of 8
Supplementary Material
Harmonic imaging for nonlinear detection of acoustic biomolecules
Rohit Nayak
1
, Mengtong Duan
2
, Bill Ling
1
, Zhiyang Jin
3
, Dina Malounda
1
, and Mikhail G. Shapiro*
1,3,4
1
Division of Chemistry and Chemical Engineering,
California Institute of Technology, Pasadena, CA, USA
2
Division of Biology and Biological Engineering, California Institute of Technology, Pasadena, CA, USA
3
Division of Engineering and Applied Science, California Institute of Technology, Pasadena, CA, USA
4
Howard Hughes Medical Institute, California Institute of Technology, Pasadena, CA, USA
*Corresponding Author
Impact of averaging repeats on xAM and HxAM imaging
To assess the effect of averaging these repetitions on the image quality of xAM and
HxAM
imaging, we performed a focused experiment. We used a tissue
-
mimicking phantom with GV
inclusion at a concentration of OD 3. We varied the number of averaging repeats in a range from
1 to 50, incrementing in steps of 5 for each test run. This testing
was carried out using a set of
five phantoms to ensure a robust sample size. The results presented in Supplementary Figure 1
reveal a correlation between the number of averaging repeats and the enhanced performance in
both xAM and HxAM imaging modalities.
However, the quality of the HxAM images consistently
exceeded that of the xAM images, regardless of the number of averaging repeats applied. The
quantitative plots show that the GV signal [SFig. 1 (a)] with HxAM imaging is at least 5 dB higher
than xAM ima
ging, while no noticeable difference in the background signal [SFig. 1 (b)] was
observed, which resulted in higher SBR [SFig. 1 (c)] and CBR [SFig. 1 (d)]. Further, the image
quality metrics show that HxAM imaging with 15 averaging repeats achieved superio
r results
compared to xAM imaging with as many as 50 repeats. Drawing from these findings, we selected
15 averaging repeats as the standard for both HxAM and xAM imaging techniques in the
subsequent experiments presented in this paper.
Harmonics imaging using wild type GVs.
To validate the critical role of nonlinear behavior in the effectiveness of HxAM imaging, our study
included experiments with wild
-
type gas vesicles (wtGVs
) embedded in tissue mimicking
phantoms. Specifically, wtGVs naturally lack nonlinear properties due to the presence of the
structurally stiffening protein gvpC. As expected, wtGVs did not demonstrate a perceptible GV
signal in AM imaging for both xAM and
HxAM imaging [SFig. 2 (a)]. Further, HxAM imaging
demonstrated no quantifiable advantage over xAM, contrary to what was observed with stripped
GVs, when compared at the same OD [SFig. 2 (b)]. This outcome highlights the importance of
nonlinear buckling in the enhanced performance of HxAM imaging. Moreover, the magnitude
Fourier spectrum analysis of wtGVs revealed no significant amplification at harmonic frequencies
[SFig. 2 (c), (d)], further substantiating our findings.
Harmonics imaging using
parabolic AM pulse sequences.
To evaluate the feasibility of using harmonic imaging to enhance the performance of parabolic AM
(pAM) pulse sequence, we conducted
in vitro
experiments with GVs embedded in a tissue
mimicking phantom at OD 3.5. The GVs inclusions were imaged using both pAM and xAM
imaging, along with its harmonic counterparts, at the same phantom cross
-
section [SFig. 3 (a)].
The acquired channel data correspo
nding to the three transmit AM pulses were stored individually
for comparative analysis. Further, for simplified ana
lysis, the quantified GV, BG signals were
normalized by respective pAM and xAM signals. Similarly, the SBR and CBR values were
normalized by pAM and xAM counterparts, respectively. The quantitative results [SFig. 3 (b)]
reveal a notable difference in the h
armonic
-
to
-
fundamental frequency ratios of the GV signal
between pAM and xAM imaging methods. Specifically, the harmonic
-
to
-
fundamental frequency
ratio for pAM stands at a lower value of approximately 0.69, while xAM registers a higher value
of approximate
ly 0.98. This leads to differing SBR, with pAM at 1.03 and xAM at 1.47. Such
differences are critical in the context of limited bandwidth of the transducer and the scanner,
suggesting that the replacement of fundamental frequencies with harmonics in pAM do
es not
significantly enhance imaging performance. In contrast, in xAM imaging, the harmonic
frequencies are on par with the fundamental frequencies in amplitude, which contributes to a
significant improvement in image quality when harmonics are incorporate
d.
The constraint on the harmonic to fundamental frequency ratio observed in pAM is predominantly
due to an artificially increased fundamental frequency component due to propagation artifacts
commonly observed with approach. The presence of these artifacts ca
n also be observed in SFig.
