PHYSICS IN MEDICINE AND BIOLOGY
Phys. Med. Biol. 55 (2010) 5515–5528
Elastographic contrast generation in optical coherence
tomography from a localized shear stress
Alex Grimwood1, Leo Garcia2, Jeff Bamber2, Jon Holmes3,
Peter Woolliams4, Pete Tomlins4and Quentin A Pankhurst1
1Davy-Faraday Research Laboratories, Royal Institution of Great Britain, 21 Albemarle Street,
London, W1S 4BS, UK
2Joint Department of Physics, Institute of Cancer Research, 15 Cotswold Road, Sutton,
Surrey SM2 5NG, UK
3Michelson Diagnostics Ltd, 11A Grays Farm Production Village, Orpington, Kent,
BR5 3BD, UK
4Biophotonics Group, National Physical Laboratory, Hampton Road, Middlesex,
TW11 0LW, UK
Received 10 June 2010, in final form 30 July 2010
Published 27 August 2010
Online at stacks.iop.org/PMB/55/5515
from a localized stress is presented.
a non-uniform, localized stress via a magnetically actuated implant.
effectiveness is demonstrated using finite-element simulations and a phantom
study provides experimental verification of this.
to a superficial cancerous lesion model represented as a stiff inclusion in
normal tissue. The lesion was best distinguished from its surroundings using
total shear strain elastograms, rather than individual strain components. In
experimental phantom studies, the lesion was imaged using optical coherence
tomography (OCT) and could still be distinguished in elastograms when not
readily identifiable in standard OCT images.
The technique involves generating
The method is applied
(Some figures in this article are in colour only in the electronic version)
There is often a relationship between tissue pathology and tissue stiffness. For example,
disease in soft tissue is commonly monitored using manual palpation (Garra et al 1997).
Such mechanical variations are exploited in elastography, where contrast is generated between
regions of high and low strains, which relate information on stiffness. Elastography has been
used across a range of modalities (Muthupillai et al 1995, Emelianov et al 2004, Rogowska
et al 2006) and elastograms have been generated using optical coherence tomography (OCT)
0031-9155/10/185515+14$30.00 © 2010 Institute of Physics and Engineering in MedicinePrinted in the UK5515
5516A Grimwood et al
for over a decade (Schmitt 1998). OCT is a non-invasive imaging modality based on low
coherence interferometry, producing micron resolution images down to a depth of roughly
as well as developmental biology, dermatology and cardiology (Podoleanu and Rosen 2008,
Manner et al 2009, Ziolkowska et al 2009, Kim et al 2009).
This paper describes a technique for generating useful contrast from a localized, non-
uniform stress. A superficial model resembling a skin lesion is used in conjunction with
an embedded, sub-millimetre, magnetizeable implant. The implant is actuated by a strong
magnet positioned outside the specimen. This produces a strain distribution with a large shear
component in the surrounding tissue. Any changes of tissue stiffness affect this distribution
and give rise to a change in contrast. Whilst the technique is demonstrated here using OCT,
such a concept could be applied to other modalities. Neither is the technique reliant on an
implant. Magnetic nanoparticles have already been demonstrated as both elastographic and
contrast agents (Oldenburg et al 2008, Crecea et al 2009) and future work could involve
magnetic localization of such particles, which can then be actuated in much the same way as
Strain generated by actuating the implant has a large shear component. Shear strains have
already been used in ultrasound (Konofagou et al 2000, Thitaikumar et al 2007) and OCT
(Filas et al 2007). Additionally, dynamic shear measurements have been described for OCT,
ultrasound and laser speckle imaging (Liang et al 2009, Tanter et al 2008, Kirkpatrick et al
2006). However, these techniques typically measure only a single directional component of
shear strain. Through finite-element simulations, we demonstrate a way in which total shear
contrast more effectively than analysing a single component. We also present experimental
verification in the form of a phantom study closely resembling the finite-element models. This
is achieved using normalized cross-correlation to track displacements between two B-scans,
followed by a least-squares strain estimation (LSQSE) algorithm to analyse strain.
