λc ∕4 OPLD in between them [1]. Many FF-OCT images are averaged to improve signal-to-noise ratio (S/N), typically resulting in the acquisition times of around 1 s. Images are virtually free of grating modulations because light diffracts mostly to the 1 diffraction order from the grating, whereas the zeroth-orders coming from the grating and the specimen are eliminated by the block. FF-OCT images at different depths of a specimen are recorded using a motor to move it along the z axis. The reference arm is attenuated by a factor of 10 with the neutral density filter to optimize the detection sensitivity [2], resulting in the reference arm reflectivity of α 0.07. A blank slide of the same thickness and material is inserted in the sample arm to compensate for the chromatic dispersion. Since dark-field minimizes incoherent light, the smallest detectable reflectivity, R m can be estimated from R m α∕8 × ξ. In the case of 50 averaged d -FF-OCT images, the estimated sensitivity is 10−9 or 90 db. In practice, R m is twice as high since the LED used here saturates only half of the FWC. Limited light budget also precludes us from demonstrating sensitivity increase in d -FF-OCT. First, to illustrate d -FF-OCT imaging on a known specimen, a USAF resolution target was imaged and compared to the conventional FF-OCT and dark-field images. For the conventional FF-OCT imaging the grating was replaced with the uncoated neutral density filter that provided ∼4% reflection. The reference and the sample beams were aligned collinearly so that the whole FOV can be captured in a single en face image [Fig. 2(a)]. For the conventional dark-field imaging, the reference beam was blocked. The image [Fig. 2(b)] confirms that the dominating contrast mechanism is light scattering since the signal is coming from the edges of the resolution bars. Very little signal comes from the flat surface, except that which is generated by random scatterers present on the surface. An image acquired with d -FF-OCT [Fig. 2(c)] looks very similar to the conventional dark-field image [Fig. 2(b)] because of the same contrast mechanism. The difference, however, is that d -FF-OCT has the optical
Letter
Fig. 2. USAF resolution target images recorded in (a) FF-OCT, (b) dark-field, and (c) d -FF-OCT. The curves in figure (d) show optical sectioning for the (b) conventional dark-field and (c) d -FF-OCT. The curve for (c) was derived by taking a line profile along the optical axis over a small scatterer that was present on the flat surface of the target. A conventional dark-field image (b) was backgroundsubtracted, as explained in the text. FOV 954 μm.
sectioning strength of 1.5 μm, which is in good agreement with the theoretical estimate of 1.4 μm, whereas conventional darkfield has no sectioning capability, as illustrated in the graph in Fig. 2(d). d -FF-OCT provides isotropical resolution of 1.5 μm since the estimated lateral resolution is also 1.5 μm. If the illumination light is converging when it goes through the beamsplitter, a surface of which is not close to the pupil plane of the objective and the antireflection coating is not perfect, some of the light may be reflected toward the camera. Such reflection does not focus precisely on the block in the pupil-conjugate plane and some of it leaks through, contributing to the background. For example, dark-field image in Fig. 2(b) is background-subtracted because the reflections coming from the beamsplitter contributed almost 50% of the total signal and were nonuniform. However, the background does not have to be subtracted in the d -FF-OCT image as it gets eliminated by the coherence gating. Generally, conventional dark-field images suffer from scattering occurring anywhere in the optical path—if a dust or some other imperfection in an optical element scatters light, it might surpass the block and be detected by a camera. d -FF-OCT enables acquisition of optically sectioned dark-field images, which helps to get rid of the unwanted reflected and scattered light by either physically blocking it or gating it through interferometric detection. The d -FF-OCT in vivo image of human skin, shown in Fig. 3(b), also clearly illustrates that light is scattered at the
Letter
Fig. 3. Images of human skin recorded in vivo at the coverslip–skin interface with (a) FF-OCT and (b) d -FF-OCT. Acquisition time: 0.6 s (20 FF-OCT images averaged).
Vol. 40, No. 14 / July 15 2015 / Optics Letters
3275
specular reflections are no longer affecting the image contrast. This is because lateral resolution for both techniques is practically the same and the main contrast generating mechanism deep in the tissue is scattering. In conclusion, we demonstrated d -FF-OCT with isotropical resolution of 1.5 μm, which consisted of a dark-field detection implemented in a Linnik-type FF-OCT system. The principle of d -FF-OCT was realized by putting a block in the central part of the pupil-conjugated plane and by replacing the reference mirror with a blazed reflective diffraction grating. d -FF-OCT enabled capturing high contrast OCT images at a coverslip–tissue interface by suppressing specular reflections by at least two orders of magnitude, which otherwise severely reduce image contrast. Funding. European Programme (312792).
