of 8
214
Vol. 11, No. 2 / February 2024 /
Optica
Research Article
Acoustic-feedback wavefront-adapted
photoacoustic microscopy
Yuecheng Shen,
1
,
Jun Ma,
2
,
Chengtian Hou,
2
,
Jiayu Zhao,
3
Yan Liu,
4
,
7
Hsun-Chia Hsu,
5
Terence T. W. Wong,
6
Bai-Ou Guan,
2
Shian Zhang,
1
,
8
AND
Lihong V. Wang
5
,
9
1
State Key Laboratory of Precision Spectroscopy, School of Physics and Electronic Science, East China Normal University, Shanghai 200241, China
2
Guangdong Provincial Key Laboratory of Optical Fiber Sensing and Communications, College of Physics & Optoelectronic Engineering, Jinan
University, Guangzhou 510632, China
3
School of Electronics and Information Technology, Sun Yat-Sen University, Guangzhou 510275, China
4
School of Optometry, Indiana University, Bloomington, Indiana 47405, USA
5
Caltech Optical Imaging Laboratory, Andrew and Peggy Cherng Department of Medical Engineering, Department of Electrical Engineering,
California Institute of Technology, 1200 East California Boulevard, Pasadena, California 91125, USA
6
Translational and Advanced Bioimaging Laboratory, Department of Chemical and Biological Engineering, Hong Kong University of Science and
Technology, Hong Kong SAR, China
7
yl144@iu.edu
8
sazhang@phy.ecnu.edu.cn
9
lvw@caltech.edu
These authors contributed equally to this work.
Received 8 November 2023; revised 22 December 2023; accepted 7 January 2024; published 5 February 2024
Optical microscopy is indispensable to biomedical research and clinical investigations. As all molecules absorb light,
optical-resolution photoacoustic microscopy (PAM) is an important tool to image molecules at high resolution without
labeling. However, due to tissue-induced optical aberration, the imaging quality degrades with increasing imaging
depth. To mitigate this effect, we develop an imaging method, called acoustic-feedback wavefront-adapted PAM (AWA-
PAM), to dynamically compensate for tissue-induced aberration at depths. In contrast to most existing adaptive optics
assisted optical microscopy, AWA-PAM employs acoustic signals rather than optical signals to indirectly determine the
optimized wavefront. To demonstrate this technique, we imaged zebrafish embryos and mouse ears
in vivo
. Experimental
results show that compensating for tissue-induced aberration in live tissue effectively improves both signal strength and
lateral resolution. With this capability, AWA-PAM reveals fine structures, such as spinal cords and microvessels, that
were otherwise unidentifiable using conventional PAM. We anticipate that AWA-PAM will benefit the
in vivo
imaging
community and become an important tool for label-free optical imaging in the quasi-ballistic regime.
© 2024 Optica
Publishing Group under the terms of the Optica Open Access Publishing Agreement
https://doi.org/10.1364/OPTICA.511359
1. INTRODUCTION
Optical microscopy is crucial to many science and engineering
fields. As all molecules absorb light, photoacoustic (PA) micros-
copy (PAM) is an important technique that employs non-ionizing
photons and low-scattering ultrasound to image molecules. By
offering universal optical absorption contrast, deep penetra-
tion, and label-free capability, PAM is suitable for a wide range
of biomedical applications, such as functional imaging of blood
oxygenation, tracking of circuiting tumor cells, lipid imaging,
and label-free histology [1–6]. PAM involves both optical exci-
tation and acoustic detection, thereby providing a great diversity
of embodiments. When biological tissue is probed with a pulsed
and focused light beam, acoustic waves are generated due to
absorption-induced heat generation, thus revealing cellular or
even sub-cellular structures with high spatial resolution deter-
mined by the optical diffraction limit. This modality, termed
optical-resolution photoacoustic microscopy (OR-PAM), was
first developed in 2008 to image microvasculatures in mice
in vivo
[7]. However, biological tissue is inherently heterogeneous, which
distorts the optical wavefront and causes optical aberration as light
propagates through. Since the performance of OR-PAM strongly
relies on the quality of the excitation focus, it inevitably suffers
from degradations in both PA amplitude and spatial resolution
while imaging at depths.
