JOURNAL OF BIOMEDICAL OPTICS 3(1), 45–54 (JANUARY 1998)

SIGNAL AND RESOLUTION ENHANCEMENTS IN

DUAL BEAM OPTICAL COHERENCE
TOMOGRAPHY OF THE HUMAN EYE
Angela Baumgartner, Christoph K. Hitzenberger, Harald Sattmann, Wolfgang Drexler,
and Adolf F. Fercher
Institute of Medical Physics, University of Vienna, Währinger Straße 13, A-1090 Vienna, Austria
(Paper JBO/IB-011 received Aug. 21, 1997; revised manuscript received Nov. 11, 1997; accepted for publication Nov. 12, 1997.)

ABSTRACT
In the past 10 years, a dual beam version of partial coherence interferometry has been developed for mea-
suring intraocular distances in vivo with a precision on the order of 0.3 to 3 mm. Two improvements of this
technology are described. A special diffractive optical element allows matching of the wavefronts of the
divergent beam reflected at the cornea and the parallel beam reflected at the retina and collimated by the optic
system of the eye. In this way, the power of the light oscillations of the interfering beams incident on the
photodetector is increased and the signal-to-noise ratio of in vivo measurements to the human retina is
improved by 20 to 25 dB. By using a synthesized light source consisting of two spectrally displaced super-
luminescent diodes with an effective bandwidth of 50 nm, and by compensating for the dispersive effects of
the ocular media, it was possible to record the first optical coherence tomogram of the retina of a human eye
in vivo with an axial resolution of ;6 to 7 mm. This is a twofold improvement over the current technology.
© 1998 Society of Photo-Optical Instrumentation Engineers. [S1083-3668(98)01601-3]

Keywords partial coherence interferometry; optical coherence tomography; resolution; dispersion; diffrac-
tive optics; human eye; retina.

1 INTRODUCTION ods are limited in their longitudinal resolution. An-

other drawback of the ultrasound technique is the
Tomographic images of retinal structures in vivo as
well as the quantitative and precise measurement mechanical contact needed between the eye and the
of the thickness of various fundus layers at arbi- instrument.
trary reproducible measurement positions are clini- In the past 10 years a new interferometric tech-
cally essential in modern ophthalmology for diag- nique for length measurement, partial coherence in-
nosis, monitoring, and therapy control of ocular terferometry (PCI), has been developed. The first
diseases. Glaucoma, macular degeneration, edema, applications of this technique to measure the axial
and many other optic neuropathies are accompa- length in human eyes in vivo were reported by
nied by alterations of the retinal thickness or the Fercher and Roth11 and Fercher, Mengedoht, and
retinal nerve fiber layer thickness. Werner.12 Since a classic interferometric setup is ex-
Retinal imaging of high lateral resolution can be tremely sensitive to longitudinal motion of an ob-
obtained by scanning laser ophthalmoscopy, and ject, a special dual beam interferometry technique
with improved depth resolution, by confocal scan- was used to eliminate the influence of longitudinal
ning laser tomography.1–4 However, the axial reso- eye motion. An electronic fringe detection system
lution of these methods is limited by the pupil di- using the Doppler principle was introduced to re-
ameter and ocular aberrations. A very fast and duce the measuring time.13,14 This method of dual-
precise method for measuring the thickness of the beam PCI enables high-precision biometry of the
retinal nerve fiber layer is scanning laser polarime-
human eye in vivo, measuring corneal thickness,14,15
try, a technique that measures the retardation of
anterior chamber depth,15,16 lens thickness,15,16 the
laser light passing through the birefringent retinal
nerve fiber layer.5,6 axial eye length,13 the thickness of the retina,13,17
Retinal thickness can be measured by ultrasonog- and various fundus layers very accurately at arbi-
raphy using high frequencies7,8 and by a noninva- trary positions of the human eye fundus. This tech-
sive optical technique employing an extension of nique has been further extended to optical coher-
slit-lamp biomicroscopy.9,10 However, both meth- ence tomography (OCT), a new imaging modality

Address all correspondence to Angela Baumgartner. Tel: 431 40480333;

Fax: 431 4024030; E-mail: [email protected] 1083-3668/98/$10.00 © 1998 SPIE

JOURNAL OF BIOMEDICAL OPTICS d

JANUARY 1998 d
VOL. 3 NO. 1 45
Downloaded From: https://www.spiedigitallibrary.org/journals/Journal-of-Biomedical-Optics on 10 Oct 2023
Terms of Use: https://www.spiedigitallibrary.org/terms-of-use
 BAUMGARTNER ET AL.

