August 15, 2002 / Vol. 27, No.

16 / OPTICS LETTERS 1415

Full range complex spectral optical coherence tomography

technique in eye imaging

M. Wojtkowski and A. Kowalczyk

Institute of Physics, Nicholas Copernicus University, Grudziadzka 5/7, Pl- 87-100 Torun, Poland

R. Leitgeb and A. F. Fercher

Institute of Medical Physics, University of Vienna, Waehringer Strasse 13, A-1090 Vienna, Austria

Received March 20, 2002

We demonstrate a new implementation of complex spectral optical coherence tomography (OCT) in biomedical
imaging. By reconstruction of both amplitude and phase we are able to use the negative and positive optical
path differences to get images of objects of considerable thickness. An accompanying reduction of coherent
noise improves the quality of the images. The property of the complex spectral OCT that permits the mea-
surement range to be increased and permits the simultaneous use of phase and amplitude in spectral systems
was not described previously. To show the potential of this technique we measured an anterior chamber of
a porcine eye in vitro. © 2002 Optical Society of America
OCIS codes: 170.4500, 070.2590, 170.4470, 170.3880, 170.0110, 090.2880.

The detection of light scattered from an object that is tered as a function of optical frequency and is called a
illuminated by light of low temporal but high spatial cross-spectral density function, GUU 共n兲:
coherence is the basis of optical coherence tomo-
graphy (OCT). During the last decade of the 20th X
century OCT was developed and became an important GUU 共n兲 苷 Grr 共n兲 1 Gnm 共n兲
n
diagnostic technique in medicine.1 The most ad- ( )
vanced application of OCT is in ophthalmology, for X
which it has proved useful for retinal imaging and 1 2 Re Gnm 共n兲exp关22pin共tn 2 tm 兲兴
quantification of the thicknesses of nerve fiber layers nfim

as well as for high-resolution imaging of the anterior ( )

chamber of the eye.2 There are two implementations X
1 2 Re Gnr 共n兲exp关22pin共tn 2 tr 兲兴 . (1)
of OCT. Historically the first was the time-domain n
version1; the so-called frequency-domain systems were
developed later.3 – 7 Among the frequency-domain The f irst two terms in Eq. (1) are intensities from the
techniques, spectral OCT is the most mature. Tomo- reference mirror and scattering centers within the ob-
grams of the human retina in vivo obtained by this ject. The next term is parasitic (so-called autocorrela-
technique were reported recently.8 Despite its several tion) and is associated with the mutual interference of
advantages, such as direct access to spectral informa- all elementary waves scattered within an object. The
tion,9 elimination of an optical delay line for depth last term contains direct information on the positions
scanning,10 and easy application to surface analy- of scattering centers along the penetration beam with
sis,11 this technique has attracted less attention than respect to the constant position of the reference mir-
the well-known time-domain version. The reason for ror. Fourier transformation of the cross-spectral den-
this is that the image obtained by spectral OCT is sity function yields
inherently disturbed by artifacts.
In 1999 Fercher et al.12 proposed and presented pre- X
liminary results of a complex spectral OCT technique FT21 兵GUU 共n兲其 苷 Grr 共t兲 1 Gnn 共t兲
based on phase reconstruction. Here we extend the n
complex spectral OCT to overcome some drawbacks of X X
the spectral methods and apply the proposed technique 1 G关t 1 共tm 2 tn 兲兴 1 G关t 2 共tm 2 tn 兲兴
nfim nfim
to get an image of the anterior chamber of the eye
in vitro. X X
1 G关t 1 共tr 2 tn 兲兴 1 G关t 2 共tr 2 tn 兲兴 , (2)
The basic technique of OCT is based on partial co- n n
herence interferometry. A beam from a light source is
split into two beams: one penetrates an object along where first-order electric f ield correlation function G is
the z axis and is ref lected back along the z axis from def ined as G共tn 兲 ⬅ 具u共t兲uⴱ 共t 1 tn 兲典.
the nth ref lecting layer, and the second beam is re- Spectral density is a real function; therefore its
f lected from the reference mirror. These two beams Fourier transform is Hermitian, and the recon-
are delayed by tn 苷 zn 兾c and tr 苷 zr 兾c, respectively. structed image is symmetrical about t 苷 0. It leads
In spectral systems the resultant interference is regis- to two mirror images, which might be superimposed.
0146-9592/02/161415-03$15.00/0 © 2002 Optical Society of America
1416 OPTICS LETTERS / Vol. 27, No. 16 / August 15, 2002