In general, optical aberration deteriorates the imaging per-
formance of all optical microscopy. To alleviate this issue, adaptive
optics (AO), which was initially developed in astronomy to com-
pensate for atmosphere-induced aberration, has been introduced
into optical microscopy. By actively modulating the incident
2334-2536/24/020214-08 Journal © 2024 Optica Publishing Group
Research Article
Vol. 11, No. 2 / February 2024 /
Optica
215
wavefront to compensate for both system-generated and tissue-
induced aberration, AO has become an effective tool for improving
image quality. To date, AO has been widely adopted in widefield
microscopy [8–13] and point-scanning microscopy [14–23]
such as confocal and multiphoton microscopy. There are a variety
of implementations of AO schemes to determine the distorted
wavefront, including measuring the optical wavefront directly
by using a Shack-Hartmann (SH) sensor [9,11,15,16,19,24–26]
and reconstructing the optical wavefront indirectly through pupil
segmentation approaches [17,18,27–29], which have been well
discussed and summarized in recent reviews [30–33]. In addition,
optical aberration can also be determined by iteratively adjusting
the wavefront until a certain image metric is optimized [34–42].
In all these existing implementations, optical measurements con-
cerning either back-reflected excitation light or emitted fluorescent
light are required to determine the optical aberration.
OR-PAM relies on optical focusing to achieve high lateral reso-
lution and acoustic time-of-arrival to achieve high axial resolution;
it produces a one-dimensional (1D) depth-resolved image (an A-
line) per laser pulse. Therefore, the quality of the excitation optical
focus is crucially important to the performance of OR-PAM. The
first attempt to introduce AO into OR-PAM was in 2010, which
aims to correct system-generated aberration [43]. By employing
an SH sensor to directly measure the backscattered light from a
white paper and a deformable mirror (DM) to compensate for
the system-generated aberration, significant improvements in
both signal strength and lateral resolution were demonstrated
[43]. In 2022, AO was employed to compensate for the spherical
aberration caused by the mismatch of refractive indices between
water and a target sample, by using a three-layer liquid crystal
device optimized for correcting this specific aberration mode [44].
Despite these efforts, however, using AO to correct tissue-induced
aberration in OR-PAM has not been demonstrated. This situation
may be due to the lack of appropriate guide stars in PAM. On one
hand, back-scattered excitation light does not function well, as the
back-reflected light may come from different regions from the focal
volume. On the other hand, the widely used fluorescent signals do
not naturally exist in PAM. To fill this void, we develop acoustic-
feedback wavefront-adapted PAM (AWA-PAM), which employs
acoustic feedback to correct for tissue-induced optical aberration
in OR-PAM. In contrast to other AO assisted three-dimensional
(3D) optical microscopy techniques that generally abandon the
correction for defocus to avoid axial shifting [45], AWA-PAM
delightfully takes the correction of defocus into consideration.
This choice is because the depth information of PAM is uniquely
determined by the time of arrival of the acoustic waves. In this
condition, optimizing the depth of the focal plane for each A-line
can considerably benefit the imaging of small features at different
depths across the field of view. To demonstrate the feasibility of
AWA-PAM, we built a microscope system by integrating a liquid
crystal based spatial light modulator and an OR-PAM system. The
tissue-induced aberration was dynamically compensated point-by-
point by optimizing the phase map displayed on the SLM with a
greedy algorithm. We will show in the next section that AWA-PAM
effectively corrects for tissue-induced aberration when imaging
in
vivo
zebrafish embryos and mouse ears and significantly improves
the image quality, revealing microstructures that are indiscernible
with conventional PAM.
