Dual OCT/SLO imager with three-dimensional tracker

Abstract

We have designed, developed, and tested a three-dimensional tracking and imaging system that uses a novel optical layout to acquire both en-face confocal images by scanning laser imaging (e.g. scanning laser ophthalmoscopy, SLO) and high-resolution depth sections by optical coherence tomography (OCT). The present application for this system is retinal imaging. The instrument is capable of sequentially collecting OCT and SLO images with the simple articulation of an optic affixed to a flip-mount. In addition, we have extended our mature transverse tracking system for full three-dimensional motion stabilization. The tracking component employs an innovative optical and electronic design that encodes transverse and depth tracking information on a single beam. We have demonstrated en face SLO imaging with a resolution of ~25 µm and depth-resolved OCT imaging with a resolution of ~10 µm. On artificial targets, transverse tracking was robust up to 1 m/s with a bandwidth of ~1 kHz and depth tracking was robust up to a velocity of ~15 cm/sec, a range of ~1 mm, and a bandwidth of a few hundred Hz. The details of the instrument, including optical and electronic design, are discussed. The system has the potential to provide clinicians and researchers with a wide variety of diagnostic information for the early detection and treatment of retinal diseases.

Full text

Page 1

PSI-SR-1218

Dual OCT/SLO Imager with Three-
Dimensional Tracker

Daniel X. Hammer
Nicusor V. Iftimia
Teoman E. Ustun
John C. Magill
R. Daniel Ferguson



Daniel X. Hammer, Nicusor V. Iftimia, Teoman E. Ustun, John C. Magill, R. Daniel
Ferguson, "Dual OCT/SLO Imager with Three-Dimensional Tracker ," presented at SPIE
Photonics West (San Jose, CA), (22-27 January2005).
Copyright © 2005 Society of Photo-Optical Instrumentation Engineers.
This paper was published in SPIE Photonics West, and is made available as an electronic
reprint (preprint) with permission of SPIE. One print or electronic copy may be made for
personal use only. Systematic or multiple reproduction, distribution to multiple locations
via electronic or other means, duplication of any material in this paper for a fee or for
commercial purposes, or modification of the content of the paper are prohibited.

Downloaded from the Physical Sciences Inc. Library. Abstract available at
http://www.psicorp.com/publications/sr-1218.shtml

Page 2

Dual OCT/SLO Imager with Three-Dimensional Tracker

Daniel X. Hammer*, Nicusor V. Iftimia, Teoman E. Ustun, John C. Magill, and R. Daniel Ferguson
Physical Sciences Inc., 20 New England Business Center, Andover MA USA 01810-1077


ABSTRACT

We have designed, developed, and tested a three-dimensional tracking and imaging system that uses a novel optical
layout to acquire both en-face confocal images by scanning laser imaging (e.g. scanning laser ophthalmoscopy, SLO)
and high-resolution depth sections by optical coherence tomography (OCT). The present application for this system is
retinal imaging. The instrument is capable of sequentially collecting OCT and SLO images with the simple articulation
of an optic affixed to a flip-mount. In addition, we have extended our mature transverse tracking system for full three-
dimensional motion stabilization. The tracking component employs an innovative optical and electronic design that
encodes transverse and depth tracking information on a single beam. We have demonstrated en face SLO imaging with
a resolution of ~25 µm and depth-resolved OCT imaging with a resolution of ~10 µm. On artificial targets, transverse
tracking was robust up to 1 m/s with a bandwidth of ~1 kHz and depth tracking was robust up to a velocity of ~15
cm/sec, a range of ~1 mm, and a bandwidth of a few hundred Hz. The details of the instrument, including optical and
electronic design, are discussed. The system has the potential to provide clinicians and researchers with a wide variety
of diagnostic information for the early detection and treatment of retinal diseases.

Keywords: Retinal tracking, optical coherence tomography, scanning laser ophthalmoscopy, low coherence
interferometry, retinal imaging

1. INTRODUCTION

Optical coherence tomography (OCT) and scanning laser ophthalmology (SLO) are two retinal imaging modalities that
have had success in diagnosis of retinal pathologies including glaucoma, diabetic retinopathy, macular degeneration,
macular edema, and retinal detachments [1-2]. In general, they measure similar information (back-scattered photons)
with different presentation views. Both SLO and OCT gate the photons with a pinhole (or single-mode fiber) placed
conjugate to the retinal focus. However, OCT additionally gates the photons using interference of low coherence light to
obtain high axial resolution cross-sections of single-scattered light. SLO, on the other hand, provides an en-face view of
single- and multiple-scattered light from within the confocal range gate. Although many groups have developed
instruments that provide an en-face retinal view of data collected from an OCT data cube (by re-ordering the typical A-,
B-, C-scan sequence) [3], their systems don’t provide conventional SLO images because the en-face view consists of
coherent range gates (of single-scattered light) taken from a single axial plane. Moreover, the data cube cannot be
collected and displayed in real time (although a single axial plane can). We have designed an innovative imaging
instrument that can sequentially (not simultaneously) collect SLO and OCT images from the retina. It is designed to
function as a rapid screening device to continuously provide high-contrast confocal SLO images of the entire retina and
intermittently provide high-resolution OCT cross-sections of regions-of-interest suspected of disease. A single source
and novel optical layout based upon a previously-reported hand-held line-scanning laser ophthalmoscope [4-5], make it
amenable to compact, inexpensive, low-power, and portable design and operation. Future designs may provide
simultaneous SLO and OCT images using a single-shot OCT approach.
All retinal imaging types are susceptible to artifact introduced by eye motion. We have established the benefit of image
stabilization for both OCT and SLO using our hardware-based active retinal tracking technology [6-11]. Retinal
tracking improves standard imaging modes but also allows for the development of new imaging techniques, for example
a slow scan technique to acquire wide-field high-dynamic range Doppler blood flow maps [12]. Herein we describe the
design and preliminary feasibility testing of a methodology to extend out tracking system to three-dimensions using the
principle of low-coherence interferometry. The system will enable transverse tracking for correction of rotational eye
motion from saccades. In addition, it will correct for head and other motion that leads to image distortion along the

