Content uploaded by Arnaud Dubois
Author content
All content in this area was uploaded by Arnaud Dubois on May 20, 2016
Content may be subject to copyright.
Full-field optical coherence microscopy with
optimized ultrahigh spatial resolution
ANTOINE FEDERICI AND ARNAUD DUBOIS*
Laboratoire Charles Fabry, CNRS UMR 8501, Institut d’Optique Graduate School, Univ. Paris-Sud, 2 avenue Augustin Fresnel,
91127 Palaiseau Cedex, France
*Corresponding author: arnaud.dubois@institutoptique.fr
Received 1 October 2015; revised 21 October 2015; accepted 21 October 2015; posted 21 October 2015 (Doc. ID 251019);
published 10 November 2015
Full-field optical coherence microscopy (FF-OCM) with
isotropic spatial resolution of 0.5 μm (in water), at 700 nm
center wavelength, is reported. A theoretical study of the
FF-OCM axial response is carried out for maximizing the
axial resolution of the system, considering the effect of op-
tical dispersion. The lateral resolution is optimized by using
water-immersion microscope objectives with a numerical
aperture of 1.2. This ultrahigh-resolution FF-OCM system
is applied to animal and human skin tissue imaging,
revealing ultra-fine in-depth structures at the sub-cellular
level. © 2015 Optical Society of America
OCIS codes: (180.1655) Coherence tomography; (170.4500) Optical
coherence tomography; (110.3175) Interferometric imaging; (180.6900)
Three-dimensional microscopy; (170.3660) Light propagation in
tissues.
http://dx.doi.org/10.1364/OL.40.005347
Optical coherence tomography (OCT) is a well-established tech-
nique for two- and three-dimensional (3D) imaging of semi-
transparent samples with micrometer-scale spatial resolution
[1–3]. A feature of OCT is that the axial resolution is usually
governed by the spectral properties of the signal, whereas the
transverse (lateral) resolution depends on the numerical aperture
(NA) of the lenses used for imaging [2]. By using a broadband
light source, the best reported axial resolution in OCT is
0.56 μm (in water) at ∼700 nm center wavelength, but with
important side lobes in the axial point spread function (PSF)
[4]. Besides, the transverse resolution in conventional OCT, es-
pecially in the frequency-domain version of OCT, is limited be-
cause relatively low-NA lenses are used to preserve a sufficient
depth of field [5]. Optical coherence microscopy (OCM) is a
particular approach of OCT that acquires en face tomographic
images. Since there is no depth of field constraint in OCM,
lenses of higher NA can be used to achieve better transverse res-
olution. Two general approaches for OCM have been developed
to date. The first approach is based on the combination of scan-
ning confocal microscopy with low-coherence interferometry
[6]. This approach requires transverse scanning using a spatially
coherent light source such as a superluminescent diode or a
short-pulse laser [6,7]. Transverse imaging resolution of 2 μm
has been achieved with OCM using water–immersion micro-
scope objectives with NA of 0.3 [8]. Another approach of
OCM is full-field optical coherence microscopy (FF-OCM), also
often referred to as full-field optical coherence tomography (FF-
OCT) [9]. FF-OCM is based on an interference microscope for
acquisition of en face tomographic images using an area camera
[10]. In FF-OCM, a low spatially coherent light source is em-
ployed for whole-field illumination without requiring transverse
scanning. Low coherence light sources such as a halogen lamp or
a light emitting diode (LED) are used in FF-OCM, ensuring
continuous broadband emission without spikes or modulations,
leading to higher quality of the axial PSF compared to that ob-
tained with broadband spatially coherent light sources [4,11].
Thus, FF-OCM is particularly appropriate for high-axial resolu-
tion imaging, while being adapted to high-transverse resolution
owing to the possibility of using high-NA microscope objectives
[12,13]. Three-dimensional high-resolution imaging of various
ex-vivo biological samples has been extensively performed with
FF-OCM [14,15], providing images close to those obtained
in conventional histology. Water-immersion microscope objec-
tives of NA up to 1.0 have been employed in FF-OCM for im-
aging of highly transparent samples [16,17]. Oil-immersion
objectives of NA 1.25 were used with quasi-monochromatic
light to achieve spatial rather than temporal coherence gating
[18]. Imaging with axial resolution of 1.0 μm was then demon-
strated, but with relatively weak penetration depth.
In this Letter, we present a FF-OCM system using broad-
band illumination and detection with 1.2 NA microscope
objectives to perform unprecedented ultrahigh-resolution 3D
imaging of highly scattering samples. The imaging axial reso-
lution has been optimized using a theoretical analysis of the
effect of optical dispersion in FF-OCM.