capable of obtaining two-dimensional sections of crease in resolution and interference fringe contrast.
the human retina.18–21 For a recent overview of We have recently demonstrated that this effect be-
OCT, see Ref. 22. Recent developments in OCT are comes dominant if measurements through the dis-
concerned with improving axial resolution23 and persive media of the eye to the retina are performed
with the development of an alternative OCT with source bandwidths larger than about 25
scheme that enables the recording of en face nm.29,30
images.24,25 In order to achieve optimum resolution of OCT
The measurement of intraocular distances by the by applying a light source with a broad emission
dual beam PCI technique is based on the detection spectrum, the dispersion of the object to be mea-
of interference fringes caused by light beams re- sured (in this case the ocular media) has to be com-
flected at the various intraocular interfaces whose pensated for. This can be achieved by placing a dis-
distances are to be measured. In order to be de- persive element in the reference arm of the
tected, the spacing of the interference fringes must Michelson interferometer that causes the same
be wider than the effective diameter of the photo- group dispersive effect as the ocular media.30 It is
detector. Since the positions of the intraocular inter- the second purpose of this paper to demonstrate
faces are measured with respect to the anterior cor- the improvement in resolution that is obtained by
neal surface as the reference surface, a concentric compensating for the object dispersion and using a
ring-shaped interference fringe system is formed light source of 50 nm bandwidth consisting of two
due to the dual beam interference of the divergent spectrally displaced SLDs. Furthermore, we present
beam reflected at the anterior corneal surface and the first OCT image recorded with this technique in
the almost parallel beam reflected at the retina and a human eye in vivo with an axial geometrical reso-
refracted by the lens and the cornea.11 Owing to the lution of about 6 mm.
wavefront mismatch of the two interfering beams,
the corresponding interference fringes are narrow
and a photodetector with a small area (50 mm diam- 2 METHODS
eter) has to be used. This limits the intensity inci- 2.1 DUAL BEAM PARTIAL COHERENCE
dent on the photodetector and hence the signal-to- INTERFEROMETER
noise ratio (S/N) obtainable with the motion-
insensitive dual beam PCI technique. In order to The theory and experimental details of measuring
improve the S/N, the PCI technique has now been intraocular distances by dual beam PCI was de-
further extended by using a diffractive optical ele- scribed in previous papers.13,14 Therefore, only a
ment that compensates for this wavefront brief summary is presented here.
mismatch.26,27 By the use of this element, both of As depicted in Figure 1, a superluminescent di-
the interfering beams reflected at the cornea and the ode emits a light beam of high spatial coherence but
retina will be collimated when they recombine at short coherence length l c which illuminates an ex-
the photodetector so that all the optical power re- ternal Michelson interferometer. The interferometer
flected by the eye may be collected. In this way, a splits the beam into two coaxial components that
significant improvement of the signal-to-noise ratio have a phase difference corresponding to twice the
is achieved. This paper demonstrates the signal im- interferometer arm length difference d. This dual
provement achieved in this way and presents the beam illuminates the eye, where each subcompo-
first retinal in vivo tomograms recorded in a human nent is reflected at the various intraocular interfaces
eye with this technique. separating regions of different refractive indices. If
In OCT, longitudinal resolution depends on the the difference in arm length of the Michelson inter-
spectral bandwidth of the light source used and on ferometer equals (within l c ) one of the intraocular
the dispersion of the media to be measured. In non- distances to be measured, the beams reflected at the
dispersive media, the resolution is approximately respective intraocular interfaces will interfere. Dur-
equal to the coherence length of the light used, ing the measurement, one of the mirrors of the
which is inversely proportional to the width of the Michelson interferometer is moved by a stepper
emission spectrum.28 Hence a broad emission spec- motor with a constant speed n, which causes a Dop-
trum yields a short coherence length and conse- pler shift of the light frequency of the correspond-
quently a good resolution. With currently used su- ing beam. In the case of a coincident path length,
perluminescent diodes (SLDs), resolutions on the the intensity of the corresponding interference pat-
order of 10 to 15 mm are achieved. By using a tern is modulated by the Doppler frequency f D
broadband Ti:Al2O3 source, a resolution of a few 52 n /l. The superimposed reflected beams are
microns was recently demonstrated in biological then focused onto the photodetector, amplified, and
tissue with a total thickness of a few hundred filtered by a bandpass filter that transmits only sig-
microns.23 However, if the tissue under investiga- nals with f D . The envelope of this signal is re-
tion is dispersive, the coherence envelope of the sig- corded as a function of the interferometer’s arm
nal broadens due to the different velocities of the length difference d with a personal computer. The
various wavelengths of the broad emission spec- signal curves recorded in this way are called PCI
trum in the dispersive medium, leading to a de- scans, or optical A-scans, and contain characteristic

46 JOURNAL OF BIOMEDICAL OPTICS d

JANUARY 1998 d
VOL. 3 NO. 1
Downloaded From: https://www.spiedigitallibrary.org/journals/Journal-of-Biomedical-Optics on 10 Oct 2023
Terms of Use: https://www.spiedigitallibrary.org/terms-of-use
 SIGNAL AND RESOLUTION ENHANCEMENTS IN DUAL BEAM OCT

Fig. 1 Schematic of the improved dispersion compensated partial coherence interferometer with the modified illumination unit employing a
diffractive optical element and the implementation of a fundus camera. BSC, beamsplitter cube; BS, beamsplitter; IOD, intraocular distance.