* hammer@psicorp.com; phone 1 978 689-0003; fax 1 978 689-3232; psicorp.com

Page 3

optical axis (z-axis). This is especially suitable for OCT, which is particularly prone to z-axis motion, since this is
typically the slow scan dimension. Three-dimensional tracking may also benefit image acquisition of rapidly moving
targets such as cardiac or laryngeal tissue. In addition, the technique may be applied to other uses, for example, to
perform dynamic focusing (for parfocality of coherence and confocal range gates) for OCT applications or auto-
focusing for other more conventional imaging applications.
1.1. Motion artifact during retinal imaging
One of the most pernicious artifacts to corrupt OCT image acquisition arises from motion. This is readily apparent in
tissues that are in constant motion, such as the eye or the heart. Imaging the eye poses an additional problem in that
laser safety standards limit the amount of illumination power that can be delivered. This places a limit on the signal-to-
noise ratio and penetration depth that can be achieved. The problem of motion is increasingly deleterious as the
resolution increases. For example, if continuous depth motion in the eye is generally on the order of several microns,
this motion will not be resolved by conventional commercial OCT systems but will become apparent when high-
resolution instruments are used. Since much of the quantification and diagnosis of disease conditions requires the
accurate measurement of boundaries (e.g., nerve fiber layer), the benefit of an increase in resolution will only
be realized by concomitant development of three-dimensional motion stabilization hardware.
Axial motion is particularly a problem for OCT since depth sections are acquired. Clinical time-domain OCT systems
use post-acquisition image analysis software to re-register individual depth scans. The re-registration algorithm uses an
image feature, such as an interface created by a large change in refractive index (e.g., the interface between the vitreous
humor and the retina or the retinal pigment epithelium, which lies between the photoreceptors and the choriocapillaris),
to correct for depth motion. Our real-time three-dimensional tracker provides a number of important advantages for
clinical applications compared to software registration algorithms. First, a software algorithm is inherently susceptible
to error since it relies on assumptions about the structure of the tissue layers to correct the image. Second, important
physiological topographical information may be degraded or entirely lost in the re-registration process. For example, a
retinal feature such as the optic nerve head may appear to be a motion artifact and cause the re-registration algorithm to
fail. Third, re-registration algorithms (for time-domain systems) do not correct for motion during a depth scan, only
motion that occurs from one depth scan to the next. In comparison, our system can correct for motion artifacts at any
time during the depth scan. Fourth, re-registration algorithms require significant image processing resources, particu-
larly if video frame rates (30 Hz) are used and acquired. Fifth and finally, real-time image information is not available
using a post-processing algorithm. Although a significant time lag between acquisition and display is not significant for
most long-term clinical imaging applications such as measurement of retinal nerve fiber layer thickness, for some thera-
peutic applications where the effect of a particular laser or drug treatment needs to be viewed in real-time (e.g., dye
angiography), our system would provide a significant benefit. Figure 1 shows retinal surface profiles and composite
OCT images acquired with and without transverse retinal tracking. The same scan path is very difficult to repeat without
the reference to fixed retinal coordinates provided by tracking. Even in the tracked case, the residual depth motion is
apparent. Alignment (in depth) and averaging reveals superior detail with tracking in the images to the right.
Motion appears to be decreasingly significant as the acquisition speed increases. Recently, a technology called spectral
domain OCT (SDOCT) has been refined for very high speed acquisition of OCT images [13]. SDOCT can acquire A-
scans at the line rate of a linear array, up to tens of kHz. This translates to frame rates exceeding 100 frames/sec for a
well-sampled transverse section. If the depth motion occurs in hundreds of milliseconds, SDOCT should be able to
acquire an entire depth section with little motion artifact in 50 ms (500 lines/10 klines/ sec). However, SDOCT imaging
will still benefit from retinal tracking for several important applications. First, even moderately-sampled three
dimensional maps made up of multiple transverse sections will require several seconds in the fastest SDOCT systems.
These maps may be difficult to construct without some means of stabilization and the full potential of SDOCT may thus
go unrealized. Second, retinal tracking enables a reduction in noise and speckle when co-adding large numbers of
transverse scans. Third, retinal tracking improves the ability to align the scan to the same retinal coordinates for
longitudinal studies to track disease progression and for other clinical applications. Although post-processing techniques
can be used to align individual scans to retinal landmarks with accuracy dependent upon the fundus imaging technique,
relocation to those coordinates for subsequent scans would still be subject to fixational precision. One example of the
importance of this third reason is in the measurement of blood flow velocity using Doppler frequency shifts. An
efficient processing algorithm that has been commonly applied to both ODT [14-15] and SDODT [16] systems involves
measurement of phase difference between adjacent A-scans. This algorithm works on the assumption that speckle