3 (a). However, xAM imaging is engineered to mitigate this artifact, making harmonic contributions
a substantial contribution. Further examination of the absolute value GV signal in pAM and xAM
imaging [SFig. 3 (b)], where assum
ing a similar harmonic response for both imaging techniques,
pAM imaging at 15.625 MHz experiences a reduction of the full amplitude receive signal by 60%
after AM cancellation, whereas xAM imaging produces a reduction of up to 80%, thereby
improving the h
armonic
-
to
-
fundamental frequency ratio. This substantial difference underscores
the distinctive imaging capabilities and benefits that xAM holds over pAM in harmonic imaging
applications.
To assess the potential of harmonic imaging in augmenting the efficacy of the parabolic AM (pAM)
pulse sequence, we undertook
in vitro
studies with GVs embedded within a tissue
-
mimicking
phantom at an OD
500
of 3.5. Imaging of the GV inclusions was performed using both pAM and
xAM techniques at 15.625 MHz, as well as their harmonic counterparts (HpAM and HxAM) at 12.5
MHz, across the same section of the phantom, as shown in [SFig. 3 (a)]. The harmonic filtere
d
images (HpAM
-
f and HxAM
-
f) were estimated by app
lying a high pass filter with a cut off frequency
of 17.5 MHz. The channel data from the three AM transmit pulses were recorded separately for a
detailed comparative evaluation. For ease of analysis, the measured GV and background signals
were normalized a
gainst their corresponding pAM and xAM signals. Similarly, the SBR values
were normalized to their respective pAM and xAM benchmarks. The findings, depicted in [SFig.
3 (b)], highlight a significant disparity in the harmonic
-
to
-
fundamental frequency ratios
of the GV
signal between the pAM and xAM methods. Specifically, the ratio for pAM is notably lower, around
0.69, compared to xAM's higher ratio of approximately 0.98, leading to a variation in SBR values
1.03 for pAM and 1.47 for xAM. This discrepancy underscores the limited utility of substituting
fundamental frequencies with harmonics in pAM, given the constr
ained bandwidth of the
transducer and scanner. In contrast, xAM imaging, where harmonic and fundamental frequencies
are comparable in amplitude, markedly improves image quality through the inclusion of harmonic
frequencies.
The observed limitation in the harmonic to fundamental frequency ratio in pAM primarily results
from an artificially elevated fundamental frequency component, a consequence of propagation
artifacts typical of this method, also observable in [SFig. 3 (a)].
However, xAM imaging is designed
to overcome such artifacts, rendering harmonic enhancements significantly more effective. A
further analysis of the absolute GV signal values in pAM and xAM [SFig. 3 (c)] reveals that,
assuming a comparable harmonic respons
e in both techniques, pAM imaging at 15.625 MHz sees
a 60% reduction in the full amplitude receive signal following AM cancellation. In comparison,
xAM imaging achieves a reduction of up to 80%, thus favoring an improved harmonic
-
to
-
fundamental frequency r
atio. This pronounced difference highlights xAM's superior imaging
capabilities and advantages over pAM in the realm of harmonic imaging.
Elevational and lateral beam profile measurement
We measured the beam profiles for xAM and HxAM imaging using an HNR
-
0500 needle
hydrophone from Onda Corporation. These measurements were performed using an automated
two
-
axis motorized stage for precise control. The hydrophone was securely mounted at the
bottom of a water
-
filled tank, while the L22
-
14vX ultrasound probe was affixed to the motorized
stage. For the imaging sequences, the xAM and HxAM scripts were run in transmit mode.
Specifically, we activated the central cluster of 64 elements for executin
g the full amplitude
transmission of the three
-
pulse xAM sequence. The motorized stage was pre
-
programmed to
conduct a raster scan with increments of 50
μ
m in the elevational direction and 100
μ
m in the
lateral direction. Data from the hydrophone was captu
red using a digital oscilloscope operating at
a sampling frequency of 9 GHz. After obtaining this data, we identified the planes corresponding
to the peak amplitude levels for both the elevational and lateral beam profiles. These selected
profiles are illu
strated in SFig. 4. The data reveal that the elevational profiles of xAM and HxAM
imaging are closely matched [SFig. 4 (a)], with neither method showing a distinct advantage over
the other. The full width half max (FWHM) for the elevational beam profile wa
s 0.866 mm and
0.855 mm for xAM and HxAM, respectively. Lateral profile analysis further indicates that the main
lobes are consistent between xAM and HxAM [SFig. 4 (b)]. Similarly, the FWHM for the lateral
beam profile was 0.806 mm and 0.785 mm for xAM and
HxAM, respectively. While HxAM exhibits
more pronounced side lobes, their amplitude does not surpass the buckling threshold of the GVs,
even at the full transmission amplitude for the pressures employed in this study. Therefore, it is
unlikely that these
side lobes would result in artifacts within the amplitude
-
modulated images.