signal peaks. From the positions of these peaks on to-noise ratio due to the wavefront mismatch of the
the d axis, the respective optical intraocular dis- two interfering beams reflected at the anterior cor-
tances (IOD) can be determined: IOD5d6l c . Using neal surface and the retina. If the probing beam is
a cursor readout to determine the signal peak posi- collimated, the light reflected at the retina will be
tion in the PCI scan provides a precision even better collimated by the optical elements of the human
than l c . To convert the optical intraocular distance eye (in case of emmetropic subjects) while the beam
into the appropriate geometrical distance, it must reflected at the cornea will be divergent. If the prob-
be divided by the group refractive index n g of the ing beam is focused at the cornea, the reflected
respective medium.13 beam from the cornea will again be collimated by
Simultaneous with the optical A-scans of the PCI the focusing lens, but the light reflected at the retina
system, the eye under investigation can be exam- will be divergent. In either case, a concentric ring-
ined by a fundus camera.27 As depicted in Figure 1, shaped interference fringe system as depicted in
the laser light illuminates the eye via a short-wave Figure 2 is formed.12 The spacing of these circular
pass beamsplitter that reflects light in the near- interference fringes becomes smaller at the periph-
infrared (IR) region (l5760 to 850 nm) and trans- ery of the fringe pattern. Ideally, the photodetector
mits the white light of the fundus camera in front of should be aligned with the center of the fringe sys-
the eye (l5450 to 650 nm). For this purpose, the tem. Because ocular motion cannot be completely
pupil of the eye to be measured has to be dilated by avoided during in vivo measurements, this ideal
a mydriaticum. First the eye is aligned with the in- alignment cannot be perfectly maintained. There-
cident laser beam; then the fundus camera is ad- fore, the diameter of the photodetector has to be
justed to its appropriate working distance in order small enough that even in the case of slight mis-
to get the fundus image showing the position of the alignments, only a single fringe is imaged onto the
incident laser spot on the retina. The image of the detector surface. A detector diameter of ;50 m m
fundus camera can be observed on-line via a visible turned out to be optimal.13 In this case, however,
and near-IR sensitive CCD camera that is connected the total light power collected by the detector is
to a personal computer via a frame grabber card to rather low, and therefore the S/N is poor.
store the appropriate fundus images and enable the In order to overcome this drawback, we used a
exact localization of the measurement position. special diffractive optical element for wavefront
matching.26,27 This sort of Fresnel’s zone lens is
2.2 WAVEFRONT MATCHING implemented in front of the eye in order to focus
The major drawback of the dual beam version of part of the incident laser beam on the vertex of the
partial coherence interferometry is the poor signal- cornea (f570 mm). The other collimated parallel

JOURNAL OF BIOMEDICAL OPTICS d

JANUARY 1998 d
VOL. 3 NO. 1 47
Downloaded From: https://www.spiedigitallibrary.org/journals/Journal-of-Biomedical-Optics on 10 Oct 2023
Terms of Use: https://www.spiedigitallibrary.org/terms-of-use
 BAUMGARTNER ET AL.

Fig. 2 Schematic of the concentric ring-shaped interference fringe system due to the dual beam interference of the divergent corneal
reflection and the parallel retinal reflection. The photodetector has to be aligned approximately coaxially to the center of the interference
fringe system.

part of the beam passes through uninfluenced (Fig- alignment of the DOE in front of the eye takes only
ure 3). This part of the beam will be focused onto a few seconds and is easy to perform. Once the eye
the retina (in case of emmetropic subjects) by the under investigation is adjusted properly, the align-
optical elements of the human eye. ment is not more sensitive to movements of the eye
The beams remitted from the anterior corneal sur- than with the previous arrangement.
face and the retina will thereby both be converted
into parallel beams when they pass again through 2.3 LIGHT SOURCE
the diffractive optical element (DOE) on their way
As already mentioned, the resolution of PCI is ap-
back to the detection unit. Assuming a reflectivity
proximately equal to the coherence length of the
of r 52% of the anterior corneal surface and r
light source used. Assuming a Gaussian shape of
51% of the retina (corresponding to data recorded
the emission spectrum, the full-width at half-
with an experimental laboratory setup), the DOE
maximum (FWHM) round-trip coherence length l c
has been designed to focus 40% of the intensity of
in a PCI system with a Michelson interferometer
the incident laser beam on the vertex of the cornea
can be calculated23:
and to let 60% pass through as a collimated parallel
beam. Under these circumstances, the theoretical
2 ln~ 2 ! l 20
visibility of the interference fringes should be ap- l C5 3 , (1)
proximately 1. Consequently, two collimated paral- p Dl
lel beams will interfere and the light power oscilla- where l 0 is the center wavelength of the source and
tions in the corresponding interferograms will be Dl is its FWHM spectral width.
much stronger than those of the narrow interfer- Within the scope of this work, two different light
ence fringes obtained without that technique. This sources have been used: A single superluminescent
improves the signal-to-noise ratio considerably. The diode (EG&G C86142E, EG&G Optoelectronics,
Canada) with l 0 5830 nm and Dl526 nm. Accord-
ing to Eq. (1), the coherence length l c 511.7 m m. For
demonstration of high-resolution OCT, a synthe-
sized light source generated by superimposing two
superluminescent diodes (EG&G C86142E, EG&G
Optoelectronics, Canada) with different center
wavelengths (l 015830 nm, Dl 1 526 nm, l 02
5855 nm, and Dl 2 525 nm) has been used. Owing
to a beat effect,31–33 these two combined light
sources have an effective spectral width of Dl eff
Fig. 3 Schematic of the principle of the diffractive optical element.
550 nm. The coherence length of this synthesized
Part of the incident laser beam will be focused on the vertex of the
cornea, the other collimated parallel part of the beam will pass light source l c '8 m m.
through uninfluenced. This part of the beam will be focused onto
the retina (in case of emmetropic subjects) by the optical elements 2.4 DISPERSION COMPENSATION
of the human eye. Both of the beams reflected at the cornea and the
retina will be collimated when they pass through the DOE on their If one of the beams in a PCI system travels through
way back to the detection unit. a dispersive medium while the other travels

48 JOURNAL OF BIOMEDICAL OPTICS d

JANUARY 1998 d
VOL. 3 NO. 1
Downloaded From: https://www.spiedigitallibrary.org/journals/Journal-of-Biomedical-Optics on 10 Oct 2023
Terms of Use: https://www.spiedigitallibrary.org/terms-of-use
 SIGNAL AND RESOLUTION ENHANCEMENTS IN DUAL BEAM OCT

through air (or if the path lengths through To obtain OCT images, several optical A-scans
dispersive media differ in length), the coherence are recorded at different equidistant angles to the
envelope of the optical A-scans broadens and vision axis. The angle increment was 0.5 deg, corre-
the resolution decreases. If the length of a disper- sponding to length increments of ;150 m m on the
sive medium in one of the interferometer arms is retina. The intensity values of the optical A-scans
L and the group dispersion of the medium are converted into pixel colors (logarithmic scale)
is dn g /dl, the width of the coherence envelope, and mounted to form a two-dimensional false color
after double passing through the medium, can image. The pixels between the individual scanning
be calculated by29,30: directions are obtained by linear interpolation.