2. Materials and method
For this experiment, elastograms were generated from finite-element simulations. These
were then compared with elastograms taken from OCT images of two phantoms in order
to verify observations from the modelled data. For all simulations and measurements, the
phantoms were treated as linear elastic solids, with Young’s moduli based on measurement
and minimal visco-elastic properties. They generally emulated diseased and healthy tissue
with one incorporating a stiff lesion and the other being uniform throughout. Mechanical
actuation was achieved by remotely pulling on a magnetizable metal implant embedded in
the phantom using an external magnet. Images prior to actuation were correlated with those
taken during a constant, applied strain to produce static elastograms. The phantom lesion
was approximately seven times stiffer than surrounding tissue, a figure previously observed in
melanoma (Kirkpatrick et al 2006).
2.1. Elastography phantom
Tissue phantoms were fabricated using a two part condensation-cured RTV silicone rubber
(MM282-A & B, ACC Silicones Ltd) mixed with silicone fluid and titanium dioxide powder
(T/1900/53, Fisher Scientific, UK) at a concentration of 1 mg g−1to promote optical
scattering. Silicone fluid (F111/50 polydimethylsiloxane, ACC Silicones Ltd) was added
during preparation to tune Young’s modulus of the phantom (Crecea et al 2009). This mixture
Elastographic contrast generation in optical coherence tomography from a localized shear stress5517
Figure 1. OCT images of control phantom (a) and inclusion phantom (b). In both images, the
phantom’s upper surface is visible as a white line (marked by the arrow). The top of the magnetic
implant can also be seen (bottom centre). Although not visible, the inclusion is situated above
the implant in (b). Horizontal banding artefacts are generated at the interfaces between the OCT
system’s four channels, which are blended together to form a single image.
Stiff phantoms were produced by adding less silicone fluid. To aid dispersion of the TiO2,
hexane fluid was added as a thinning agent at a volumetric ratio of 1:1 before sonication.
Finally, the mixture was degassed, allowing the hexane to evaporate off without affecting the
curingprocess(Bisaillonetal2008), andthencastintomoulds. Amagnetizablemetalimplant
was incorporated by casting the phantom in a series of layers, such that the top of the implant
was 1.58 mm from the phantom’s surface. The implant was a chrome steel ball-bearing with
a diameter of 0.79 mm (Grade 25 AISI 52100 chrome steel, Simply Bearings Ltd, UK).
Two phantoms were constructed (figure 1). The first was a homogenous block with a
height of 4 mm and sides of length 40 mm. The second had equal dimensions and modulus,
but also contained an inclusion made from stiffer rubber and centred above the implant. The
inclusion was a disc 0.8 mm in height, with a diameter of 2.0 mm, positioned with its upper
face flush against the top surface of the phantom. The scattering properties for bulk and
inclusion were closely matched, with little variation in optical contrast between the two, as
can be seen in figure 1(b). The inclusion was securely bonded within the phantom by adding
it before allowing the surrounding rubber to cure.
The apparent Young’s moduli of the phantoms were measured by casting test objects and
subjecting them to compression tests (Instron 3342 with 2519-103 load-cell, Instron, UK). For
the bulk and inclusion, these were measured as 36.7 ± 7 kPa and 268.5 ± 31 kPa respectively
at a compression rate of 2% s−1.
2.2. OCT probe and actuator apparatus
A diagram of the experiment is shown in figure 2. It consisted of a multi-beam swept-source
OCT system (EX1301, Michelson Diagnostics Ltd, UK) and magnetic actuation assembly.