As mentioned earlier, assuming sufficiently broadband
illumination, the axial resolution in OCT is governed by the
temporal coherence of the light source [12]. According to the
Wiener–Khinchin theorem, the temporal coherence function is
related to the power spectral density of the source by a Fourier
transform. If supposing a Gaussian shape of the power spectral
density with respect to wave number σ1∕λ, the imaging
axial resolution can be expressed as
Letter Vol. 40, No. 22 / November 15 2015 / Optics Letters 5347
0146-9592/15/225347-04$15/0$15.00 © 2015 Optical Society of America
Δzlc
2ns
2ln2
πns
1
Δσ2ln2
πns
λmaxλmin
Δλ;(1)
where lcis the temporal coherence length; Δσand Δλλmax −
λmin are the full width at half-maximum (FWHM) of the spec-
trum expressed with respect to wavenumber and wavelength,
respectively; and nsis the refractive index of the sample. By
introducing λ0as the geometric mean of λmin and λmax , Eq. (1)
can be rewritten as
Δz2ln2
πns
λ2
0
Δλ;(2)
which is the usual expression of the axial resolution in OCT
[2]. If the spectrum is not too broad, λ0can be approximately
considered as the center wavelength or the peak wavelength of
the spectrum.
Equation (1) shows that for a given spectral width Δλ,a
lower λmin results in a better axial resolution. To estimate
the best axial resolution that can be achieved in OCT, we as-
sume that λmin is fixed, and we look for an optimized value of
λmax. Equation (1) can then be expressed with respect to only
one variable, λmax. Since Δzstrictly decreases for λmax greater
than λmin, we can conclude that for a Gaussian frequency spec-
trum with a fixed value of λmin, the larger the spectral width Δλ,
the better the axial resolution. We point out that if λ0is con-
sidered as the arithmetic mean of λmin and λmax , instead of the
geometric mean, an optimum for the axial resolution is found
for λmax 3λmin, which is not true.
The above analysis was directly derived from Eq. (1) and did
not take into account optical dispersion or absorption within
the sample, which are major issues when imaging in depth.
Absorption, which is mostly due to water, the main component
of biological tissues, affects the shape of the spectrum especially
for wavelengths beyond 1.2 μm. Dispersion mismatch in the
interferometer can lead to axial resolution deterioration when
the group velocity dispersion of the immersion medium differs
from that of the sample. It is then tightly linked to the spectral
width. Other than developing a FF-OCM system dedicated
to a particular sample, such a mismatch cannot be totally com-
pensated for and must be considered. On the other hand,
dispersion mismatch resulting from imperfections of the beam
splitter, or from differences between the two objectives, can easily
be corrected by setting a glass plate in each arm of the interfer-
ometer [19]. Hence, only dispersion mismatch induced by the
sample is considered in the analysis presented below.
In [20], an analytical expression has been established dem-
onstrating the effect of dispersion on the effective coherence
length with a Gaussian spectrum. However, this study did not
consider the change of the immersion medium thickness result-
ing from the dynamic focusing process specific to FF-OCM
[21]. The formula given in [20] needs to be adapted, taking
into account the variation of thickness of the immersion
medium, called δLim, to express the effective axial resolution
Δzd, at depth Ls, in the sample,
Δz2
dΔz2F2ϕ2;sLs−ϕ2;i mδLim
nsΔz2
2
:(3)
In the above expression, Δzrepresents the axial resolution at
the surface of the sample (i.e., without dispersion), ϕ2;s and
ϕ2;im are the group velocity dispersion [20] of the sample
and of the immersion medium, respectively. Fis a constant
factor whose value depends on the definition of the axial res-
olution. With the usual FWHM criterion, F4c2ln 2.To
better reveal the spread of the interferogram when dispersion
mismatch occurs, it is more appropriate to define the axial res-
olution as the full width of the fringe envelop at 1/e of the
maximum. In that case, the effective axial resolution is denoted
by Δzd;1∕eand F4c2. Since δLim is equal to Lsnim∕ns[19]
where nsand nim are the mean refractive indexes of the sample
and the immersion medium, respectively, Δzd;1∕eonly depends
on the distance Ls. Then, it is worth noticing that even though
the immersion medium has the same mean refractive index
as the sample, the GVDs have to be identical to avoid axial
resolution deterioration.