F S DG
2 1/2 2.6 IN VIVO MEASUREMENTS AND LASER
dn g
l C,m 5 l 2C 1 3L3Dl . (2) SAFETY
dl
In vivo measurements were performed in healthy,
The mean group dispersion of the ocular media is volunteer subjects after full informed consent was
approximately 21.831025 /nm34 in the near infra- obtained. Optical A-scans were performed over a
red. Assuming a mean axial eye length of 24 mm, maximum length increment of 3 mm, which takes
the signal width, after passing through the eye me- approximately 0.4 s. Performing eight scans at most
dia to the retina and returning, can be calculated by for each measurement direction for further im-
Eq. (2) to be ;16.2 m m in the case of the single SLD provement of S/N by averaging, the maximum
with Dl526 nm and ;22.5 m m in the case of a time of continuous illumination of one point on the
combination source with Dl eff550 nm. This means retina is about 4 to 5 s. During this time, the eye is
that the resolution in the case of a larger source illuminated with a laser light power of approxi-
bandwidth is not improved, but degraded. In order mately 200 mW or an intensity of 520 m W/cm2 (av-
to achieve the optimum resolution obtainable with eraged over a 7-mm aperture). This is permitted for
the broadband light source, the dispersion of the 25 min for a wavelength of l5830 nm in the case of
ocular media has to be compensated for. This can intrabeam viewing.36 Since only 60% of the incident
be achieved by placing a dispersion-compensating light power is focused at the retina, the laser safety
element in the longer arm of the external Michelson limit can be easily met.
interferometer (Figure 1). This element must fulfill
the condition 3 RESULTS
3.1 IMPROVEMENT OF THE SIGNAL-TO-
L el S D
dn g
dl
5L ob
el
S D
dn g
dl ob
, (3)
NOISE RATIO
In order to quantify the improvement in the signal-
to-noise ratio from the use of the diffractive optical
where the indices el and ob refer to the compensa- element, measurements were taken on a model eye
tion element and the object, respectively. As a dis- that consisted of a biconvex lens (f530 mm) and a
persion compensation element we used a plane- mirror (simulating the fundus of the eye) at a dis-
parallel plate of BK7 optical glass with L el512 mm tance approximately 30 mm from the lens. Various
and (dn g /dl) el'431026 /nm at a wavelength of neutral density filters were put between the lens
l>840 nm. 35 This element fulfills Eq. (3) to within a and the mirror in order to reduce the intensity re-
few percent and therefore compensates for most of flected at the mirror with respect to that reflected at
the dispersive effects of the ocular media.30 the anterior lens surface. The model eye fundus
measurements were performed under the same ex-
perimental settings as was the case for in vivo mea-
2.5 TOMOGRAPHIC MEASUREMENTS surements. A single SLD (l 0 5830 nm, Dl526 nm)
For tomographic recordings, a separate fixation was used to illuminate the model eye with a light
light was installed in front of the eye under inves- power of approximately 200 mW. According to the
tigation, which encloses an adjustable vertical and model eye measurement, the application of the
horizontal angle with the measuring beam and is DOE increased the signal-to-noise ratio by about 40
collimated in order to make the subject look toward to 45 dB compared with previous measurements
infinity. This allows the rotation of the eye to be without that element.
measured in order to adjust a defined arbitrary In vivo measurements have been performed at
angle between the vision axis and the measuring various positions along the papillomacular axis of
direction. Once the eye looks at the fixation target, the fundus, within the angular range of 5 deg tem-
it is aligned with the measuring beam so that the poral to 20 deg nasal. As an example, Figure 4
two image points of the anterior corneal surface shows an optical A-scan obtained in a healthy hu-
and the retina are centered and become confocal on man eye in vivo at an angle of 4.5 deg nasal to the
the detector. For observation purposes, the detec- vision axis. The signal intensity (ordinate), which is
tion plane is imaged onto a CCD camera where the proportional to the interference fringe contrast, is
confocal alignment can be monitored. plotted as a function of the interferometer’s arm

JOURNAL OF BIOMEDICAL OPTICS d

JANUARY 1998 d
VOL. 3 NO. 1 49
Downloaded From: https://www.spiedigitallibrary.org/journals/Journal-of-Biomedical-Optics on 10 Oct 2023
Terms of Use: https://www.spiedigitallibrary.org/terms-of-use
 BAUMGARTNER ET AL.

clearly in a comparison with Figure 4 that the

signal-to-noise ratio is improved significantly by
the DOE (on the order of 20 to 25 dB). In Figure 4
the signals from the different fundus layers can be
clearly differentiated from noise although the step-
per motor and the measuring electronics of the PCI
system using the DOE are not yet optimized.

