High-resolution ultrasonic imaging using an etalon detector array
Sheng-Wen Huang,1,a?Yang Hou,2Shai Ashkenazi,1and Matthew O’Donnell3
1Department of Biomedical Engineering, University of Michigan, Ann Arbor, Michigan 48109, USA
2Department of Electrical Engineering and Computer Science, University of Michigan,
Ann Arbor, Michigan 48109, USA
3Department of Bioengineering, University of Washington, Seattle, Washington 98195, USA
?Received 15 June 2008; accepted 26 August 2008; published online 15 September 2008?
A photoacoustic imaging system was built and tested to demonstrate the feasibility of
high-resolution low-noise ultrasonic imaging based on parallel detection using polymer etalons. Its
capability of detecting ultrasound at different elements simultaneously in the optical end was
verified by imaging three 49 ?m beads. An average noise-equivalent pressure of 3.6 kPa over 50
MHz for 50 ?m diameter detection elements in a two-dimensional array with a diameter of 1.35
mm and a detection bandwidth of 75 MHz at –3 dB was measured. These results demonstrate the
potential of polymer etalons for high-frame-rate high-resolution three-dimensional photoacoustic
and ultrasound pulse-echo imaging. © 2008 American Institute of Physics.
High-frequency ?above 20 MHz? ultrasonic imaging,
including pulse-echo1,2and photoacoustic ?also called optoa-
coustic? imaging,3,4has been applied to applications demand-
ing high resolution. However, no two-dimensional ?2D?
piezoelectric array has been realized for real-time high-
resolution three-dimensional ?3D? ultrasonic imaging. In
fact, high-frequency piezoelectric 2D arrays would naturally
suffer from increased noise levels and wiring and fabrication
complexities due to large element count and small element
size and spacing. To avoid these issues, one way is to detect
and generate ultrasound optically.5
Three properties make resonant optical ultrasound
transducers ?ROUTs?5–17suitable for realizing dense large-
element-count high-frequency 2D detection arrays. First, the
signal-to-noise ratio ?SNR? or sensitivity of a ROUT does
not directly depend on its element size and can be improved
by increasing the probing light. Second, no complex wiring
and electromagnetic interference exist around a ROUT
array. Third, the space for optoelectrical transduction and
electronics is not limited by the array size or element spac-
ing. In short, a ROUT array’s performance and complexity
are irrelevant to its element size.
Polymer Fabry–Pérot etalons are a special type of
ROUT.5,10–14Their basic structure is a thin polymer layer
with semitransparent mirrors coated on both its sides. The
optical reflectivity of an etalon is a function of wavelength.
When the probing light experiences a round-trip phase shift
of 2m?, where m is an integer, in the etalon, resonance oc-
curs and the reflectivity drops. In the presence of ultrasound,
pressure modulates the etalon thickness and therefore the
phase shift. At a wavelength near resonance, the phase
modulation is transformed into reflectivity modulation with
high gain, and ultrasound detection can be performed by
measuring the reflected light power as a function of time.
An ultrasound detection element can be formed on an
etalon by focusing the probing light onto an area of interest.
Such optically defined elements can be as small as the order
of 10 ?m. To form an array, different areas can be probed in
sequence.10In this way, however, imaging frame rate is
greatly limited. Therefore, to achieve high frame rate, dis-
tributing the probing light at a fixed wavelength over a large
area to enable parallel detection is preferred.13,14Note that
although charge-coupled device ?CCD? arrays can be used
for parallel detection at a time instant,9,12they are not suit-
able for high-frame-rate imaging where low noise and
parallel detection at ?1000 instants ?i.e., 1000 consecutive
time samples? are required. Element-by-element wavelength
adjustment11is also infeasible in this case.