Page 4

between adjacent lines is reasonable correlated. Any motion will degrade the accuracy of the measurement to varying
degrees by invalidating this assumption. It should be noted that the reasons enumerated above apply both to transverse
tracking and depth tracking. Finally, although SDOCT is exciting new technology, it is less proven than time-domain
OCT for a variety of pathologies and may also eventually prove costly for clinical use. Subjects with retinal diseases
routinely scanned in the clinic present a host of problems (poor or no fixation, cloudy aqueous and vitreous humor, low-
grade cataract, etc.) that degrade performance in comparison to the young, healthy subjects scanned during SDOCT
research instrumentation development. Many of these issues can be vastly improved with the use of a retinal tracker.
We are currently testing a time-domain clinical optical coherence tomography system (Stratus OCT, Carl Zeiss Meditec
Inc.) with a second-generation retinal tracker [10-11]. Thus three-dimensional tracking is important even for high-speed
SDOCT systems. The three-dimensional tracking platform we have demonstrated herein can be applied to both time-
domain and spectral-domain OCT imaging systems.

Fig. 1. Retinal surface profile and composite images for three radial disc scans on three separate days with and without tracking in a
normal subject. Profiles are vertically aligned at a single point to enable comparison of scans in the presence of depth (z) motion.
Composite images are created by averaging nine individual scans after application of a post-processing z-axis re-registration algorithm.
1.2. Techniques for depth motion stabilization
Several techniques have been developed for medical and non-medical applications for the measurement of motion in
one axis using interferometry. For example, laser Doppler vibrometry detects the shift in fringe pattern that results when
two laser beams illuminate the sample that are at equal angles from the direction normal to the sample motion [17]. For
medical applications that involve low coherence interferometry, there are at least two paths that one may envision for a
depth tracking system. First, one may detect tissue edges for each A-scan and correct the z position of subsequent scans
by any position shift that may occur. This adaptive ranging (AR) technique has been used to correct the scan range for
cardiac and gastroenterological endoscopic OCT applications for large diameter vessels or where there is difficulty in
catheter placement [18]. AR has a bandwidth that is a fraction of the A-scan rate, which is determined by the properties
of the reference arm delay line scanner for time domain OCT or the linear array camera integration time for SDOCT.
For example, to correct an ideal sinusoidally-varying disturbance, one would need an A-scan rate such that the edge
position difference between adjacent A-scans is small to minimize errors that occur from correcting the next scan with
the position error from the previous scan. In other words, this non-deterministic approach results in a phase lag that

Page 5

limits bandwidth. AR resolution is limited by the image pixel size, which in turn depends on the coherence length of the
source. The detection and correction range will also be limited by the reference arm scanner or linear array integration
time. However, the depth range may be quite large, even larger than that used for imaging, as long as the instantaneous
velocity is not such that the edge moves completely out of the image window. Also, for SDOCT the A-scan rates are
quite high (~30 kHz). Since AR uses OCT image information to correct position, this method is the real-time equivalent
to the post-processing software algorithms that are currently employed to flatten an image or remove depth motion
artifact. As such, they suffer from the same disadvantages, most notably, any surface topographical information will be
lost in the flattened images. For some applications and tissue types (e.g., cardiac endoscopy), this disadvantage is not
significant. In summary, this technique is useful for removal of slow scan offsets that do not require preservation of
tissue topographical information.
A second method to detect and correct depth motion is to use the Doppler frequency shift that occurs when a tissue
moves. This approach, henceforth named Doppler ranging (DR), detects the Doppler shift, converts the frequency shift to
tissue velocity, and finally integrates the velocity signal to obtain an absolute position. Since the image and tracking
beams are not co-located at the focus with the tracking beam stationary and the image beam scanning transversal with
respect to the sample, topographical boundaries are maintained. The bandwidth of DR can be quite high (hundreds of Hz
to kHz): limited by the detection electronics and the bandwidth of the z-scanner. As in AR, the range will be limited by
the reference arm galvanometer. However since this technique must integrate the detected signal to measure position, DR
is susceptible to offsets. Moreover, very low frequency drifts that cannot be detected by the system may also cause the
system to lose lock. Although the tracking instrument was designed and constructed to implement this second technique,
the first technique was also tested to remove low frequency drifts (e.g., a hybrid approach) and as a means of depth
tracking in and of itself. However, once AR is included as a correction to a system that uses DR, the advantage of tissue
topographical preservation is lost.