Let us consider human skin epidermis as a sample to be
imaged. Knowing its refractive index properties [22], we can
simulate the evolution of the axial resolution Δzd;1∕eas a func-
tion of the spectral width Δλat several distinct imaging depths,
by varying λmax and keeping λmin constant. We choose to fix
λmin to 500 nm, considering that for lower wavelengths light
absorption in skin is too important. Results of the numerical
simulations are presented in Fig. 1(a). At a given imaging depth
in the epidermis, a unique value of Δλ(i.e., a value of λmax )is
found to optimize the axial resolution. As expected, the larger
the imaging depth, the narrower the spectrum must be to re-
duce the effect of dispersion mismatch. Figure 1(b) shows the
evolution of the optimal axial resolution as a function of the
imaging depth in epidermis, considering water as an immersion
medium. In spite of the optimization of the spectrum width
according to the depth, a degradation of the optimal axial res-
olution with depth cannot be avoided. If the spectrum is not
adjusted during the acquisition of a z-stack of images, a com-
promise has to be done between superficial and deep axial
resolution, considering the fact that a stretching of the axial
response also reduces the signal contrast and, therefore, the
imaging penetration depth.
A way to avoid optical dispersion in OCT is to use a narrow
spectrum combined with high-NA, objectives as done in [18].
However, although potentially high axial resolution can be
achieved, based on spatial coherence gating due to narrow
depth of field, this induces important side lobes in the axial
PSF [12]. This leads to a degradation of the sectioning ability
and image quality. We measured the axial PSF of a FF-OCM
system equipped with high-NA microscope objectives, consid-
ering either a broad spectrum (KG1 filter from Schott) or a
narrow spectrum (10 nm wide FB710-10 filter). By represent-
ing the amplitude of the axial PSF in logarithmic scale
[Fig. 2(b)], which is commonly done as post data treatment
Fig. 1. (a) Axial resolution Δzd;1∕e(at 1/e) with respect to the
FWHM of the Gaussian spectrum, at different depths Lsin human
epidermis, taking into account dispersion mismatch. The cross mark-
ers point out the maximal reachable axial resolution for a given epi-
dermis layer thickness. (b) Maximal reachable axial resolution versus
depth in the epidermis. The cross markers correspond to those of (a).
5348 Vol. 40, No. 22 / November 15 2015 / Optics Letters Letter
for displaying OCT images, a significant degradation is high-
lighted when sectioning is achieved with spatial coherence
gating rather than temporal coherence gating.
The sectioning ability provided by high-NA objectives asso-
ciated to broadband illumination leads to reduced side lobes ow-
ing to the combination of the two coherence gating effects.
However, with a very broad spectrum, the effect of spatial coher-
ence gating is in practice negligible compared to the effect of tem-
poral coherence gating and has no impact on the axial resolution.
In this Letter, we report on a FF-OCM system whose im-
aging axial resolution has been optimized using the theoretical
study presented previously. The experimental setup is depicted
in Fig. 3.
The FF-OCM system uses a halogen lamp as the light
source in a Köhler illumination arrangement. The interferom-
eter is based on the Linnik configuration [12,13], employing
two 1.2 NA plan apochromatic water-immersion microscope
objectives (CFI Plan Apo VC 60XWI from Nikon) with a
300 μm-working distance. Since these high-NA objectives
are aberration corrected for imaging through a 145 μm thick
cover glass, we use the bottom surface of such a cover glass (see
Fig. 3) as a reference surface offering and, at the same time, a
low reflectivity (4%) as required for optimizing the detection
sensitivity in FF-OCM [13]. An identical cover glass is
placed above the sample, as indicated in Fig. 3. These high-
performance objectives are equipped with a correction collar
system permitting fine aberration correction depending on the
cover glass thickness. Two glass windows, referred to as DCP in
Fig. 3, are placed in the interferometer to compensate for
residual dispersion mismatch. Precise angular alignment of the
two windows was carried out to maximize the axial resolution.
A piezoelectric transducer (PZT) is used to oscillate the refer-
ence surface to perform phase-shifting interferometry [13]. En
face tomographic images are acquired by translating the refer-
ence arm step by step (0.3 μm step) to move the temporal co-
herence gate into the tissue, while the microscope objective of
the sample arm is moved simultaneously to adjust the focus.
Because of the small working distance of the objectives, the
imaging depth is limited to ∼150 μm.Theimagesareprojected
via an achromatic 400 mm focal length tube lens onto a 2D
CMOS-detector (OWL-CL from Raptor Photonics) with 320 ×
256 pixels, a frame rate up to 345 Hz, and a full well capacity of
170,000 electrons. Based on an InGaAs photodiode array with
extended sensitivity in the visible, this camera has a broad spec-
tral response expanding from 400 to 1800 nm (see Fig. 4).