3.2 RESOLUTION IMPROVEMENT IN OCT

IMAGING
To demonstrate the improvement in axial resolu-
tion, we recorded three OCT tomograms of the
same area of the same human eye in vivo (Figure
6—color plate). A horizontal section through the
Fig. 4 Optical A-scan recorded in a human eye in vivo at an angle fovea centralis, from 5 deg temporal to 5 deg nasal
of 4.5 deg nasal to the vision axis using the diffractive optical
element. The signal intensity, which is proportional to the interfer-
is shown. Figure 6(a) was recorded with a single
ence fringe contrast, is plotted as a function of the interferometer’s SLD (Dl526 nm) without dispersion compensa-
arm length difference d . Several peaks from different retinal micro- tion. Figure 6(b) was recorded with the same light
structural layers can be differentiated very clearly from noise. Peak source, but with dispersion compensation, and Fig-
1 probably arises from the inner limiting membrane, peak 2 prob- ure 6(c) was recorded with the combination source
ably originates from the retinal pigment epithelium. The distance
between peaks 1 and 2 corresponds to the optical thickness of the
(Dl eff550 nm) and with dispersion compensation.
retina (391 mm). The peaks behind peak 2 are probably caused by Several microstructural layers can be observed in
choroidal structures. these figures: the inner limiting membrane, the
foveal depression, the nearly transparent retina,
and highly reflective layers at the posterior side of
length difference d (abscissa). In this case, d was the retina. The first of these layers is probably the
scanned over a range of 2 mm, from 33.5 to 35.5 inner limiting membrane (ILM) that separates the
mm. Several peaks from different retinal micro- retinal structures from the vitreous. In Figure 6(a),
structural layers can be differentiated very clearly two different layers approximately 390 mm behind
from noise. Peak 1 probably arises from the inner the ILM can be observed; however, these partly
limiting membrane, the closest layer to the vitreous. overlap (the color coding on a logarithmic scale
Peak 2 originates probably from the retinal pigment tends to reduce the visibility of closely related de-
epithelium. The distance between peaks 1 and 2 tails). The axial resolution in this figure is
corresponds to the optical thickness of the retina ;16.5 m m optical distance (measured from optical
(391 mm). The peaks behind peak 2 are probably A-scans, see Figure 7) or 12 mm geometrical dis-
caused by choroidal structures. tance, assuming a group refractive index of ;1.4
To demonstrate the improvement in the signal-to- for the retina. In Figure 6(b), the resolution im-
noise ratio achieved by the DOE, Figure 5 shows an proved to ;13.5 m m optical distance (;9.5 m m
optical A-scan carried out with the previous version geometrical distance) because of dispersion com-
of the PCI without the application of the diffractive pensation. The three layers observed behind the
optical element. This measurement was performed ILM are much better separated. Finally, the best
at a horizontal angle of 10 deg nasal. It can be seen resolution is obtained with the broadband synthe-
sized light source with dispersion compensation
[Figure 6(c)]. The resolution, as measured from op-
tical A-scans, is ;9 to 10 mm optical distance or ;6
to 7 mm geometrical distance. The three layers pos-
terior to the retina are now imaged as three narrow,
well-separated bands.
To elucidate the improvement in resolution, we
have taken two individual optical A-scans from the
data sets used to synthesize the tomograms of Fig-
ures 6(a) and 6(c) and superimposed them. The re-
sult is shown in Figure 7. These A-scans correspond
to a measuring angle of 1.5 deg nasal to the vision
axis and cover an optical distance of ;100 m m. The
dashed line shows the signal recorded with the
single SLD without dispersion compensation; the
solid line is the signal obtained with the combina-
Fig. 5 Optical A-scan recorded in a human eye in vivo without tion source and with dispersion compensation. The
using the DOE at an angle of 10 deg nasal to the vision axis. improvement of the resolution is obvious: the

50 JOURNAL OF BIOMEDICAL OPTICS d

JANUARY 1998 d
VOL. 3 NO. 1
Downloaded From: https://www.spiedigitallibrary.org/journals/Journal-of-Biomedical-Optics on 10 Oct 2023
Terms of Use: https://www.spiedigitallibrary.org/terms-of-use
 SIGNAL AND RESOLUTION ENHANCEMENTS IN DUAL BEAM OCT

Fig. 6 Retinal tomograms recorded in a human eye in vivo demonstrating different longitudinal resolutions, depending on the spectral
properties of the light source used. A horizontal section of the fundus across the fovea from 5 deg temporal to 5 deg nasal is shown
corresponding to a width of approximately 3 mm on the retina. The vertical scale indicates the optical distance to the anterior corneal
surface in millimeters. (a) Single SLD with Dl526 nm without dispersion compensation. (b) Single SLD with Dl526 nm with dispersion
compensation. (c) Synthesized light source of two spectrally displaced SLDs with Dl eff550 nm and with dispersion compensation.