The OCT light source (HSL-2000-10 MDL, Santec Corp.) produced a central wavelength
of 1305 ± 15 nm with a 150 nm sweep and a peak power output of 15 mW. Pixel size was
4.3 µm in tissue (at a bulk refractive index of 1.35), and B-scans were attained using a sample
rate of 14.2 MHz and A-scan rate of 10 kHz. B-scans comprised four imaging beams each
with a 0.25 mm depth of focus centred at adjacent axial positions producing a 1 mm effective
focal range and a point spread function of 10.9 ± 2.0 mm axially and 8.9 ± 2.0 mm laterally
at FWHM for each channel (Tomlins et al 2009). No smoothing is performed on the OCT
5518 A Grimwood et al
Figure 2. Schematic of the OCT probe and magnetic actuation apparatus.
image signal, although signals from the four beams are combined to form a single image,
which sometimes results in visible horizontal banding. The operating software offers some
control over how the four channels are blended to produce the final image, allowing the user
to mitigate this effect.
The OCT system imaged a phantom placed on a glass platform underneath which was
mounted the magnetic actuator. Cross-section images were taken through the upper part of the
phantom 7 mm wide and 1.8 mm deep. As seen previously in figure 1, this area incorporated
the phantom’s upper surface, the top part of the magnetic implant and the entire stiff inclusion
(where present). The actuator was constructed from a custom-built three-element Halbach
array of neodymium–iron–boron magnets fixed to a vertical translation stage (PT1/M-Z7,
The Halbach array was chosen for its ability to produce relatively strong magnetic fields and
high field gradients. These were measured to be 0.4 T and 67 Tm−1, respectively, at a 5 mm
± 0.1 mm distance using a Hall probe. As the separation between the magnet and the implant
varied, so the force experienced by the implant changed. This was measured by attaching an
implant to a force balance and comparing its weight change relative to the distance from the
array. A force–distance profile was compiled and used to predict the force exerted on implants
in the phantom.
2.3. Finite-element model
The commercial FEM software Comsol Multiphysics 3.4 (COMSOL, USA) was used to
generate a 2D finite-element model resembling a cropped, central cross-section through the
Elastographic contrast generation in optical coherence tomography from a localized shear stress5519
Figure 3. Schematic of the phantom’s 2D finite-element model incorporating a circle for the
spherical implant and rectangle for the stiff inclusion. The model’s width is half that of the real
tissue phantoms with a width 20 mm, height 4 mm and Young’s modulus 37 kPa for the bulk.
A 2D model of reduced dimensions was adopted because there was insufficient computer
memory to simulate the phantom’s entire 3D structure at a suitable mesh density. This model
comprised 2090 mesh points and incorporated a rigid circle of diameter 0.79 mm to resemble
the implant. The circle was assigned the mechanical properties of high-strength alloy steel,
as chosen from the default Multiphysics materials library. In a second model, a rectangular
inclusion (width 2.0 mm, depth 0.8 mm) with elastic modulus, E = 269 kPa, was positioned
between the circle and model’s upper surface to mimic a stiff inclusion (figure 3). Axial forces
were applied to the circle, resembling those experienced by the implant, so that it underwent
an axial displacement towards the model’s lower surface of ∼23.5 µm.
2.4. Data analysis
Elastograms were produced in Matlab by analysing displacements in the phantoms and using
this information to estimate strain. Displacement tracking was conducted by a normalized
cross-correlation algorithm (Brezinski 2006) with a square reference window 25 pixels wide.
Strain was estimated from the displacements using standard least-squares strain estimation
(LSQSE). Fluctuations in the displacement data can be amplified by this process, so it is
imperative that there is as little noise as possible (Kallel and Ophir 1997). In practice, the
LSQSE was executed in two steps. First, a 15 pixel 1D median filter was applied along
the axis of interest to reduce any high frequency noise in the displacement map. Second, a
20 pixel 1D window is placed around a point of interest in the filtered displacement map
parallel to the axis of interest. A first-order polynomial fit is then applied to the windowed
data, and its gradient recorded as the differential strain value for that location. This process
can be repeated for each element of a displacement map until a full elastogram is produced.