In this study, a photoacoustic imaging system was built
and tested to demonstrate the feasibility of high-resolution
low-noise ultrasonic imaging based on parallel detection us-
ing polymer etalons. As shown in Fig. 1?a?, the output light
from a continuous-wave tunable laser ?HP 8168F, Agilent
Technologies, Santa Clara, CA? was approximately colli-
mated by a lens ?L1? to illuminate an etalon. A unity-
magnification two-lens system comprised of L1 and L2
mapped the light reflected from the etalon onto a photode-
tector plane. The etalon was made on a glass substrate by
coating in sequence a 30 nm gold mirror, a 6 ?m SU-8
?MicroChem Corp., Newton, MA? polymer layer, another 30
nm gold mirror, and an additional 1.5 ?m SU-8 protection
As shown in Fig. 1?b?, one tip of a fiber with a 50 ?m
?in diameter? core was put on the photodetector plane to
deliver light from a 50 ?m element on the etalon surface to
a photoreceiver ?1811-FC, New Focus, San Jose, CA?. A
2D translation stage, driven by two motorized actuators
?T-LA60, Zaber Technologies Inc., Richmond, BC, Canada?,
scanned this tip to emulate a photodetector array. Detecting
a?Electronic mail: email@example.com.
FIG. 1. ?Color online? Setup for photoacoustic imaging. BS, PD, and L
stand for beamsplitter, photodetector, and lens, respectively.
APPLIED PHYSICS LETTERS 93, 113501 ?2008?
0003-6951/2008/93?11?/113501/3/$23.00© 2008 American Institute of Physics
light using this array is equivalent to sensing ultrasound
on a corresponding array on the etalon surface. The photore-
ceiver output was amplified by 30 dB ?AU-1310–1103-BNC,
MITEQ, Hauppauge, NY? and then digitized using an oscil-
loscope ?WaveSurfer 432, LeCroy, Chestnut Ridge, NY?.
Note that the system’s optical end is capable of parallel prob-
ing and image quality will remain the same if a physical
photodetector array of the same quality as the photoreceiver
in the system is used to replace mechanical scanning to en-
able high-frame-rate imaging.
We estimated the light intensity distribution on the ultra-
sonic array plane ?i.e., the etalon surface? by measuring that
on the photodetector plane at an off-resonance optical wave-
length, 1560 nm. The result is shown in Fig. 2?a?. The
−3 dB spot size was 1.3 mm in diameter. The reflection
spectra measured at the positions indicated by circles in this
figure are shown in Fig. 2?b?. The resonance wavelengths at
different positions were close to each other, meaning that this
etalon has a uniform thickness. We then defined a circular
array whose element positions are indicated by x-marks in
Fig. 2?a?. This array has 373 elements, an element spacing of
50 ?m, a diameter of ?1.05 mm, and an optical quality
factor of ?340. To quantify the smoothness of the array, we
fitted the light intensity distribution at the circle positions to
a 2D Gaussian function and measured the root-mean-squared
?RMS? fitting error to be 13% of the mean intensity. Accord-
ing to the spectra shown in Fig. 2?b?, 1545.2 nm was picked
as the probing wavelength across the whole array.
A photoacoustic imaging experiment was conducted us-
ing the imaging system. Three black polystyrene beads
?BK050, Microgenics Corp., Fremont, CA? with a diameter
of 49 ?m were fixed in a gel made from water and 1%
agarose ?GPG/LE, American Bioanalytical, Natick, MA?.
This phantom was then put in de-ionized water with the
beads ?1.1 mm from the etalon. A 532 nm pulsed laser
?Surelite I-20, Continuum, Santa Clara, CA? illuminated the
beads at 28 mJ/cm2to generate photoacoustic signals,
which were recorded by the circular array with a total prob-
ing power of 5 mW. All element data were bandpass filtered
over 20–36 MHz and then inputted to a delay-and-sum
beamforming algorithm with envelope detection to recon-
struct a 3D image.
Figure 3?a? shows the –6 dB isosurface of the image.
The ?x,y,z? coordinate system is defined in Fig. 1?a? with
the center of the array as its origin. The central positions
of the beads were ?114, 194, 1034?, ?56, 60, 1132?, and
?206,106,1174? ?m, and the –6 dB ?relative to individual
peaks? widths of the bead images were ?84, 84, 98?, ?82, 94,
90?, and ?100,90,94? ?m. The –6 dB length of the band-
pass filter was 53 ns, leading to a lower bound of 78 ?m to
the –6 dB axial ?z? resolution. The –6 dB lateral resolution
of a circular array can be estimated as 1.41??F number?