2. RESULTS
A. Principle of AWA-PAM
We start by describing the operational principle of AWA-APM,
which is schematically shown in Fig. 1. In conventional PAM, as
shown in Fig. 1(a), pulsed light is focused into the biological tissue
to locally induce ultrasonic waves, which are subsequently detected
by a focused ultrasonic transducer through confocal geometry. The
peak-to-valley value of the measured ultrasonic wave, defined as
Ult
ras
onic
transducer
Lens
Microv
essels
PA signal
(a)
Tissue
-ind
uced
aberration
Ultrasonic
transducer
Lens
PA signal
(b)
Modulated
wavefront
Feedback
loop
Com
puter
Com
puter
Microv
essels
Spatial
light
modulator
Conv
ent
ional
PAM
AWA-P
AM
Fig.1.
Schematics of the operational principle of AWA-PAM. (a) In conventional PAM, pulsed light is focused into the biological tissue to locally induce
ultrasonic waves, which are measured by a focused ultrasonic transducer. The heterogeneity of the biological tissue distorts the wavefront of the focused
light, decreasing the signal strength and resolution. (b) In AWA-PAM, a spatial light modulator modulates the wavefront of the pulsed excitation light to
compensate for tissue-induced aberration. The phase map displayed by the SLM is the inverse phase of the distorted wavefront, thereby nullifying the dis-
tortion and creating a sharp focus at depths. To obtain the desired phase map, a feedback loop between the SLM, the computer, and the ultrasonic trans-
ducer is established to optimize the PA amplitude.
Research Article
Vol. 11, No. 2 / February 2024 /
Optica
216
the PA amplitude, is generally used to reflect the local absorption of
the tissue within the optical focus. In this condition, the full width
at half maximum (FWHM) size of the optical focus determines the
lateral resolution. In practice, however, optical aberration aggre-
gates as the excitation light propagates due to the heterogeneity of
biological tissue, which broadens the focus and deteriorates the res-
olution. Moreover, as light is not tightly focused and energy is not
highly concentrated, the measured PA amplitude also decreases for
small features. As a result, tissue-induced aberration deteriorates
both lateral resolution and signal strength.
Inspired by the effectiveness of adaptive optics in fighting
against tissue-induced optical aberration in optical microscopy, we
developed AWA-PAM to effectively mitigate this issue, as shown
in Fig. 1(b). The key enabling point is to modulate the excitation
light by using an SLM before it enters the tissue. One would expect
that tissue-induced aberration can be compensated by modulat-
ing the incident wavefront, resulting in a sharp focus at depths.
Due to the special detecting scheme with ultrasound, AWA-PAM
adopts an indirect wavefront sensing approach by employing
acoustic feedback to estimate the desired phase map through a
greedy optimization algorithm. To begin with, the SLM displays
a planar phase map such that AWA-PAM effectively functions as
conventional PAM. Then, a series of ordered Zernike polynomials
are loaded to the SLM in order. These Zernike polynomials are
orthogonal to each other over circular pupils [46], and each one
effectively represents one type of optical aberration. A feedback
loop is established among the SLM, the ultrasonic transducer,
and the computer. The corresponding coefficients for each order
of the Zernike polynomials are traversed and determined in an
ergodic manner [47,48]. This feedback loop guarantees continu-
ous enhancement in PA amplitudes after correcting aberrations for
each order, leading to the formation of a high-quality focus eventu-
ally. The AWA-PAM is relatively simple and does not require direct
optical wavefront sensing to determine tissue-induced aberration,
which is particularly suitable for PAM with ultrasonic detec-
tion. This process is essentially similar to that of feedback-based
wavefront shaping with acoustic feedback [49–52]. However,
the scattering process inside biological tissue is almost random,
which requires a large number of modes to be tested to generate a
focus with acceptable quality. Thus, this inefficient method makes
it almost impossible to implement for
in vivo
studies. In com-
parison, optical aberrations are known to be well-represented by
the Zernike polynomials. In this condition, the feedback process
is quite efficient, as one needs to consider only a few low-order
Zernike polynomials to account for optical aberration. Therefore,
point-by-point compensation for spatially inhomogeneous optical
aberration can be realized even for live tissue. Furthermore, some of
the previous works allow only focusing light to indefinite locations
and are not suitable for 3D imaging, unlike AWA-PAM.