A large choice of spectral bands is made possible by inserting
appropriate optical filters in the illumination path to match the
optimal configuration predicted by theory. To optimize the
spectrum regarding the axial resolution for imaging at depths
in skin between ∼10 and ∼50 μm(see Fig. 1), the KG1 band
pass filter is placed in the illumination path (resultant detected
power spectral density shown in Fig. 4). The theoretical and
experimental interferograms obtained with the KG1 filter are
shown in Fig. 5.
One can notice the high similarity between the simulation
[Fig. 5(a)] and the measurement [Fig. 5(b)], both indicating a
superficial axial resolution of Δz0.50 μmin water at a
center wavelength of ∼700 nm. Owing to the use of high-
NA microscope objectives, the system is capable of 0.5 μm spa-
tial resolution imaging in all spatial directions. To the best of
our knowledge, such an isotropic spatial resolution has never
been reached in OCT.
To extract the OCT signal from the acquired interferometric
images, a phase-shifting method based on a sinusoidal oscillation
of the reference surface is implemented [12]. Accumulation of
images is done to increase the detection sensitivity as usual in
Fig. 2. Axial PSF amplitude measured with an effective NA of ∼0.7
with broadband (blue curve) and narrow-band (red curve) illumina-
tion, represented in (a) linear and (b) logarithmic scale.
Fig. 3. FF-OCM experimental setup. AD, aperture diaphragm; F,
KG1 filter; FD, field diaphragm; BS, beam splitter; DCP, dispersion
compensation plate; MO, 1.2 NA microscope objective; PZT, piezo-
electric transducer; L1, lens; L2, tube lens (doublet).
Fig. 4. Detected power spectral density considering the absorption
by the immersion medium (water) of the microscope objectives, with-
out (dotted black curve) and with (blue curve) a KG1 filter inserted
into the illumination path. Without a KG1 filter, the power spectral
density is mainly limited by the quantum efficiency of the camera.
Fig. 5. (a) Theoretical and (b) experimental interferograms, using a
KG1 filter. The FWHMs of the theoretical and experimental interfero-
grams are identical, equal to 0.50 μm (in water).
Letter Vol. 40, No. 22 / November 15 2015 / Optics Letters 5349
FF-OCM [13,19]. By accumulating 50 en face images, leading
to an effective frame rate of ∼1Hz, a detection sensitivity of
71 dB is achieved. This relatively low detection sensitivity com-
pared to other FF-OCM systems is due to the high electronical
noise of the InGaAs array detector [19,23].
En face and reconstructed B-scan images of Xenopus Laevis
frog skin are presented in Fig. 6. The high spatial resolution
enables to see fine sub-cellular structures, such as cell mem-
branes and nuclei. One can observe that the density of the cells
increases with depth. These types of structures are also visible in
human skin as shown in Fig. 7. In this sample, the nuclei ap-
pear dark aside the light cytoplasm [24,25]. While a high axial
resolution is achieved, the imaging penetration depth is a bit
lower than in conventional FF-OCM [13–15], as a conse-
quence of the larger numerical aperture of the objectives. The
larger numerical aperture results in higher sensitivity to optical
aberrations, which reduces the signal amplitude as the imaging
depth increases. Nevertheless, one can note, especially in the
B-scan images, that the axial resolution does not perceptibly
degrade with depth and enables the distinction of thin layers
in the whole imaged volume.
We have developed a FF-OCM system with the highest spa-
tial resolution (0.5 μm in water, isotropic) ever achieved in
OCT, to the best of our knowledge. Besides, this was achieved
at a center wavelength of ∼700 nm, which is well adapted to
the imaging of biological tissues regarding optical absorption
and scattering. A theoretical study of the effect of optical
dispersion mismatch in the interferometer has been carried
out, providing the optimal spectral width depending on the
imaging depth. An experimental validation of this study has
been performed by 3D imaging of skin samples with a penetra-
tion depth close to 100 μm. As the 1.2 NA objectives used in
this FF-OCM system are very sensitive to optical aberrations,
implementation of adaptive optics could significantly improve
the imaging penetration depth [26].
Funding. TecSan (ANR-08-TECS-012-03).
Acknowledgment. The authors are grateful to Odile
Bronchain (Institut des Neurosciences Paris-Saclay).
REFERENCES
1. D. Huang, E. A. Swanson, C. P. Lin, J. S. Schuman, W. G. Stinson, W.
Chang, M. R. Hee, T. Flotte, K. Gregory, C. A. Puliafito, and J. G.
Fujimoto, Science 254, 1178 (1991).