3.1. Finite-element simulations
Displacement maps produced using the FEM software were constructed for an actuation force
of 890 µN in both the control and inclusion phantoms (figure 4). These illustrate the non-
5520A Grimwood et al
Figure 4. Finite-element simulations of the phantoms showing lateral displacement in the control
phantoms (d). Note that the colour scales in (a) and (b) are much smaller than (c) and (d). Colour
bar units are metres.
the inclusion, whilst the axial displacement field broadens to incorporate it. The overall result
is a quantifiable change in the distribution and magnitude of displacement in the presence of a
stiff inhomogeneity. However, due to the field’s non-uniform nature, limited contrast between
the inclusion and its surroundings can be seen in the displacement data.
Analyzing strain data provides a more useful approach. Figure 5 shows the axial normal
strain through the centre of both phantoms where its magnitude is greatest.
The inclusion causes a change in the surrounding strain field, a step change in the strain
occurs at the inclusion boundary and strain outside of the inclusion is intensified. These
features provide useful contrast mechanisms. First, in the presence of an inclusion, strain
inside the stiff region is attenuated whilst the field outside intensifies. Second, the sudden
change in magnitude at the inclusion’s boundary is a means of delineating its geometry.
However, contrast is only generated from normal strain in regions where the measured field is
of sufficient magnitude. This region is confined to a narrow band directly above the implant
which is much smaller than the overall width of the inclusion. Outside of this area, normal
would be compounded.
Calculating the magnitude of total shear strain from its two components (axial shear
strain and lateral shear strain) provides contrast over a wider area. As can be seen in figure 6,
sufficient contrast is generated to distinguish the 2 mm inclusion despite producing a strain
field with an inherent null zone at its centre. The null zone arises where phantom deformation
Elastographic contrast generation in optical coherence tomography from a localized shear stress5521
Figure 5. Axial normal strain through the centres of both control and inclusion phantoms. The
inclusion boundary can be seen at 0.8 mm depth.
is primarily in the direction of motion of the implant. This results in a large normal strain
component, but little or no shear.
3.2. Experimental results
lateral and axial components of motion were constructed using normalized cross-correlation
and these are shown in figure 7. It can be seen from the images that noise is present in the
data, as are horizontal banding artefacts arising from the channel boundaries. Despite this,
there is still general agreement between experimental and simulated data. A region of high
noise is visible along the top left-hand side in the inclusion phantom images, which was the
result of decorrelation arising from lack of scatterers outside of the phantom (as can be seen
in figure 1(b)).
Estimations of axial normal strain are plotted in figure 8. As with those shown for the
simulation data (figure 5), these were calculated using the LSQE method. Banding artefacts
and noise make it impossible to identify the inclusion boundary.
The poor signal for normal strain makes it increasingly desirable that useful shear strain
images could be produced (figure 9). By depicting the magnitude of total shear strain, it
is possible to form an elastogram in which the previously invisible inclusion can now be
distinguished. Its boundaries are visible in figure 9(b), with the exception of the central region
the strain distribution, especially in figures 9(c) and (d). These two elastograms were formed
from lateral displacement data, where the artefacts are again caused by the seams between
each OCT channel in the original images (figure 1).
Tofurtherclarifythetechnique’s performance, theratiobetweenaverage totalshearstrain
and a background reading was plotted. This was achieved by considering the total shear strain
elastograms and measuring their mean background in a 50 × 50 pixel region situated in an
area of low strain. Next, the mean strain for every A-line between a depth of 0.215 mm and
0.645 mm (a distance of 50 pixels) was recorded. The ratio between each A-line’s mean value
and the background measurement was calculated for control and inclusion phantoms using
experimental and FEM data. The results are plotted in figure 10. Figure 10(d) illustrates the
inclusion’s effect on the strain ratio. As would be expected, the experimental data exhibit a
large noise component compared to FE simulations, resulting in smaller peaks at the inclusion
5522 A Grimwood et al
Figure 6. Magnitude values of total shear strain for (a) the control and (b) inclusion phantoms, as
well as the lateral shear strain (c) and (d) and the axial shear strain components (e) and (f).
boundary. In the experimental data, the inclusion was located slightly off-centre, which
accounts for any misalignment with the simulated results.