2. MATERIALS AND METHODS
2.1 Optics
The optical schematic and photographs of the front-end of the dual tracking SLO/OCT imaging system are shown in
Figs. 2 and 3. Imaging and tracking on artificial targets took place at the first retinal conjugate and the ophthalmoscopic
lens (OL) was inserted for the retinal imaging and tracking. The SLO portion of the system is similar to previous line-
scanning instruments patented by PSI and described more thoroughly in other publications [4-5]. This design achieves
nearly the same confocality as a flying-spot scanner without the need for a fast moving resonant scanner or expensive
spinning polygon. It uses a cylindrical lens (CL) to focus the beam to a line and a single imaging galvanometer (IG) to
scan the line across the sample. A mirror with a hole drilled in the center separates entrance and exit pupils. This “split
aperture” (SA) placed near a pupil conjugate reduces image artifacts from back-reflections. The light reflected from the
sample is imaged onto a linear array detector (LA) where synchronization between the array and the galvanometer yield
high-contrast confocal images of the sample. Several different linear arrays were tested including a CMOS camera built
by PSI, a CMOS camera built by Fairchild Imaging Inc, and a CCD camera built by Basler Vision Technologies Inc.
Each camera had different parameters (sensitivity, pixel size, line rate, etc.). The cylindrical lens is affixed in a flip-
mount so that when it is moved out of the optical axis, a single beam is focused on the sample rather than a line for OCT
imaging. Automatic mechanical articulation of the CL mount will be implemented in future systems. In this setup, the
OCT beam can be scanned transversely with the SLO imaging galvanometer and a dichroic beamsplitter mounted to a
second galvanometer. The OCT and tracking sources are fiber-coupled into the front-end with fiber collimators (FC)
mounted on stages to adjust optical path length.
The (SLO or OCT) imaging beam passes through a set of tracking galvanometers (TG), which move as transverse
motion is detected to keep the beam locked to fixed sample coordinates. The principle of operation of the patented
retinal tracker has been described more thoroughly in previous publications [6-11]. In essence, the tracking beam is
dithered with resonant scanners on a target and changes in target reflectivity associated with motion are processed and
fed back as error signals into the drive electronics of the tracking galvanometers. In these previous implementations of
the transverse tracker, a confocal reflectometer is placed immediately posterior to the resonant scanners. To add depth
tracking to this system, a low coherence source is used with a fiber interferometer in the tracking beam path. Thus
transverse and depth tracking are achieved with the same beam in this design. As will be shown in Section 2.2, both
transverse and depth motion signals can be extracted from the detection electronics.

Page 6

Fig. 2. Optical schematic of the front-end of the tracking SLO/OCT imager. All imaging/tracking took place at the retinal conjugate
(r). OL: ophthalmoscopic lens (not used in Phase I), SL: scan lens, TG: tracking galvanometers, IG: imaging galvanometer, D:
dichroic beamsplitter, RS: resonant scanners, M: turning mirror, FC: fiber collimator, SA: split aperture, CL: cylindrical lens, IO:
imaging objective, LA: linear array detector, FM: flip mount, AOM: acousto-optic modulator, OPL: optical path length adjust. AOM
in the sample path was used for dispersion compensation only (i.e., it was not driven with RF signal).



Fig. 3. Photograph of the front-end of the tracking SLO/OCT imager. a. Bench-top tracking SLO/OCT imager. b. Close-up detail of
instrument optics.

The back-end of both imaging and tracking sub-systems consist of fiber interferometers. The optical schematic and
photographs of the back-end are shown in Figs. 4 and 5. The tracking fiber interferometer uses a standard fiber
Michelson configuration with the exception of the acousto-optic modulator (AOM) in the reference path. The AOM is
configured in a double-pass arrangement to modulate the signal at 75 MHz (twice the drive frequency). The double-
pass arrangement, used to compensate for the angular dispersion produced by the AOM, is illustrated in the inset in
Fig. 4. The AOM is used to shift the frequency of the depth information signal away from the transverse reflectance
signal at the dither frequency (currently 8 kHz) and also to utilize standard off-the-shelf RF electronic components for

Page 7

Fig. 4. Optical schematic of the back-end of the tracking SLO/OCT imager. I: iris aperture, R: retroreflector, M: mirrors, G: grating,
L: lens, g: galvanometer. The inset shows the double-pass arrangement of the AOM to compensate angular dispersion.