2. A. F. Fercher, J. Biomed. Opt. 1, 157 (1996).
3. A. M. Zysk, F. T. Nguyen, A. L. Oldenburg, D. L. Marks, and S. A.
Boppart, J. Biomed. Opt. 12, 051403 (2007).
4. B. Povazay, K. Bizheva, A. Unterhuber, B. Hermann, H. Sattmann,
A. F. Fercher, W. Drexler, A. Apolonski, W. J. Wadsworth, J. C.
Knight, P. S. J. Russell, M. Vetterlein, and E. Scherzer, Opt. Lett.
27, 1800 (2002).
5. M. Wojtkowski, R. Leitgeb, A. Kowalczyk, T. Bajraszewski, and A. F.
Fercher, J. Biomed. Opt. 7, 457 (2002).
6. J. A. Izatt, E. A. Swanson, J. G. Fujimoto, M. R. Hee, and G. M. Owen,
Opt. Lett. 19, 590 (1994).
7. A. Aguirre, P. Hsiung, T. Ko, I. Hartl, and J. Fujimoto, Opt. Lett. 28,
2064 (2003).
8. Y. Chen, S.-W. Huang, A. D. Aguirre, and J. G. Fujimoto, Opt. Lett. 32,
1971 (2007).
9. E. Beaurepaire, A. C. Boccara, M. Lebec, L. Blanchot, and H. Saint-
Jalmes, Opt. Lett. 23, 244 (1998).
10. L. Vabre, A. Dubois, and A. C. Boccara, Opt. Lett. 27, 530 (2002).
11. W. Drexler, U. Morgner, F. Kärtner, C. Pitris, S. Boppart, X. Li, E.
Ippen, and J. Fujimoto, Opt. Lett. 24, 1221 (1999).
12. A. Dubois, L. Vabre, A. C. Boccara, and E. Beaurepair e, Appl. Opt. 41,
805 (2002).
13. A. Dubois, K. Grieve, G. Moneron, R. Lecaque, L. Vabre, and C.
Boccara, Appl. Opt. 43, 2874 (2004).
14. A. Dubois, G. Moneron, K. Grieve, and A. C. Boccara, Phys. Med.
Biol. 49, 1227 (2004).
15. K. Grieve, M. Paques, A. Dubois, J. Sahel, C. Boccara, and J.-F. Le
Gargasson, Invest. Ophthalmol. Visual Sci. 45, 4126 (2004).
16. W. J. Choi, K. S. Park, T. J. Eom, M.-K. Oh, and B. H. Lee, Opt. Lett.
37, 2784 (2012).
17. W. J. Choi, D. I. Jeon, S.-G. Ahn, J.-H. Yoon, S. Kim, and B. H. Lee,
Opt. Express 18, 23285 (2010).
18. A. Safrani and I. Abdulhalim, Opt. Lett. 37, 458 (2012).
19. A. Dubois, G. Moneron, and C. Boccara, Opt. Commun. 266, 738
(2006).
20. C. K. Hitzenberger, A. Baumgartner, W. Drexler, and A. F. Fercher, J.
Biomed. Opt. 4, 144 (1999).
21. S. Labiau, G. David, S. Gigan, and A. C. Boccara, Opt. Lett. 34, 1576
(2009).
22. H. F. Ding, J. Q. Lu, W. A. Wooden, P. J. Kragel, and X. H. Hu, Phys.
Med. Biol. 51, 1479 (2006).
23. W. Oh, B. Bouma, N. Iftimia, S. Yun, R. Yelin, and G. Tearney, Opt.
Express 14, 726 (2006).
24. M. Rajadhyaksha, M. Grossman, D. Esterowitz, and R. H. Webb, J.
Invest. Dermatol. 104, 946 (1995).
25. C. C. Tsai, C. K. Chang, K. Y. Hsu, T. S. Ho, M. Y. Lin, J. W. Tjiu, and
S. L. Huang, Biomed. Opt. Express 5, 3001 (2014).
26. A. Kumar, W. Drexler, and R. A. Leitgeb, Opt. Express 21, 10850
(2013).
Fig. 6. En face images of Xenopus Laevis frog skin (a)–(d) at 10, 14,
18, and 30 μm depth, respectively, and reconstructed B-scan image (e).
Scale bars: 10 μm.
Fig. 7. En face FF-OCM images of human skin acquired at depths
of (a) 8 and (b) 30 μm. B-scan image (c) obtained from a z-stack. Scale
bars: 10 μm.
5350 Vol. 40, No. 22 / November 15 2015 / Optics Letters Letter
A preview of this full-text is provided by Optica Publishing Group.
Content available from Optics Letters
This content is subject to copyright. Terms and conditions apply.