Strain ratios were also produced for a vertical band bisecting the inclusion and passing
through a region of high shear. This time, instead of taking mean A-line values, the average
strain along each row of pixels within the band was recorded. These results are shown in
figure 11. The peaks in the strain ratio in figure 11(b) and especially 11(d) are the result of
decorrelation and banding artefacts. The artefacts may also account for the undulating, low
frequency variations in the signal, which can clearly be seen in figure 11(b) at the inclusion
The implant produces a non-uniform strain field with both normal and shear components.
Elastograms of total shear proved useful for distinguishing a lesion represented by a stiff
Elastographic contrast generation in optical coherence tomography from a localized shear stress5523
Figure 7. Displacement maps produced using normalized cross-correlation for lateral components
of motion in the control phantom (a) and inclusion phantom (b), and for axial components in the
control (c) and inclusion (d) phantoms. Inset pictures are the corresponding simulations previously
shown in figure 4. Colour bar units are metres.
phantoms. The inclusion boundary is at 0.8 mm depth.
Axial normal strain estimations through the centre of both control and inclusion
inclusion in the tissue phantoms, although a central null zone was present. By comparison,
normal strain was confined to this small area, but suffered from poor signal to noise. It should
is also generated in the central region. There may be advantages to combining normal and
5524 A Grimwood et al
Figure 9. Magnitudes of the total shear strain estimations in (a) the control phantom and (b) the
inclusion phantom, along with the lateral shear strain components (c) and (d) and axial shear strain
components (e) and (f). Inset pictures show the corresponding simulations. Banding artefacts can
be clearly seen in the lateral shear elastograms.
shear strain elastograms in future studies, but only if the signal to noise can be improved,
especially with respect to normal strain data. The simulated axial normal strain (figure 5)
indicated a decrease below 1.3 mm; it is not clear from experimental data whether this actually
occurs. It may arise because the simulated implant was treated as a free body. Similarly, it
is not apparent what causes an increase in normal strain towards the phantom surface; it is
suggested to arise from the surface boundary’s resistance to deformation.
of speckle correlation across channel boundaries. The channels have staggered focal regions
and separate beams, leading to differences in their phase, such that speckle is uncorrelated
between channels (Popescu et al 2007). Another plausible explanation is break-through
Elastographic contrast generation in optical coherence tomography from a localized shear stress5525
background value and show the horizontal band within which mean A-line values were recorded.
Plots of the shear strain ratio across (b) the control and (d) inclusion phantoms are depicted for
both experimental data and FEM simulations.
noise of spikes in the tuneable laser spectrum. Both of these processes reduce the image
tracking algorithm’s effectiveness, leading to horizontal banding in displacement data, which
is accentuated by the LSQSE. A channel blending utility can be used to smooth the transition,
but was observed to have a little effect on image tracking performance. In the future, banding
could be mitigated by processing the strain data from each channel separately and combining
the resultant elatograms into a single image.
The phantom model provided a general representation of melanoma; however, it did not
take into account average values of skin stiffness, viscoelasticity, anisotropy or the extremely
also assumed that the phantom’s compressive modulus and finite-element model’s tensile
5526 A Grimwood et al
Figure 11. Images (a) and (c) illustrate the band in which mean strain values for each row of pixels
were recorded. Graphs (b) and (d) show the change in the strain ratio with depth for control and
inclusion phantoms, respectively. The inclusion boundary in (b) is at 0.8 mm.