B. Compensating for System Aberration
We first compensate for the system aberration by imaging a car-
bon fiber. The experimental setup of AWA-PAM is schematically
shown in Fig. 2(a) with a detailed description in Appendix A. To
guarantee sufficient accuracy under static conditions, the step
size for traversing the coefficients for a given Zernike mode was
set to 0.2
π
rads, and the range was from
10
π
to
+
10
π
rads.
Phase wrapping was performed. Using PA amplitudes as feedback,
Fig. 2(b) shows the signal enhancement contributed from each
order. Among them, the 1st order (i.e., piston) has no effect, while
the 2nd and 3rd orders that correspond to tip and tilt are the two
largest contributors. These two corrections mainly account for the
misalignment of the SLM. Besides, the correction for the 4th order
that corresponds to defocus helps, possibly due to correction for
the curvature of the SLM, the use of a lens pair and depth-induced
aberration. Corrections for the Zernike polynomials from the 5th
order to the 8th order also provide certain contributions. In con-
trast, the Zernike polynomials after the 10th order have negligible
effects, indicating that high-order aberration does not exist in the
current system. Overall, compensating for system aberration effec-
tively enhances PA amplitudes by about three times. The image
of the focus shown in Fig. 2(c) confirms that, in free space, the
lateral resolution can be brought back to the theoretically predicted
3
μ
m after correcting for the system aberration. Since system-
generated aberration is static, a fixed phase map corresponding
to correcting the system aberration is always used as the initial
phase map before subsequent wavefront optimization to correct for
sample-induced aberration.
C. Imaging Spinal Cords of Zebrafish Embryos
In Vivo
The feasibility of AWA-PAM to compensate for tissue-induced
aberration was demonstrated through imaging spinal cords of
zebrafish embryos
in vivo
. The zebrafish used for imaging is two
days after fertilization. After anesthetizing the zebrafish, we first
performed whole-body imaging through conventional PAM,
which is shown in Fig. 3(a). Then, we proceeded to execute AWA-
PAM to demonstrate its superiority in revealing structures that
were hindered by optical aberrations. Given that imaging the
whole fish is time-consuming, wavefront correction was performed
on a smaller region that mainly contains the spinal cord, which is
denoted by the white dashed square in Fig. 3(a). To avoid causing
artifacts due to beam drifting, we restrict the orders of the Zernike
polynomials from 4 to 10 (i.e., excluding tip, tilt, and piston)
during
in vivo
experiments. The step size of the coefficient for
traversing each Zernike order was set to
π
rads, and the optimiza-
tion range was from
4
π
to
+
4
π
rads. The isoplanatic patch
size was measured to be 15
×
15
μ
m
2
(with the quantification
procedures described in Supplement 1, Note 1) and we used the
same corrective wavefront for the measurements taken within the
isoplanatic patch. Higher Zernike orders, smaller step sizes, finer
isoplanatic patches, and larger optimization ranges can be targeted,
which is at the cost of increased imaging time and will be discussed
later. Figures 3(b) and 3(c) show the obtained 3D image and its
corresponding 2D maximum amplitude projection (MAP) image
of the spinal cords of the zebrafish embryo by employing AWA,
respectively. As a comparison, the obtained 3D image and 2D
MAP image for the same region without using AWA correction are
shown in Figs. 3(d) and 3(e), respectively. It is clear that AWA-PAM
provides richer structural information than that obtained without
AWA correction. Notably, AWA correction includes the adjust-
ment of the 4th order of the Zernike polynomials that represents
defocusing, allowing spinal cords at different depths to be identi-
fied. Such an implementation presents a key difference between
AWA-PAM and other AO-assisted 3D optical microscopy. For
example, the spinal cord enclosed within the white dashed box in
Fig. 3(c) could hardly be seen without AWA correction [Fig. 3(e)]
but becomes visible with the assistance of AWA correction. The
depth information for the 2D MAP image acquired with AWA
correction [Fig. 3(c)] was extracted based on the time-of-flight
information of the ultrasonic wave and provided in Fig. 3(f ).