??acoustic central wavelength?,18ignoring apodization due
to light distribution. Therefore, the system’s –6 dB lateral
?x,y? resolutions were lower bounded to 73, 80, and 83 ?m
at the beads’ positions. Considering the beads’ physical di-
ameter and the system’s resolution, the image sizes are rea-
sonable. A 2D image on the plane crossing the central posi-
tions of the beads is shown in Fig. 3?b?.
To improve SNR to a reasonable level, we incorporated
an erbium-doped fiber amplifier ?EDFA? ?KPS-BT2-C-30-
SLM-PM-PB-FA, Keopsys, Lannion, France? into the optical
path. The EDFA was driven by the tunable laser and output-
ted 1 W light to probe the etalon with a −3 dB spot size of
1.3 mm in diameter. The electronic amplifier following the
photoreceiver was removed. The noise-equivalent pressure
FIG. 2. ?Color online? ?a? Estimated light intensity distribution on the etalon
surface at 1560 nm, an off-resonance optical wavelength. The −3 dB spot
size was 1.3 mm in diameter. ?b? Reflection spectra measured at the posi-
tions indicated by circles in ?a? indexed row by row from the top-left corner.
FIG. 3. ?Color online? ?a? –6 dB isosurface of a 3D photoacoustic image of
three 49 ?m beads. ?b? A 2D image on the plane crossing the central posi-
tions of the beads displayed over a 15 dB dynamic range. The probing
wavelength was 1545.2 nm across the whole imaging array indicated by
x-marks in Fig. 2?a?.
113501-2Huang et al. Appl. Phys. Lett. 93, 113501 ?2008?
?NEP?, a measure of the minimum detectable pressure,
of a 50 ?m diameter element at the spot center was mea-
sured using a calibrated transducer.17With 0.294 mW dc
light to the photoreceiver, this detection element measured
28 mV ac at 30 kPa and an RMS noise level of 2.9 mV over
10–60 MHz. Therefore, its sensitivity and NEP over the 50
MHz band were ?28 mV?/?30 kPa?=0.93 mV/kPa and
?2.9 mV?/?0.93 mV/kPa?=3.1 kPa, respectively.
Using the above measurement as a reference, we esti-
mated NEPs at other light powers. Given a light power Pdc,
only the corresponding RMS noise level n?Pdc? was mea-
sured since the sensitivity is proportional to Pdc. Specifically,
NEP?Pdc?=n?Pdc?/??0.93 mV/kPa??Pdc/?0.294 mW??.
Figure 4 shows the monotonically decreasing function NEP
?Pdc?. At low light levels where thermal noise is the domi-
nant noise, NEP is inversely proportional to Pdc. The de-
crease in NEP with Pdcgradually lessens due to increased
shot noise and laser noise. Fitting the measured n2?Pdc? to a
second degree polynomial led to n2?Pdc???0.63 mV2?
+?13.7 mV2/mW?Pdc+?43.0 mV2/mW2??Pdc?2with a co-
efficient of determination R2of 0.9998.
Based on NEP?Pdc?, the average NEP of 50 ?m
elements across the central 1.35 mm diameter area of the
light spot was estimated to be 3.6 kPa over 50 MHz, not
only a two-order improvement over a CCD approach9
but also lower than the NEP of a 75 ?m piezoelectric
PVDF ?polyvinylidene fluoride? transducer ?HPM075/1,
which isatleast4.2 kPa ?=?60 ?V????50 MHz?/
?100 MHz??1/2/?10 nV/Pa?? over the same bandwidth con-
sidering only the noise from its matched preamplifier ?HP1,
Precision Acoustics?. If a smaller element diameter such as
25 ?m is required, the same level of NEP can be obtained
by adjusting the probing light intensity accordingly.
NEP can be reduced by reducing noise or improving
sensitivity. By monitoring the EDFA output using an inde-
pendent photodetector, the second degree term in the poly-
nomial noise model can be greatly eliminated. To make more
sensitive etalons, softer polymers can be used. For example,
according to our experience, polydimethylsiloxane etalons
are more than twice as sensitive as SU-8 etalons. Sensitivity
can also be improved by increasing etalons’ quality factor.