modulus were identical. This did not affect the nature of the results and still allowed for
comparisons in a strain pattern to be made between simulated and experimental data. Another
limitation of the phantom model was the magnetic implant. The phantoms were fabricated by
hand; thus, theimplantwasneverperfectlyalignedwiththeinclusion. Also, theHalbacharray
was positioned manually and both of these factors increased the likelihood of out-of-plane
motion leading to decorrelation in the displacement. However, the array’s field density at
a given depth is relatively uniform across the imaging region and no adverse effects were
Decorrelation is still a primary consideration however, especially for in vivo imaging
where Brownian motion and other movement artefacts are increasingly predominant. It is
desirable that speckle motion unrelated to the implant is kept to a minimum; thus, B-scan rates
Elastographic contrast generation in optical coherence tomography from a localized shear stress5527
need to be high. Additionally, implant displacement must be small enough so that it does not
cause decorrelation itself. A maximum axial displacement between scans was observed to be
roughly 7 pixels. There is also an upper limit for scan rate dictated by the signal to noise ratio,
whereby the displacement between two frames must be sufficiently higher than background
Future work could conceivably use an actuated implant, similar to that described here, for
pre-clinical small animal in vivo applications or ex vivo work on cell cultures. It is also worth
noting that contrast generation from highly localized, non-uniform displacements has been
the subject of previous research in both ultrasound and OCT communities (Oldenburg and
Boppart 2010, Melodelima et al 2007). However, using total shear strain to generate contrast
is especially applicable to techniques, which incorporate non-uniform strain distributions with
a shear component, such as acoustic radiation force imaging (ARFI) where a small region of
tissue is excited using a focused acoustic pulse (Dahl et al 2009).
The study demonstrates a technique for generating contrast from a highly localized, non-
uniform stress. A silicone rubber phantom with a stiff inclusion 2.0 mm × 0.8 mm in
size was used to represent a melanoma skin lesion model. The inclusion was shown to be
distinguishable from shear–strain elastograms despite not being visible in standard, multi-
channel OCT images. Finite-element simulations were also in agreement with experimental
results. Banding artefacts and high noise restricted performance, but some of these issues
can be addressed in future by processing each OCT channel separately. Potential applications
for the imaging technique have been discussed and include pre-clinical in vivo work. Other
methods of actuation, such as indentation, or highly localized magnetic nanoparticles could
also be adopted to reduce invasiveness. There might also be further potential for using shear
strain elastography with OCT to quantify the degree of bonding between a lesion and its
surroundings, which would in turn help ascertain its malignancy (Konofagou et al 2000).
This work was funded jointly by the Engineering and Physical Sciences Research Council
(EPSRC) and Michelson Diagnostics Ltd through an EPSRC Industrial CASE award. We
would like to thank the National Physical Laboratory’s (NPL) biophotonics group for their
help with phantom production. Our gratitude also goes to those members of the optical
imaging team at the Institute of Cancer Research (ICR) who provided feedback on the study.
Finally, we would like to acknowledge the staff at Michelson Diagnostics Ltd for supporting
our use of their equipment.
Bisaillon C-E, Lamouche G, Maciejko R, Dufour M and Monchalin J-P 2008 Deformable and durable phantoms with
controlled density of scatterers Phys. Med. Biol. 53 N237–47
Brezinski M E 2006 Optical Coherence Tomography: Principles and Applications (Burlington, MA: Academic)
Crecea V, Oldenburg A L, Liang Xing, Ralston T S and Boppart S A 2009 Magnetomotive nanoparticle transducers
for optical rheology of viscoelastic materials Opt. Express 17 23114–22
Dahl J J, Dumont D M, Allen J D, Miller E M and Trahey G E 2009 Acoustic radiation force impulse imaging
for non-invasive characterization of carotid artery atherosclerotic plaques: a feasibility study Ultrasound Med.