demodulation. The drive frequency of the AOM is adjustable between 20 and 40 MHz but has maximum efficiency near
40 MHz. The 3-dB cut-off of the RF balanced detector is 80 MHz. A retro-reflector attached to a galvanometer with a
lever arm is used to position the range gate within the sample volume. The detected signal is processed with electronics
described in the next section and resultant error signals passed to the galvanometer-driven retro-reflector (Fig. 4) for
depth tracking and the tracking galvanometers (Fig. 2) for transverse tracking.
The imaging interferometer uses two couplers in a balanced configuration. This configuration improves signal-to-noise
ratio by suppression of common-mode noise in the interferometer arms. A rapid scanning optical delay line (RSODL) is
used to generate depth (A-) scans by scanning the coherence range gate within the sample focal volume. Multiple depth
scans are concatenated to generate a depth section. The RSODL uses a grating to independently control group and phase
velocity so that no phase modulation occurs when the group delay is scanned [19]. Previous configurations have used
resonant scanners to achieve A-scan rates of 8 kHz and video frame rates (30 frames/sec) for a medium resolution
frames (256 pixels) [20]. Since image speed was not important for our investigation, we achieved a range of 5 mm and a
rate of 500 A-scans/sec (1 frame/sec) using a galvanometer.
2.2 Tracking Electronics
The tracking system was designed to use a single beam to encode depth and transverse sample motion information. The
detector signal is amplitude modulated by the dither signal at ~8 kHz and frequency modulated by the Doppler shift at
twice the AOM modulation frequency (~75 MHz). This approach, although elegant from the stand-point of optical
complexity (i.e., a single beam is used and a confocal reflectometer is not needed), has never been proven in an actual
tracking system and also requires substantial additional processing to electronically separate depth and transverse
motion signals. A major concern is whether low-noise signals with sufficient amplitude can be extracted and used for
transverse tracking after undergoing frequency demodulation. In the confocal reflectometer signals are acquired from a
relatively long confocal range gate and therefore even with a beam power of only 15 µW, enough signal is returned
from the retina for lock-in detection and tracking. Conversely, in an OCT system light is detected from only a very

Page 8

narrow coherence range gate and therefore a beam power >500 µW is usually required to obtain signals from the retina.
Therefore, for bulk tissue tracking, a source with a narrow band is desirable.
An alternate design was developed that uses a confocal reflectometer to optically separate the depth and transverse
tracking signals. Normally, a split aperture is placed at a pupil conjugate to stop corneal reflections. However, since the
depth tracking component uses a low coherence interferometer, corneal reflections will not be detected in the light
directly returned from the tissue. Therefore a split aperture can also be used to separate light used for depth tracking
(directly reflected) from transverse tracking (outer ring). This alternate technique has the advantage that the transverse
tracking component is essentially the same as those proven in other fielded systems (TSLO and TOCT). Both methods
(optically and electronically separating the signals) were tested.
The system electronics block diagram and photograph are shown in Fig. 5. The signal received by the detector is band-
pass filtered (BPF) and amplified with a low noise amplifier (LNA). The signal is then split (S) to be separately
processed for frequency (depth motion) and amplitude (transverse motion) demodulation. The former is accomplished
with a custom printed circuit board, the depth tracking board (DTB) designed, developed, and tested for this project and
shown in Fig. 6.


Fig. 5. Electronics and signal processing design. a. Block diagram. b. Photograph. BPF: band-pass filter, LNA: low noise amplifier,
S: power splitter, RF: radio frequency, LO: local oscillator, IF: intermediate frequency, LPF: low-pass filter.



Fig. 6. Printed circuit board designed and constructed for frequency demodulation. a. Block diagram. b. Photograph. VCO: voltage-
controlled oscillator, PLL: phase-locked loop.

Page 9

The transverse motion signal is extracted by down-converting the detector signal with a plug-in RF mixer and detecting
low level amplitude changes at the dither frequency with a lock-in amplifier. The “RF Out” signal from the DTB is used
as the local oscillator (LO) signal into the RF mixer. Amplitude modulation has been removed from this signal; it is
purely modulated at the instantaneous Doppler frequency of the sample. (The design originally amplified and band-pass
filtered this signal; in practice this processing was not necessary.) The intermediate frequency (IF) output from the
mixer is a low frequency (8 kHz) amplitude modulated signal. This signal is connected to a dual-phase lock-in amplifier
(Scitec Instrument LTD) to get the in-phase and quadrature error signals for closed-loop transverse control via software
proportional-integral-derivative (PID) control loop and tracking galvanometers.
In the depth tracking portion of the system, there are actually three nested control loops used each for RF carrier signal
regeneration, interferometer reference arm position control, and galvanometer position. The first is used to regenerate a
clean frequency modulated signal absent of amplitude modulations for transverse tracking (see above). This loop, from
which the sample velocity signal is also derived, must be designed for a specific range of velocities (Doppler frequency
shifts). It thus has a limited linear range of operation (i.e., the capture range). The second loop is simply a software
proportional control loop solely resident and executed on a real-time data acquisition board. This loop takes the depth
signal (sample velocity), integrates to obtain position and feeds that signal to the retro-reflector galvanometer board.
The third is the closed-loop electronics of the retro-reflector galvanometer driver board itself. This loop uses the galva-
nometer position sensor to lock to a position determined by the input voltage from the second software loop.
The block diagram and photograph of the DTB is shown in Fig. 6. To accomplish carrier regeneration and frequency
demodulation, the depth tracking electronics use a traditional phase-locked-loop (PLL) design, consisting of PLL IC
(LMX2306), low-pass filter, and voltage control oscillator (VCO). This control loop generates a DC voltage (velocity)
signal that is proportional to the Doppler frequency shift detected in the signal. The VCO gain determines the range
over which the depth tracking electronics will operate. Note that the loop itself is only necessary for carrier regeneration
– a voltage proportional to the instantaneous Doppler frequency shift from the sample is found from mixing the detected
signal with a frequency from the VCO.
The signal from the power splitter is used as the reference signal in the DTB by the PLL IC and connected to “Ref In”
(Figs. 5-6). An active low pass filter (LPF) is used in the PID loop of the PLL. This matches the dynamic range of the
VCO to the PLL IC. Since the software control loop integrates velocity to obtain position, it is important to remove any
electronically-derived DC offsets in the input signal. This is accomplished with an additional offset-correction circuit
(LPF and instrumentation amplifier). The “Depth Out” signal is passed to the depth controller (software and
galvanometer control loops). As discussed above, the output from the VCO (“RF out”) is used to down convert the
signal coming from the power splitter for amplitude demodulation. Lock detect circuitry has been implemented to
monitor the locking status of the DTB. “Lock Det Out” is zero when the board is locked and one when it is not. In
addition, a test VCO is populated on the board to test the operation of the board in the absence of a reference signal.
A PCI real-time data acquisition processor board (DAP, Microstar Laboratories Inc.) was used to process the transverse
and depth tracking software control loops referred to in the previous section. The DAP board acquired error, position,
and reflectance signals and output three signals, two to the transverse tracking galvanometer boards and one to the depth
tracking retro-reflector galvanometer board. A PCI daughter board was used to extend the number of DAP outputs from
2 to 6 channels. A digital framegrabber board (PCI-1424 or PCI-1428, National Instruments Inc.) was used to acquire
SLO images. A high-speed data acquisition board (DAQ, PCI-6115, National Instruments Inc.) was used to acquire
OCT A-scan signals and to generate analog out (AO) signals to drive the SLO image galvanometer synchronized with
the linear array and the RSODL galvanometer offset when adaptive ranging was used. In addition, various triggers and
counters were used over the real-time system integration (RTSI) line between framegrabber and DAQ boards. A
function generator and audio amplifier were used to drive the sample targets attached to a medium-sized 8-Ω speaker to
create a z-motion disturbance. Software was written to acquire, process, and display the SLO and OCT images, initiate
tracking control and alter control parameters, and manage all other hardware tasks (e.g., multiple PCI board
communication and system timing) and software processing. The software was written in LabVIEW (National
Instruments Inc.) and C.
b