Research Article
Vol. 11, No. 2 / February 2024 /
Optica
217
532
nm pulsed
laser
Lens
Lens
Pinhole
Half-w
ave
plate
Spatial
light
modulator
Lens
Mirror
Beam
splitter
Ultrasonic
transducer
Tissue
Motorized
stage
1
2
345
678
9
10
100
105
110
115
120
125
130
Enhancem
ent
ratio
(%
)
Zernik
e order
s
(c)
(b)
5 μm
(a)
Phot
odiode
Polarizin
g
beam
splitter
Fig.2.
Compensating for inherent system aberration. (a) Schematic illustration of the AWA-PAM system. The light being focused into the tissue is mod-
ulated by a spatial light modulator, which dynamically compensates for spatially inhomogeneous tissue-induced aberration. (b) Signal enhancement con-
tributed by each order of Zernike modes. (c) An image of the focus after compensation for system aberration, captured through an optical microscope.
This image shows that different spinal cords are indeed located
at different depths. Detailed information on the optimization
performance and phase maps at different locations can be found in
Supplement 1, Note 2.
D. Imaging Microvascular Structures of Mouse Ear
In
Vivo
The effectiveness of AWA-PAM was also demonstrated by imaging
microvascular structures of mouse ears
in vivo
. The mouse was
anesthetized by isoflurane during the entire experiment, and the
ear to be imaged was in a natural state, i.e., without being pressed or
flattened, which can otherwise restrain blood flow. The same set-
ting used for imaging zebrafish embryos was used here. Figures 4(a)
and 4(b) show the 3D microvascular images of the mouse ear from
two different views obtained with AWA correction, and its corre-
sponding 2D MAP image is also shown in Fig. 4(c). In comparison,
3D images and corresponding 2D MAP image obtained with-
out AWA correction are illustrated in Figs. 4(d)–4(f ). Although
big vessels can be identified in both images, the AWA correction
effectively helps to identify microvascular structures at depths.
For example, by scrutinizing the regions highlighted by the white
dashed boxes in Fig. 4(c), one could hardly see similar features in
the same regions in the image obtained without AWA correction
[Fig. 4(f )]. Moreover, by comparing the MAP images belonging
to different categories, we note that signal strengths obtained with
AWA correction are generally stronger than those obtained without
AWA correction. Furthermore, Figs. 4(g) and 4(h) plot the profiles
along the two white dashed lines labeled in Figs. 4(c) and 4(f ). A
direct comparison shows that richer structural information can be
observed with AWA correction (the red solid curve). In contrast,
the 1D profiles obtained without AWA correction, denoted by
the dashed blue curves, are almost informationless. As before,
for the 2D MAP image with AWA correction activated, the depth
information based on the time-of-flight information of the ultra-
sonic wave is provided in Fig. 4(i). As we can see from the figure,
microvascular structures that are hardly seen in Fig. 4(f ) are at a
depth that is distinctively different from the focal plane (13.1 mm).
These observations demonstrate the superior performance of
AWA-PAM over conventional PAM in revealing microstructures
that are wiped out by optical aberrations. Detailed information
on the optimization performance and phase maps at different
locations can be found in Supplement 1, Note 3.
3. DISCUSSION
In principle, AWA-PAM relies on the assumption that the tightest
focus produces the strongest PA amplitude. Unlike two-photon
fluorescent microscopy, which employs nonlinearity, PA is gen-
erally considered a linear process. Nonetheless, when imaging
capillaries in practice, Supplement 1, Note 4 suggests that the
lateral resolution degrades to tens of microns at the depths of
a few hundred microns. Since the lateral resolution without
AWA correction is larger than the diameters of many capillaries,
AWA-PAM can increase the PA amplitude and achieve a tighter
focus, thus revealing more microscopic features. Moreover, even
for very big vessels shown in Fig. 4, we found a roughly 20%–
30% signal enhancement by using AWA correction (detailed in