Biol. 35 707–16
5528 A Grimwood et al Download full-text
Emelianov S Y, Aglyamov S R, Shah J, Sethuraman S, Scott W G, Schmitt R, Motamedi M, Karpiouk A and
Oraevsky A 2004 Combined ultrasound, optoacoustic and elasticity imaging Proc. SPIE 5320 101–12
Filas B A, Efimov I R and Taber L A 2007 Optical coherence tomography as a tool for measuring morphogenetic
deformation of the looping heart Anat. Rec. 290 1057–68
Garra B S et al 1997 Elastography of breast lesions: initial clinical results Radiology 202 79–86
Kallel F and Ophir J 1997 A least-squares strain estimator for elastography Ultrason. Imaging 19 195–208
Khatyr F, Imberdis C, Vescovo P, Varchon D and Lagarde J-M 2004 Model of the viscoelastic behaviour of skin in
vivo and study of anisotropy Skin Res. Technol. 10 96–103
Kim J-S et al 2009 Optical coherence tomography evaluation of zotarolimus-eluting stents at 9-month follow-up:
comparison with sirolimus-eluting stents Heart 95 1907–12
Kirkpatrick S J, Wang R K, Duncan D D, Kulesz-Martin M and Lee K 2006 Imaging the mechanical stiffness of skin
lesions by in-vivo acousto-optical elastography Opt. Express 14 9770–9
Konofagou E E, Harringan T and Ophir J 2000 Shear strain estimation and lesion mobility assessment in elastography
Ultrasonics 38 400–4
for measuring material mechanical properties Opt. Lett. 34 2894–6
Manner J, Thrane L, Norozi K and Yelbuz T M 2009 In vivo imaging of the cyclic changes in cross-sectional
shape of the ventricular segment of pulsating embryonic chick hearts at stages 14 to 17: a contribution to the
understanding of the ontogenesis of cardiac pumping function Dev. Dyn. 238 3273–84
Melodelima D, Bamber J C, Duck F A and Shipley J A 2007 Transient elastography using impulsive ultrasound
radiationforce: apreliminarycomparisonwithsurfacepalpationelastographyUltrasoundMed.Biol. 33 959–69
Muthupillai R, Lomas D, Rossman P, Greenleaf J, Manduca A and Ehman R 1995 Magnetic resonance elastography
by direct visualization of propagating acoustic strain waves Science 26 1854–7
Oldenburg A L and Boppart S A 2010 Resonant acoustic spectroscopy of soft tissues using embedded magnetomotive
nanotransducers and optical coherence tomography Phys. Med. Biol. 55 1189
Oldenburg A L, Crecea V, Rinne S A and Boppart S A 2008 Phase-resolved magnetomotive OCT for imaging
nanomolar concentrations of magnetic nanoparticles in tissues Opt. Express 16 11525–39
Podoleanu AGh and Rosen R B 2008 Combinations of techniques in imaging the retina with high resolution Prog.
Retin. Eye Res. 27 464–99
Popescu D P, Hewko M D and Sowa M G 2007 Speckle noise attenuation in optical coherence tomography by
compounding images acquired at different positions of the sample Opt. Commun. 269 247–51
Rogowska J, Patel N, Plummer S and Brezinski M E 2006 Quantitative optical coherence tomographic elastography:
method for assessing arterial mechanical properties Br. J. Radiol. 79 707–11
Schmitt J M 1998 OCT elastography: imaging microscopic deformation and strain of tissue Opt. Express 3 199–211
Tanter M et al 2008 Quantitative assessment of breast lesion viscoelasticity: initial results using supersonic shear
imaging Ultrasound Med. Biol. 34 1373–86
Thitaikumar A, Krouskop T A, Garra B S and Ophir J 2007 Visualization of bonding at an inclusion boundary using
axial-shear strain elastography: a feasibility study Phys. Med. Biol. 52 2615–33
Tomlins P H, Ferguson R A, Hart C and Woolliams P D 2009 Point-spread function phantoms for optical coherence
tomography NPL Report OP 2
Tomlins P H and Wang R K 2005 Theory, developments and applications of optical coherence tomography J. Phys.
D: Appl. Phys. 38 2519–35
Ziolkowska M, Philipp C M, Liebscher J and Berlien H P 2009 OCT of healthy skin, actinic skin and NMSC lesions
Med. Laser Appl. 24 256–64