Page 10

3. RESULTS
3.1 Imaging
SLO images were acquired from a variety of test targets included an Air Force resolution chart. Figure 7a shows SLO
images from the instrument taken with the Fairchild linear array camera. The Air Force resolution chart demonstrates a
resolution better than the 5-3 group-element in the direction parallel to the linear array (i.e., direction of line scan) and
better than 5-1 group-element in the direction perpendicular (i.e. along the linear array. The resolution parallel is
determined by the magnification (1.3) and the line scan increment. The resolution perpendicular is determined by the
magnification and pixel pitch (7 µm). The 5-1 and 5-3 group-elements are 32.0 and 40.3 line-pairs/mm or a resolution
of 31 and 25 µm, respectively. This is 2-3 times the diffraction-limit of ~10 µm and quite sufficient for wide-field
retinal images. The field of view obtained on the PSI camera (with larger pixels than the Fairchild and Basler) was 13 ×
6 mm. Typical retinal images acquired from the LSLO from different subjects is show in Figs. 7b and c. No processing
has been performed on these images. Since infrared light scatters less, the 905-nm illumination source provides deeper
tissue penetration than shorter wavelengths. The result is immediately apparent, especially for subjects with a low
pigmentation: the choroidal vasculature is faintly visible beneath the retina. The retinal vessels in all the images appear
sharp and the lumen is clearly visible for the larger vessels. An area of hypopigmentation can be seen in Fig. 7c (arrow).

Fig. 7. SLO images of (a) an AF resolution chart, (b) the optic nerve head of a human volunteer, and (c) the macula of a
second human volunteer.

An example of preliminary OCT images obtained from the instrument are shown in Fig. 8. Note the penetration into a
diffuse target (standard business card shown in Fig. 8a). The reflectivity of the layers of the artificial eye in Fig. 8b are
roughly similar to that of a live eye. Figure 8c shows very preliminary in vivo OCT images of the retina of a human
volunteer. Several layers are clearly delineated including the nerve fiber layer with gradually decreasing thickness in the
A-scans closest to the fovea (indicated by arrow). This final image was taken without tracking and was not optimized
for dispersion compensation.

Fig. 8. OCT images of (a) a white business card, (b) an artificial eye (inset shows picture of eye), and (c) the retina of a
human volunteer.

Page 11

The primary artificial target used to test depth tracking was a target attached to a speaker. Although we were able to
achieve interference signals that could be processed for transverse and depth tracking from specular targets (mirrors and
retroreflecting tape), signals from diffuse targets proved more difficult to obtain. This was due to the lower signal-to-
noise ratio (SNR) of the depth tracking interferometer determined by SLD power, AOM efficiency, and detector band-
width. Future systems will implement methods for improving SNR to obtain interference signals from diffuse tissues.
The first step in the demonstration of depth tracking feasibility was to record the detected signals that were processed by
the depth tracking board (DTB). As discussed in Section 2.2, the detected signal is frequency modulated at twice the
AOM drive frequency and amplitude modulated by reflectance variation at the resonant scanner (i.e., dither) frequency.
The former is the output of the DTB. Figure 9 shows signals obtained for speaker frequencies between 1 and 15 Hz
(Doppler shifts between 5 and 70 kHz, velocities between ~2 and 30 mm/sec).
y = -0.0058x + 4.5514
R2= 0.937
y = 0.0044x + 4.5575
R2= 0.9542
4.1
4.2
4.3
4.4
4.5
4.6
4.7
4.8
4.9
5
0 10 2030 40 50 60 7080
Doppler Frequency (kHz)
- Direction
+ Direction
Depth Signal (V)
G-7107
Fig. 9. Signals from DTB.