In vitro
imaging of engineered MDA cells expressing acoustic reporter genes
Supplementary Figure 5 displays
in vitro
ultrasound images of engineered MDA cells, post
-
collapse of the GVs for xAM and HxAM imaging, at different cell concentrations. These post
-
collapse images are the background counterparts to those depicted in Figure 3, illustrating the
effect of GV collaps
e on the imaging results.
In vivo
imaging of mice liver
Supplementary Figure 6 depicts the axial line plots corresponding to GV and background signal
at varying depths of xAM and HxAM images [Fig. 4 (c)], corresponding to IV injection of GVs at
the lower concentration. Analysis of the magnitude Fourier spectra
associated with GV and
background signals in the xAM and HxAM images [Fig. 4 (c)] show a distinct presence of harmonic
contribution for HxAM in the presence of GV, which disappears with the collapse of the GVs, as
identified as background in [SFig. 6 (b)].
Supplementary Figure 1: Ultrasound imaging of GV at varying acquisition averaging repeats
. The
quantitative plots show
(a)
GV signal at OD
500
3.5 and the corresponding
(b)
BG signal in the tissue
mimicking phantom material as a function of averaging repeats for both HxAM and xAM imaging. The
corresponding
(c)
SBR and
(d)
CBR metric. N=5 phantom samples.
Signal amplitude presented in
decibels (dB rel. raw) are calculated based on the absolute signal recorded directly from the Verasonics
scanner without further normalization.
Supplementary Figure 2: In vitro ultrasound imaging of wtGVs in tissue
-
mimicking phantoms.
(a)
xAM and
HxAM images of wtGVs at a relatively high concentration of OD3.
(b)
Quantitative analysis of xAM and HxAM
imaging with respect to wtGVs, corresponding background and comparison with GVs at matched concentration.
The bar plots report the magnitude of GV and background signal, and the corresponding performance metrics of
ST
R and CTR.
(c)
Magnitude Fourier spectrum associated with xAM and HxAM images of wtGVs
relative to the
background, and
(d)
the corresponding spectral power. N=5. Errorbars: standard error of the mean (SEM).
Signal
amplitude presented in decibels (dB rel. raw) are calculated based on the absolute signal recorded directly from
the Verasonics scanner without further normalization.
Supplementary Figure 3: Comparative imaging and signal analysis using a tissue mimicking
phantom
embedded with a GV inclusion.
(a)
The top row displays images from pAM, HpAM, and harmonic filtered HpAM
(HpAM
-
f) imaging of a GV inclusion at OD 3.5. The bottom row presents the corresponding images for xAM, HxAM,
and harmonic filtered HxAM (HxAM
-
f), captured at the same imaging cross
-
se
ction.
(b)
Barplots display a
quantitative assessment of the GV and background signals within the ROIs identified in (a), along with estimates of
the SBR. The top and bottom rows detail the GV and background signals
normalized to the GV signal obtained from
pAM and xAM imaging at 15.625 MHz, respectively. The SBR values are accordingly normalized against the pAM
and xAM counterparts at 15.625 MHz.
(c)
displays the absolute values of the GV signal from (b), juxtaposed with the
GV signal captured at full amplitude transmit before AM cancellation. N = 5 samples; error bars represent standard
error of the mean.
Supplementary Figure 4.
Hydrophone measured beam profile of xAM and HxAM imaging.
The top
and bottom rows correspond to elevational and lateral beam profiles of xAM (blue) and HxAM (orange)
imaging.
Supplementary Figure 5: In vitro ultrasound imaging of engineered mammalian cells with collapsed
gas vesicles.
This figure displays ultrasound images of engineered mammalian cells expressing acoustic
reporter genes, following the collapse of the GVs. The top row presents xAM images, and the bottom row
shows HxAM images, across various cell concentrations. These im
ages are the post
-
collapse counterparts
to those depicted in Figure 3, illustrating the effect of GV collapse on the imaging results. Data were co
llected
from N=5 sample sets
. Signal amplitude presented in decibels (dB rel. raw) are calculated based on the
absolute signal recorded directly from the Verasonics scanner without further normalization.
Supplementary Figure 6: In vivo ultrasound imaging of mice liver after intravenous injection of purified
GVs at OD
500
10.
(a)
Axial line plots corresponding to GV and background signal at varying depths of images
corresponding to IV injection of GVs. The solid line plots and the corresponding dotted double
-
sided bands
represent mean and the standard error, respectively, estimate
d across all columns of the ultrasound
images in Figure 4 (c).
(b)
Magnitude Fourier spectra associated with GV and background signals in xAM and
HxAM images reported in Figure 4 (c). N=5 mice and the error bars represent standard error of the mean.
Signal amplitude presented in decibels (dB rel. raw) are calculated based on the absolute signal recorded
directly from the Verasonics scanner without further normalization.