JOURNAL OF BIOMEDICAL OPTICS d

JANUARY 1998 d
VOL. 3 NO. 1 51
Downloaded From: https://www.spiedigitallibrary.org/journals/Journal-of-Biomedical-Optics on 10 Oct 2023
Terms of Use: https://www.spiedigitallibrary.org/terms-of-use
 BAUMGARTNER ET AL.

flected by the eye may be collected by the photode-

tector. Hence the intensity of the interfering beams
is much higher, yielding a better S/N. By using a
single diffractive optical element, we were able to
improve the S/N by about 20 to 25 dB in in vivo
measurements. Compared with the rather complex
spatial filtering method used by Chen et al.38 our
solution has the advantage of consisting of a single,
lightweight optical element. The lens with the cen-
tral hole,38 on the other hand, has the drawback that
only a narrow parallel beam can pass through the
hole, which restricts the possible resolution on the
retina because of the small aperture angle and con-
sequently the rather large diameter of the
Fig. 7 Optical A-scans recorded with a single SLD without disper- diffraction-limited probing beam on the retina.
sion compensation [dashed line, corresponding to Figure 6(a)], The axial resolution of OCT systems currently
and with the combination source with dispersion compensation used for retinal imaging is claimed to be on the
[solid line, corresponding to Figure 6(c)] at a measuring angle of
1.5 deg nasal. The improvement in resolution can be clearly ob- order of 10 to 15 mm.20–22 Systems with higher reso-
served. lution using broadband Ti:Al2O3 sources have so
far been used only for ranging and imaging of tis-
sues with a total thickness of a few hundred
FWHM of the first peak is ;16.5 m m (optical dis- micrometers.23,39 These systems were carefully de-
signed to compensate for the dispersion of the op-
tance) in the case of the dashed curve and ;10 m m
tical elements within the measurement and the ref-
in the case of the solid curve. This corresponds to
erence arm of the interferometer. The additional
geometrical distances of ;13 and 7 mm, respec-
dispersive effects of a few hundred micrometers of
tively. The second signal peak shows approxi-
tissue have only a negligible effect on the signal
mately the same factor of improvement. Because
width and hence the resolution. However, if mea-
the two scans were recorded in different sessions,
surements are performed through the relatively
the overall signal shape is slightly different (it is not
thick ocular media to the retina, the dispersive ef-
possible to maintain exactly the same measuring
position on the retina in two successive sessions). fects of these media have to be compensated for if
broadband light sources are used. If a light source
with a bandwidth greater than that of a single SLD
4 DISCUSSION (Dl;25 nm) is used, the width of the coherence
One of the drawbacks of the dual beam version of envelope increases and hence the resolution is de-
PCI and OCT was the poor signal-to-noise ratio. graded if the dispersion is not compensated for.29,30
This was caused by the wavefront mismatch of the We have recently demonstrated a signal width of
beams reflected at the retina and the cornea (which 5 mm (geometrical distance) in an optical A-scan of
is used as a reference surface). Different methods of a human retina in vivo by employing a broadband
overcoming this drawback have been suggested: SLD (Dl;60 nm) with dispersion compensation.30
Sekine, Minegishi, and Koizumi developed an axial Since the output power of this SLD was lower and
eye length measurement system based on the tun- the spatial coherence of the source was worse than
ing of the wavelength of a laser diode. They guided that of usual SLDs, the S/N was rather poor in that
the light to the cornea and the retina through sepa- case. Therefore we preferred to use a synthesized
rate optical paths with identical optical path lengths light source with Dl eff>50 nm generated by super-
using spatial filtering.37 Reflections from the cornea imposing two spectrally displaced SLDs in this
and the retina return through their respective paths work. With this light source and dispersion com-
and interfere with each other after being matched pensation, we demonstrated the first in vivo OCT
in wavefront shape. The application of this tech- image of the retina of a human eye with an axial
nique for measuring axial eye length in 21 adults in resolution of ;6 to 7 mm (geometrical distance).
vivo has been demonstrated. This is approximately a twofold improvement over
Wavefront matching with the use of a special lens existing technology. We were able to resolve struc-
with a hole in its optical center was reported by tures posterior to the retina that have not been
Chen et al.38 Half of the illuminating beam is fo- shown by OCT. To identify the histologic origin of
cused on the cornea by the special lens, the other these structures, further work has to be done.
part passes through the hole in the center and is We want to point out that the images of Figure 6
focused by the lens of the eye onto the retina. Both were recorded with an experimental laboratory
of the beams reflected at the cornea and the retina setup in which each scanning angle had to be ad-
are recollimated by the special lens and the eye justed manually. Therefore the total scanning time
lens, respectively, so that all the optical power re- was several minutes and the lateral resolution is

52 JOURNAL OF BIOMEDICAL OPTICS d

JANUARY 1998 d
VOL. 3 NO. 1
Downloaded From: https://www.spiedigitallibrary.org/journals/Journal-of-Biomedical-Optics on 10 Oct 2023
Terms of Use: https://www.spiedigitallibrary.org/terms-of-use
 SIGNAL AND RESOLUTION ENHANCEMENTS IN DUAL BEAM OCT