The next demonstration of feasibility was to close the loop and characterize tracking performance on a target moving
along the z-axis. The loop was closed with a software switch, which sent the depth position signal processed by the
DAP board to both the depth tracking retro-reflector galvanometer and the OCT imaging RSODL galvanometer. OCT
images were collected from a mirror attached to a speaker as it was driven at a range of frequencies from ~5 to 150 Hz
and a displacement (amplitude) up to 1 mm. The A-scan rate for this demonstration was 200 scans/sec (0.4 frames/sec).
For a stationary speaker, the OCT image is a line a few pixels wide. While OCT images were continuously collected,
the tracker was turned on at low amplitude and the amplitude was increased until just before lock was lost. Tracking
was then turned off. Figure 10 shows the OCT images acquired for three frequencies (~30, 50, 120 Hz). With the
tracker off, the displacement approached 1 mm for all frequencies. With tracking on, the depth motion of the speaker is
tracked and the amplitude decreased. For smaller displacements, the mirror appears completely flat. For larger
displacements, such as those shown in Fig. 10, the tracking accuracy is reduced. This was somewhat expected since the
closed-loop control must have a finite error signal to act upon. For a system that is not critically damped, a rapid
disturbance will exhibit some overshoot. Thus, the data suggests the tuning of the control loops (especially the retro-
reflector galvanometer board) was not optimized. In future implementations, some effort will be made to better design
and implement the depth tracking closed-loop control. To summarize the depth tracking data, we achieved reasonably
accurate and robust tracking with a velocity up to 15 cm/sec, a bandwidth of ~5 to >150 Hz, and a range up to 1 mm.

Page 12

Fig. 10. OCT image of moving front surface reflector before and after tracking is turned off. Reflector is attached to speaker driven
at: a. 30 Hz, b. 50 Hz, and 120 Hz. Horizontal bars illustrate 1 mm displacement. Illumination power = 450 µW, scale bar = 1 mm.

4. DISCUSSION AND CONCLUSIONS

The goal of this nascent program was to establish the feasibility of extension of our tracking technology to three-
dimensions and to integrate a prototype three-dimensional tracker into a novel dual SLO/OCT imager. We have
successfully demonstrated feasibility and preliminary design and operation of the imager, but much work remains to
develop a prototype system suitable for use in a research or clinical environment.
To summarize performance results, the novel dual SLO/OCT imager acquired SLO images with a field of view (FOV) of
~13 × 6 mm (50 deg diagonal), a resolution of ~25 µm, and a speed up to 75 frames/sec (512 × 512). OCT images were
acquired from ~5 × 5 mm sections with a resolution <10 µm and a speed up to 1 frame/sec (512 × 512). The z-tracker
locked to ~1-mm axial movements at velocities up to 15 cm/sec with a bandwidth of ~5 to >150 Hz. The x-y tracker
locked to movements with a peak-to-peak magnitude > 7 mm, velocities up to 0.5 m/sec, an accuracy of < 15 µm and a
closed-loop bandwidth of nearly 1 kHz. The primary hurdle that was not overcome in the investigation was related to
detecting signal levels from diffuse targets with the depth tracker. This prevented full implementation of the elegant
heterodyne methodology for transverse tracking. A confocal reflectometer was therefore used for transverse tracking.
There are several improvements that are necessary to move the system toward eventual clinical operation or use in
specific research applications. First, the depth tracker will be refined and re-designed to provide tracking from diffuse
targets and the retina. Second, full integration of depth and transverse tracking will be accomplished and tested. This
will allow in-vivo demonstration of full, simultaneous three-dimensional tracking. Finally, the OCT imaging will be
converted to a spectral domain system to allow for higher acquisition speeds and a further reduction in the number of
moving parts (i.e., elimination of the delay line galvanometer). This system re-configuration and optimization will allow
realization of stabilization in three-dimensions in a unique dual-mode imaging instrument.

ACKNOWLEDGEMENTS

This work was supported by NIH Grant EB000586.

REFERENCES

1. Optical Coherence Tomography of Ocular Diseases, eds. C. A. Puliafito, M. R. Hee, J. S.Schuman, and J. G.
Fujimoto, Slack Inc., Thorofane NJ, 1996.
2. A. E. Elsner, D. Bartsch, J. J. Weiter, and M. E. Hartnett, “New Devices for Retinal Imaging and Functional
Evaluation,” in Practical Atlas of Retinal Disease and Therapy, ed. W. R. Freeman, Lippincott-Raven Publishers,
Philadelphia, 1997.
3. J. A. Rogers, A. Podoleanu, G. Dobre, D. A. Jackson, and F. W. Fitzke, "Topography and volume measurements
of the optic nerve usingen-face optical coherence tomography," Opt. Express 9, 533-545 (2001),
http://www.opticsexpress.org/abstract.cfm?URI=OPEX-9-10-533.
4. R. D. Ferguson, “Line-Scan Laser Ophthalmoscope,” U.S. Patent #6,758,564 and others pending.
5. D.X. Hammer, R. D. Ferguson, T. E. Ustun, G. Maislin, R. H. Webb, “Hand-held digital line-scanning laser
ophthalmoscope (LSLO)” in Ophthalmic Technolgies XIV, Eds: Manns, Soderberg, and Ho, SPIE Vol. 5314, pg.
161-169 (2004).