Additionally, terms that originate from the mutual In Fig. 2 we present a comparison of tomographic
interference between waves ref lected within the object images of the porcine eye’s anterior chamber obtained
[third and fourth terms of Eq. (2)] mix with those with the aid of spectral OCT and of complex spectral
that provide direct information on the position of the OCT (left and right, respectively). The f irst method
ref lecting surfaces (the last two terms). In effect, is insensitive to the sign of optical path difference
the resultant image is obscured by artifacts and by d, i.e., the range of negative d is symmetrical to the
coherent noise. To eliminate these shortcomings we range of positive d. Therefore there are overlapping
used a complex method that allows the complex form mirror images. Furthermore, close to zero, optical
of the last term in Eq. (1) to be reconstructed: path difference parasitic autocorrelation terms are
X located that effectively reduce the signal-to-noise
Ĉnr 共n兲 苷 Gnr 共n兲exp关22pin共tn 2 tr 兲兴 ratio. Hence the recognition of morphological details
n is severely impaired. In contrast, the complex tech-
苷 C共n兲exp关if共n兲兴 . (3) nique is free of any ambiguities caused by overlapping
mirror images as well as of the presence of disturbing
terms. The useful depth range doubles to 0.4 cm.
To calculate the complex function Ĉnr 共n兲 we adopted The cornea, iris, lens, and some post-mortem changes
a f ive-frame method.13 From the f ive consecutive in eye structure are clearly visible.
measurements of the spectra GUU 共n兲 共n兲 taken with a To illustrate the signal-to-noise ratio enhancement
phase shift in increments of p兾2, the phase f共n兲 and we present a comparison of two axial reconstructions
the amplitude C共n兲 are calculated according to the obtained by spectral and complex spectral OCT (Fig. 3).
following formulas: Both parts of the f igure are normalized to the signal at
the border of iris. Application of the complex method
2关G 共2兲 2 G 共4兲 兴 , reduces the noise level ⬃2.5 times. Therefore the
f 苷 arctan (4a)
2G 共3兲 2 G 共5兲 2 G 共1兲
1
C苷 共 兵2关G 共2兲 2 G 共4兲 兴其2 1 关2G 共3兲 2 G 共5兲 2G 共1兲 兴2 兲1兾2 , (4b)
4

where G 共n兲 苷 GUU 关f共n兲 1 n共p兾2兲兴 , n 苷 1, 2, . . . , 5.