Union’s
Seventh
Framework
Acknowledgment. The authors wish to thank Yaron Bromberg for fruitful discussions and valuable suggestions. REFERENCES
Fig. 4. Images of a fresh lamb brain slice recorded at the coverslip– brain interface with (a) FF-OCT and (b) d -FF-OCT. Individual myelinated axons are clearly visible in image (b) because of the increased image contrast achieved by rejecting specular reflections. Acquisition time: 1.5 s (50 FF-OCT images averaged).
edges which, in this particular case, is created around the skin wrinkles or other voids in contact with a coverslip. The bright parts in the FF-OCT image [Fig. 3(a)] are generated by the specular reflections coming from the coverslip–air interface. Some other places produce less specularly reflected light because of the smaller coverslip–skin refractive index mismatch. Specular reflections are blocked in the d -FF-OCT image [Fig. 3(b)], with the suppression factor measured to be more than 20 db. Furthermore, we demonstrate that d -FF-OCT can discern clearly myelinated axons in a fresh lamb brain slice [Fig. 4(b)] that cannot be seen in the conventional FF-OCT image [Fig. 4(a)] because of the strong specular reflections completely dominating the signal. This example is a clear demonstration that d -FF-OCT can be used to acquire OCT images close to a highly reflective surface, which would otherwise severely reduce the image contrast. d -FF-OCT images visually look very similar to FF-OCT images below the coverslip, where
1. E. Beaurepaire, A. C. Boccara, M. Lebec, L. Blanchot, and H. SaintJalmes, Opt. Lett. 23, 244 (1998). 2. A. Dubois, L. Vabre, A. C. Boccara, and E. Beaurepaire, Appl. Opt. 41, 805 (2002). 3. A. Dubois, G. Moneron, K. Grieve, and A. C. Boccara, Phys. Med. Biol. 49, 1227 (2004). 4. E. Coron, E. Auksorius, A. Pieretti, M. M. Mahé, L. Liu, C. Steiger, Y. Bromberg, B. Bouma, G. Tearney, M. Neunlist, and A. M. Goldstein, Neurogastroenterol. Motil. 24, e611 (2012). 5. E. Auksorius, Y. Bromberg, R. Motiejuenaite, A. Pieretti, L. Liu, E. Coron, J. Aranda, A. M. Goldstein, B. E. Bouma, A. Kazlauskas, and G. J. Tearney, Biomed. Opt. Express 3, 661 (2012). 6. S. L. Stott, C.-H. Hsu, D. I. Tsukrov, M. Yu, D. T. Miyamoto, B. A. Waltman, S. M. Rothenberg, A. M. Shah, M. E. Smas, G. K. Korir, F. P. Floyd, A. J. Gilman, J. B. Lord, D. Winokur, S. Springer, D. Irimia, S. Nagrath, L. V. Sequist, R. J. Lee, K. J. Isselbacher, S. Maheswaran, D. A. Haber, and M. Toner, Proc. Natl. Acad. Sci. 107, 18392 (2010). 7. E. Dalimier and D. Salomon, Dermatology 224, 84 (2012). 8. T. Louie, C. Lee, D. Hsu, K. Hirasuna, S. Manesh, M. Staninec, C. L. Darling, and D. Fried, Lasers Surg. Med. 42, 898 (2010). 9. L. Vabre, V. Loriette, A. Dubois, J. Moreau, and A. C. Boccara, Opt. Lett. 27, 1899 (2002). 10. P. T. Goedhart, M. Khalilzada, R. Bezemer, J. Merza, and C. Ince, Opt. Express 15, 15101 (2007). 11. M. Vaez-Iravani, “Single laser bright field and dark field system for detecting anomalies of a sample,” U.S. patent 5,798,829 (August 25, 1998). 12. M. Villiger, C. Pache, and T. Lasser, Opt. Lett. 35, 3489 (2010). 13. S. E. Schausberger, B. Heise, C. Maurer, S. Bernet, M. Ritsch-Marte, and D. Stifter, Opt. Lett. 35, 4154 (2010). 14. Z. Ansari, Y. Gu, M. Tziraki, R. Jones, P. M. W. French, D. D. Nolte, and M. R. Melloch, Opt. Lett. 26, 334 (2001). 15. Y. Choi, T. D. Yang, K. J. Lee, and W. Choi, Opt. Lett. 36, 2465 (2011).
![[Optical coherence tomography].](/assets/img/hold.png)