rather poor (;150 m m). This is not a major draw- 8. S. Tane, J. Kohono, J. Horikoshi, K. Kondo, K. Ohashi, A.
Komatsu, and T. Kakehashi, ‘‘The study of the microscopic
back of the dual beam PCI technique, but merely a biometry of the thickness of the human retina, choroid, and
question of the simple scanning device we used in sclera by ultrasound,’’ Nippon Ganka Gakki Zasshi 88, 1412–
our laboratory work. On the other hand, the images 1417 (1984).
of Figure 6 demonstrate the main advantage of the 9. R. C. Zeimer, M. T. Mori, and B. Khoobehi, ‘‘Feasibility test
of a new method to measure retinal thickness noninva-
dual beam technique: its exceptional stability in an sively,’’ Invest. Ophthal. Vis. Sci. 30, 2099–2105 (1989).
axial direction. In spite of the long measuring time, 10. S. Asrani, R. Zeimer, M. F. Goldberg, and S. Zou, ‘‘Applica-
the individual A-scans that were recorded at differ- tion of rapid scanning retinal thickness analysis in retinal
ent angles with time intervals of tens of seconds diseases,’’ Ophthalmology 104(7), 1145–1151 (1997).
11. A. F. Fercher and E. Roth, ‘‘Ophthalmic laser interferom-
were mounted on the OCT images without any ad- etry,’’ Proc. SPIE 658, 48–51 (1986).
ditional image processing required to correlate the 12. A. F. Fercher, K. Mengedoht, and W. Werner, ‘‘Eye length
longitudinal positions of the A-scans. This might be measurements by interferometry with partially coherent
helpful in cases where digital postprocessing of the light,’’ Opt. Lett. 13, 186–189 (1988).
13. C. K. Hitzenberger, ‘‘Optical measurement of the axial eye
images could lead to loss of fine image details. length by laser doppler interferometry,’’ Invest. Ophthalmol.
Vis. Sci. 32, 616–624 (1991).
5 CONCLUSION 14. C. K. Hitzenberger, ‘‘Measurement of corneal thickness by
low-coherence interferometry,’’ Appl. Opt. 31, 6637–6642
The dual beam partial coherence interferometry (1992).
technique has been improved by the use of a dif- 15. W. Drexler, A. Baumgartner, O. Findl, C. K. Hitzenberger,
H. Sattmann, and A. F. Fercher, ‘‘Submicrometer precision
fractive optical element for wavefront matching, biometry of the anterior segment of the human eye,’’ Invest.
achieving a significant improvement in the signal- Ophthalmol. Vis. Sci. 38(7), 1304–1313 (1997).
to-noise ratio. To enhance resolution of OCT and 16. A. Baumgartner, C. K. Hitzenberger, W. Drexler, H. Satt-
PCI for in vivo measurements of the posterior struc- mann, and A. F. Fercher, ‘‘Measurement of the anterior
structures of the human eye by partial coherence interfer-
tures of the human eye, light sources with broad ometry,’’ Proc. SPIE 2330, 146–151 (1995).
emission spectra were used together with compen- 17. W. Drexler, C. K. Hitzenberger, H. Sattmann, and A. F.
sation of the first-order group dispersion of the ob- Fercher, ‘‘Measurement of the thickness of fundus layers by
ject to be measured, i.e., the ocular media. partial coherence tomography,’’ Opt. Eng. 34, 701–710
(1995).
High precision and resolution in vivo fundus bi- 18. D. Huang, E. A. Swanson, C. P. Lin, J. S. Schuman, W. G.
ometry and imaging were obtained at various posi- Stinson, W. Chang, M. R. Hee, T. Flotte, K. Gregory, C. A.
tions on the human retina. Hence this improved Puliafito, and J. G. Fujimoto, ‘‘Optical coherence tomogra-
version of the dual beam PCI is able to obtain local- phy,’’ Science 254, 1178–1181 (1991).
19. A. F. Fercher, C. K. Hitzenberger, W. Drexler, G. Kamp, and
ized tomography and biometry of the human eye H. Sattmann, ‘‘In vivo optical coherence tomography,’’ Am.
with high axial resolution. J. Ophthalmol. 116, 113–114 (1993).
20. M. R. Hee, J. A. Izatt, E. A. Swanson, D. Huang, J. S. Schu-
Acknowledgments man, C. P. Lin, C. A. Puliafito, and J. G. Fujimoto, ‘‘Optical
coherence tomography of the human retina,’’ Arch. Ophthal-
Financial support from the Austrian Science Foun- mol. 113(3), 325–332 (1995).
dation (FWF grant P 9781-MED) is acknowledged. 21. C. A. Puliafito, M. R. Hee, C. P. Lin, E. Reichel, J. S. Schu-
man, J. S. Duker, J. A. Izatt, E. A. Swanson, and J. G.
Fujimoto, ‘‘Imaging of macular diseases with optical coher-
REFERENCES ence tomography,’’ Ophthalmology 102, 217–229 (1995).
1. R. H. Webb, ‘‘Scanning laser ophthalmoscope,’’ in Noninva- 22. A. F. Fercher, ‘‘Optical coherence tomography,’’ J. Biomed.
sive Diagnostic Techniques in Ophthalmology, B. R. Masters, Opt. 1(2), 157–173 (1996).
ed., pp. 438–450, Springer-Verlag, New York (1990). 23. B. Bouma, G. J. Tearney, S. A. Boppart, M. R. Hee, M. E.
2. J. F. Bille, A. W. Dreher, and G. Zinser, ‘‘Scanning laser to- Brezinski, and J. G. Fujimoto, ‘‘High-resolution optical co-
mography of the living human eye,’’ in Noninvasive Diagnos- herence tomographic imaging using a mode-locked Ti:Al2O3
tic Techniques in Ophthalmology, B. R. Masters, ed., pp. 528– laser source,’’ Opt. Lett. 20, 1486–1488 (1995).
547, Springer-Verlag, New York (1990). 24. A. G. Podoleanu, G. M. Dobre, D. J. Webb, and D. A. Jack-
3. K. Rohrschneider, R. O. W. Burk, F. E. Kruse, and H. E. son, ‘‘Coherence imaging by use of a Newton rings sam-
Völcker, ‘‘Reproducibility of the optic nerve head topogra- pling function,’’ Opt. Lett. 21, 1789–1791 (1996).
phy with a new laser tomographic scanning device,’’ Oph- 25. A. G. Podoleanu, G. M. Dobre, D. J. Webb, and D. A. Jack-
thalmology 101, 1044–1049 (1994). son, ‘‘Simultaneous en-face imaging of two layers in the Hu-
4. A. V. Menezes, M. Giunta, L. Chisholm, P. T. Harvey, R. man retina by low-coherence reflectometry,’’ Opt. Lett. 13,
Tuli, and R. G. Devenyi, ‘‘Reproducibility of topographic 1039–1041 (1997).
measurements of the macula with a scanning laser ophthal- 26. B. Möller, G. Rudolph, A. Klopffleisch, K. H. Donnerhacke,
moscope,’’ Ophthalmology 102(2), 230–235 (1995). and A. Dorsel, ‘‘Application of diffractive optics for axial
5. A. W. Dreher and K. Reiter, ‘‘Retinal laser ellipsometry: a eye length measurement using partial coherence interferom-
new method for measuring the retinal nerve fiber layer etry,’’ Proc. SPIE 2930, 175–182 (1996).
thickness distribution,’’ Clin. Vis. Sci. 7, 481–488 (1992). 27. A. Baumgartner, B. Möller, C. K. Hitzenberger, W. Drexler,
6. R. N. Weinreb, S. Shakiba, and L. Zangwill, ‘‘Scanning laser and A. F. Fercher, ‘‘Measurement of the posterior structures
polarimetry to measure the nerve fiber layer of normal and of the human eye in vivo by partial coherence interferometry
glaucomatous eyes,’’ Am. J. Ophthalmol. 119(5), 627–636 using diffractive optics,’’ Proc. SPIE 2981, 85–93 (1997).
(1995). 28. M. Born and E. Wolf, ‘‘Interference and diffraction with par-
7. K. Emi, Y. Kobayashi, S. Chujo, C. Fujioka, and M. tially coherent light,’’ in Principles of Optics, pp. 491–555,
Yokoyama, ‘‘The biometry of each thickness of the human Pergamon Press, Oxford (1980).
retina, choroid and sclera by using ultrasound and Fourier 29. C. K. Hitzenberger, W. Drexler, A. Baumgartner, and A. F.
analysis—at the foveola,’’ Nippon Ganka Gakki Zasshi 87, Fercher, ‘‘Dispersion effects in partial coherence interferom-
74–79 (1983). etry,’’ Proc. SPIE 2981, 29–36 (1997).