Page 13

6. R. D. Ferguson, “Servo Tracking System Utilizing phase-Sensitive Detection of Reflectance Variation,” U. S.
Patent #5,767,941 and #5,943,115.
7. D. X. Hammer, R. D. Ferguson, J. C. Magill, M. A. White, A. E. Elsner, and R. H. Webb, “Image stabilization for
scanning laser ophthalmoscopy,” Opt. Express 10, 1542-1549 (2002),
http://www.opticsexpress.org/abstract.cfm?URI=OPEX-10-26-1542.
8. D. X. Hammer, R. D. Ferguson, J. C. Magill, M. A. White, A. E. Elsner, and R. H. Webb, “Tracking scanning laser
ophthalmoscopy (TSLO),” Proceedings of SPIE, Ophthalmic Technologies XIII, Eds. F. Manns, P. G. Söderberg,
and A. Ho, Vol. 4951, 208-217 (2003).
9. D. X. Hammer, R. D. Ferguson, J. C. Magill, M. A. White, A. E. Elsner, and R. H. Webb, “Compact scanning laser
ophthalmoscope with high speed retinal tracker,” Appl. Opt. 42, 4621-4632 (2003).
10. R. D. Ferguson, D. X. Hammer, L. A. Paunescu, S. Beaton, and J. S. Schuman, “Tracking Optical Coherence
Tomography,” Opt. Lett., 29, 2139-2141 (2004).
11. Daniel X. Hammer, R. Daniel Ferguson, John C. Magill, Lelia Adelina Paunescu, Siobahn Beaton, Hiroshi
Ishikawa, Gadi Wollstein, and Joel S. Schuman, “An Active Retinal Tracker for Clinical Optical Coherence
Tomography Systems,” Jounal of Biomedical Optics, in press.
12. R. D. Ferguson, D. X. Hammer, A. E. Elsner, R. H. Webb, S. A. Burns, and J. J. Weiter, "Wide-field retinal
hemodynamic imaging with the tracking scanning laser ophthalmoscope," Opt. Express 12, 5198-5208 (2004),
http://www.opticsexpress.org/abstract.cfm?URI=OPEX-12-21-5198.
13. J. F. de Boer, B. Cense, B. H. Park, M. C. Pierce, G. J. Tearney, B. E. Bouma, “Improved signal-to-noise ratio in
spectral-domain compared with time-domain optical coherence tomography, Opt. Lett. 28, 2067-2069 (2003).
14. Y. Zhao, Z. Chen, C. Saxer, S. Xiang, J. F. de Boer, and J. S. Nelson, “Phase-resolved optical coherence
tomography and optical Doppler tomography for imaging blood flow in human skin with fast scanning speed and
high velocity sensitivity,” Opt. Lett., 25, 114-116 (2000).
15. Y. Zhao, Z. Chen, C. Saxer, Q. Shen, S. Xiang, J. F. de Boer, and J. S. Nelson, “Doppler standard deviation
imaging for clinical monitoring of in vivo human skin blood flow,” Opt. Lett., 25, 1358-1360 (2000).
16. B. R. White, M. C. Pierce, N. Nassif, B. Cense, B. H. Park, G. J. Tearney, and B. E. Bouma, “In vivo dynamic
human retinal blood flow imaging using ultra-high-speed spectral domain optical Doppler tomography,” Opt.
Express, 11, 3490-3497 (2003) http://www.opticsexpress.org/abstract.cfm?URI=OPEX-11-25-3490.
17. G. R. Ball, A. Huber, and R. L. Goode, “Scanning laser Doppler vibrometry of the middle ear ossicles,” J. Ear Nose
Throat, 76, 213-218 (1997).
18. N. V. Iftimia, B. E. Bouma, J. F. de Boer, B. H. Park, B. Cense and G. J. Tearney, “Adaptive ranging for optical
coherence tomography”, Opt. Express, 17, 4025-34 (2004),
http://www.opticsexpress.org/abstract.cfm?URI=OPEX-12-17-4025.
19. G. J. Tearney, B. E. Bouma, and J. G. Fujimoto, “High-speed phase- and group-delay scanning with a grating based
phase control delay line,” Opt. Lett., 22, 1811-1813 (1997).
20. A. M. Rollins, M. D. Kulkarni, S. Yazdanfar, R. Ungarunyawee, and J. A. Izatt, “In vivo video rate optical
coherence tomography,” Opt. Express., 3, 219-229 (1998), http://www.opticsexpress.org/abstract.cfm?URI=OPEX-
3-6-219.