Using a previously developed method,17the detection
bandwidth of the system was measured to be 75 MHz at
–3 dB, which can provide three- to fourfold better resolution
compared to previous CCD-based systems.9,12In the imaging
experiment, the main band of signals was 0–40 MHz. There-
fore, we did not choose a filter with a higher band to achieve
better resolution. Nonetheless, according to the 75 MHz
bandwidth, for small objects higher resolutions can be
achieved using reasonable F numbers without sacrificing
SNR. For example, a bandpass filter over 40–72 MHz to-
gether with a unity F number leads to –6 dB resolutions of
less than 40 ?m in all directions for photoacoustic imaging.
The resolution of ultrasound pulse-echo imaging can be even
better because of two-way focusing, especially in the axial
direction where less than 20 ?m is expected.
The system described above can perform high-resolution
3D photoacoustic imaging with a reasonable SNR. Incorpo-
rating a photodetector array will immediately provide high
frame rates, theoretically equal to the repetition frequency of
the pulsed laser. To enable ultrasound imaging, etalons may
be integrated with ultrasound transmitters such as a gold
Support from NIH ?Grant No. EB-003455? is gratefully
1M. Vogt, K. Kaspar, P. Altmeyer, K. Hoffmann, and S. El Gammal,
Frequenz 55, 12 ?2001?.
2D. J. Coleman, R. H. Silverman, A. Chabi, M. J. Rondeau, K. K. Shung, J.
Cannata, and H. Lincoff, Ophthalmology 111, 1344 ?2004?.
3R. J. Zemp, R. Bitton, M.-L. Li, K. K. Shung, G. Stoica, and L. V. Wang,
J. Biomed. Opt. 12, 010501 ?2007?.
4S. Sethuraman, J. H. Amirian, S. H. Litovsky, R. W. Smalling, and S. Y.
Emelianov, Opt. Express 15, 16657 ?2007?.
5Y. Hou, J.-S. Kim, S. Ashkenazi, S.-W. Huang, L. J. Guo, and M.
O’Donnell, Appl. Phys. Lett. 91, 073507 ?2007?.
6J.-P. Monchalin, Appl. Phys. Lett. 47, 14 ?1985?.
7J. D. Hamilton, T. Buma, M. Spisar, and M. O’Donnell, IEEE Trans.
Ultrason. Ferroelectr. Freq. Control 47, 160 ?2000?.
8V. Wilkens, J. Acoust. Soc. Am. 113, 1431 ?2003?.
9M. Klann and C. Koch, IEEE Trans. Ultrason. Ferroelectr. Freq. Control
52, 1546 ?2005?.
10S. Ashkenazi, Y. Hou, T. Buma, and M. O’Donnell, Appl. Phys. Lett. 86,
11P. C. Beard, IEEE Trans. Ultrason. Ferroelectr. Freq. Control 52, 1002
12M. Lamont and P. C. Beard, Electron. Lett. 42, 187 ?2006?.
13S.-W. Huang, Y. Hou, S. Ashkenazi, and M. O’Donnell, Proc.-IEEE Ul-
trason. Symp. 2007, 719 ?2007?.
14S.-W. Huang, Y. Hou, S. Ashkenazi, and M. O’Donnell, Proc. SPIE 6856,
15C.-Y. Chao, S. Ashkenazi, S.-W. Huang, M. O’Donnell, and L. J. Guo,
IEEE Trans. Ultrason. Ferroelectr. Freq. Control 54, 957 ?2007?.
16A. Maxwell, S.-W. Huang, T. Ling, J.-S. Kim, S. Ashkenazi, and L. J.
Guo, IEEE J. Sel. Top. Quantum Electron. 14, 191 ?2008?.
17S.-W. Huang, S.-L. Chen, T. Ling, A. Maxwell, M. O’Donnell, L. J. Guo,
and S. Ashkenazi, Appl. Phys. Lett. 92, 193509 ?2008?.
18X. Chen, K. Q. Schwarz, and K. J. Parker, J. Acoust. Soc. Am. 94, 2979
19See http://www.acoustics.co.uk/products/hpm075-1 for specification in-
20Y. Hou, J. S. Kim, S. Ashkenazi, M. O’Donnell, and L. J. Guo, Appl.
Phys. Lett. 89, 093901 ?2006?.
FIG. 4. ?Color online? NEP vs power of detection light.
113501-3 Huang et al. Appl. Phys. Lett. 93, 113501 ?2008?