JOURNAL OF BIOMEDICAL OPTICS d

JANUARY 1998 d
VOL. 3 NO. 1 53
Downloaded From: https://www.spiedigitallibrary.org/journals/Journal-of-Biomedical-Optics on 10 Oct 2023
Terms of Use: https://www.spiedigitallibrary.org/terms-of-use
 BAUMGARTNER ET AL.

30. C. K. Hitzenberger, A. Baumgartner, W. Drexler, and A. F. herence interferometry,’’ Exp. Eye Res., accepted for publi-
Fercher, ‘‘Partial coherence interferometry of the retina with cation.
5 mm resolution by compensation of object dispersion,’’ J. 35. W. J. Tropf, M. Thomas, and T. J. Harris, ‘‘Properties of
Biomed. Opt., submitted for publication. crystals and glasses,’’ in Handbook of Optics, Vol. II, 2nd ed.,
31. D. N. Wang, Y. N. Ning, K. T. V. Grattan, A. W. Palmer, and M. Bass, E. W. Van Stryland, D. R. Williams, and W. L.
K. Weir, ‘‘Optimized multiwavelength combination sources Wolfe, eds., Chap. 33, McGraw-Hill, New York (1995).
for interferometric use,’’ Appl. Opt. 33, 7326–7333 (1994). 36. American National Standards Institute, Safe Use of Lasers,
32. Y. J. Rao, Y. N. Ning, and D. A. Jackson, ‘‘Synthesized ANSI Z 136.1, American National Standards Institute, New
source for white-light sensing systems,’’ Opt. Lett. 18, 462– York (1986).
464 (1993). 37. A. Sekine, I. Minegishi, and H. Koizumi, ‘‘Axial eye-length
33. W. Drexler, O. Findl, A. Baumgartner, K. Strenn, G. Rainer, measurement by wavelength-shift interferometry,’’ J. Opt.
C. K. Hitzenberger, A. F. Fercher, and R. Menapace, ‘‘Dual Soc. Am. 10, 1651–1655 (1993).
beam optical coherence tomography and topography of the 38. S. Chen, D. N. Wang, K. T. V. Grattan, A. W. Palmer, and G.
human eye—a clinical feasibility study,’’ Proc. SPIE 2930, L. Dick, ‘‘A compact optical device for eye-length measure-
183–193 (1996). ment,’’ IEEE Photonics Technol. Lett. 5, 729–731 (1993).
34. W. Drexler, C. K. Hitzenberger, A. Baumgartner, O. Findl, 39. X. Clivaz, F. Marquis-Weible, and R. P. Salathé, ‘‘1.5 mm
H. Sattmann, and A. F. Fercher, ‘‘Investigation of dispersion resolution optical low coherence reflectometry in biological
effects in ocular media by multiple wavelength partial co- tissues,’’ Proc. SPIE 2083, 338–346 (1993).

54 JOURNAL OF BIOMEDICAL OPTICS d

JANUARY 1998 d
VOL. 3 NO. 1
Downloaded From: https://www.spiedigitallibrary.org/journals/Journal-of-Biomedical-Optics on 10 Oct 2023
Terms of Use: https://www.spiedigitallibrary.org/terms-of-use