Finally,
X
FT21 兵C共n兲exp关if共n兲兴其 苷 G关t 2 共tr 2 tn 兲兴 (5)
n

yields structural information on the object that is free

from parasitic terms as well as from a mirror image.
As a result the accessible depth range expands by a
factor of 2.
The OCT instrument is based on a Michelson in- Fig. 1. Schematic of the complex spectral OCT device.
terferometer setup (Fig. 1) with a 50兾50 cube beam SLD, superluminescent diode.
splitter. The light source is a superluminescent diode
emitting at 810 nm with a 20-nm FWHM and a maxi-
mum output power of 2 mW. The detection unit con-
sists of a diffraction grating (1800 grooves兾mm) and a
16-bit CCD camera (64-kHz sampling rate of an ana-
log-to-digital converter, 1024 3 128 pixels, full vertical
binning, and a full well capacity of 106 e2 ). To intro-
duce a phase shift into the reference beam we place the
mirror upon a piezo translator.
The transverse scanner and the cornea were placed
at the focal point of the object lens, OL 共 f 苷 80 mm兲.
This design ensures that the direction of the beam is
insensitive to the scanner angle. The optical system
produces a focal spot size of Dx ⬃ 35 mm and a focal
depth of 3 mm. Spectra from the CCD camera were
transferred to a personal computer, analyzed, and dis-
played by a program written in LabView.
The measurements were performed on porcine eyes
in vitro. The region of interest was the anterior cham-
ber. At each transverse scanning position a set of five Fig. 2. Tomographic images of porcine eye’s anterior
spectra was recorded. The images consisted of 900 chamber in vitro obtained by left, spectral OCT and right,
horizontal lines. The reference mirror’s position is in- complex spectral OCT. LS, lens; IR, iris; CR, cornea;
dicated by arrows in the f igures that follow. P, post-mortem changes.
 August 15, 2002 / Vol. 27, No. 16 / OPTICS LETTERS 1417

stability of the object. An object displacement of as

little as l兾10 during a f ive-frame cycle completely
precludes faithful reconstruction. Therefore for
in vivo measurements the system has to collect all five
spectra simultaneously.
The authors thank Tomasz Bajraszewski and Iwona
Gorczynska for help and J. Anuszkiewicz, J. Kowal-
ski, H. Schiller, and J. Wojtylewski of the mechanical
workshop of the Nicolas Copernicus University Physics
Institute for their professional support. This research
was financed by grant 502-F from Nicolas Coperni-
cus University. M. Wojtkowski’s e-mail address is
[email protected].

References
1. D. Huang, E. A. Swanson, C. P. Lin, J. S. Schu-
man, 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. F. Fercher, C. K. Hitzenberger, G. Kamp, and S. Y.
El-Zaiat, Opt. Commun. 117, 43 (1995).
4. A. Eigensee, G. Hausler, J. M. Hermann, and M. W.
Fig. 3. Axial reconstructions obtained by (a) spectral and Lindner, Proc. SPIE 2925, 169 (1996).
(b) complex spectral OCT. Both plots are normalized to 5. F. Lexer, C. K. Hitzenberger, A. F. Fercher, and
the signal at the border of the iris. M. Kulhavy, Appl. Opt. 36, 6548 (1997).
6. U. H. P. Haberland, V. Blazek, and H. J. Schmitt,
J. Biomed. Opt. 3, 259 (1998).
7. I. Zeylikowich, A. Gilerson, and R. R. Alfano, Opt. Lett.
resultant signal-to-noise ratio is increased by 7.5 dB; 23, 1797 (1998).
the f inal dynamic range of the complex spectral 8. M. Wojtkowski, R. Leitgeb, A. Kowalczyk, and
method is 74 dB. A. Fercher, Proc. SPIE 4619 (to be published).
The benefits of the method presented here are the 9. R. Leitgeb, M. Wojtkowski, C. K. Hitzenberger,
absence of overlapping mirror images, doubled depth M. Sticker, A. Kowalczyk, and A. F. Fercher, Opt. Lett.
range, higher signal-to-noise ratio, and simple inter- 25, 820 (2000).
pretation of thick biological samples. The price to be 10. A. F. Zuluaga and R. Richards-Kortum, Opt. Lett. 24,
519 (1999).
paid for these improvements is a f ivefold increase in 11. G. Hausler and M. W. Lindner, J. Biomed. Opt. 3, 21
the measurement time. In this system it takes about (1998).
2 min to collect the complex-method data. However, 12. A. F. Fercher, R. Leitgeb, C. K. Hitzenberger,
an upgrade of the analog-to-digital converter to a H. Sattmann, and M. Wojtkowski, Proc. SPIE 3564,
1-MHz pixel rate shortens this time to 6 s. A sec- 173 (1999).
ond disadvantage is that this method requires high 13. J. Schmit and K. Creath, Appl. Opt. 34, 3610 (1995).