A composite set of ovarian tissue features extracted from photoacoustic spectral data, beam envelope, and co-registered ultrasound and photoacoustic images are used to characterize malignant and normal ovaries using logistic and support vector machine (SVM) classifiers. Normalized power spectra were calculated from the Fourier transform of the photoacoustic beamformed data, from which the spectral slopes and 0-MHz intercepts were extracted. Five features were extracted from the beam envelope and another 10 features were extracted from the photoacoustic images. These 17 features were ranked by their p-values from t-tests on which a filter type of feature selection method was used to determine the optimal feature number for final classification. A total of 169 samples from 19 ex vivo ovaries were randomly distributed into training and testing groups. Both classifiers achieved a minimum value of the mean misclassification error when the seven features with lowest p-values were selected. Using these seven features, the logistic and SVM classifiers obtained sensitivities of 96.39±3.35% and 97.82±2.26%, and specificities of 98.92±1.39% and 100%, respectively, for the training group. For the testing group, logistic and SVM classifiers achieved sensitivities of 92.71±3.55% and 92.64±3.27%, and specificities of 87.52±8.78% and 98.49±2.05%, respectively.
Laser diodes are widely used in diffuse optical tomography (DOT) systems but are typically expensive and fragile, while light-emitting diodes (LEDs) are cheaper and are also available in the near-infrared (NIR) range with adequate output power for imaging deeply seated targets. In this study, we introduce a new low-cost DOT system using LEDs of four wavelengths in the NIR spectrum as light sources. The LEDs were modulated at 20 kHz to avoid ambient light. The LEDs were distributed on a hand-held probe and a printed circuit board was mounted at the back of the probe to separately provide switching and driving current to each LED. Ten optical fibers were used to couple the reflected light to 10 parallel photomultiplier tube detectors. A commercial ultrasound system provided simultaneous images of target location and size to guide the image reconstruction. A frequency-domain (FD) laser-diode-based system with ultrasound guidance was also used to compare the results obtained from those of the LED-based system. Results of absorbers embedded in intralipid and inhomogeneous tissue phantoms have demonstrated that the LED-based system provides a comparable quantification accuracy of targets to the FD system and has the potential to image deep targets such as breast lesions.
KEYWORDS: Acquisition tracking and pointing, Imaging systems, Field programmable gate arrays, Ovary, Tumors, Ultrasonography, Tissue optics, Data acquisition, Photoacoustic spectroscopy, Digital signal processing
Coregistered ultrasound (US) and photoacoustic imaging are emerging techniques for mapping the echogenic anatomical structure of tissue and its corresponding optical absorption. We report a 128-channel imaging system with real-time coregistration of the two modalities, which provides up to 15 coregistered frames per second limited by the laser pulse repetition rate. In addition, the system integrates a compact transvaginal imaging probe with a custom-designed fiber optic assembly for in vivo detection and characterization of human ovarian tissue. We present the coregistered US and photoacoustic imaging system structure, the optimal design of the PC interfacing software, and the reconfigurable field programmable gate array operation and optimization. Phantom experiments of system lateral resolution and axial sensitivity evaluation, examples of the real-time scanning of a tumor-bearing mouse, and ex vivo human ovaries studies are demonstrated.
In this paper, human ovarian tissues with malignant and benign features were imaged ex vivo by using an opticalresolution photoacoustic microscopy (OR-PAM) system. Several features were quantitatively extracted from PAM images to describe photoacoustic signal distributions and fluctuations. 106 PAM images from 18 human ovaries were classified by applying those extracted features to a logistic prediction model. 57 images from 9 ovaries were used as a training set to train the logistic model, and 49 images from another 9 ovaries were used to test our prediction model. We assumed that if one image from one malignant ovary was classified as malignant, it is sufficient to classify this ovary as malignant. For the training set, we achieved 100% sensitivity and 83.3% specificity; for testing set, we achieved 100% sensitivity and 66.7% specificity. These preliminary results demonstrate that PAM could be extremely valuable in assisting and guiding surgeons for in vivo evaluation of ovarian tissue.
A high-throughput ultrasound/photoacoustic probe for delivering high contrast and signal-to-noise ratio images was designed, constructed, and tested. The probe consists of a transvaginal ultrasound array integrated with four 1mm-core optical fibers and a sheath. The sheath encases transducer and is lined with highly reflecting aluminum for high intensity light output and uniformity while at the same time remaining below the maximum permissible exposure (MPE) recommended by the American National Standards Institute (ANSI). The probe design was optimized by simulating the light fluence distribution in Zemax. The performance of the probe was evaluated by experimental measurements of the fluence and real-time imaging of polyethylene-tubing filled with blood. These results suggest that our probe has great potential for in vivo imaging and characterization of ovarian cancer.
Human ovarian tissue features extracted from photoacoustic spectra data, beam envelopes and co-registered ultrasound and photoacoustic images are used to characterize cancerous vs. normal processes using a support vector machine (SVM) classifier. The centers of suspicious tumor areas are estimated from the Gaussian fitting of the mean Radon transforms of the photoacoustic image along 0 and 90 degrees. Normalized power spectra are calculated using the Fourier transform of the photoacoustic beamformed data across these suspicious areas, where the spectral slope and 0-MHz intercepts are extracted. Image statistics, envelope histogram fitting and maximum output of 6 composite filters of cancerous or normal patterns along with other previously used features are calculated to compose a total of 17 features. These features are extracted from 169 datasets of 19 ex vivo ovaries. Half of the cancerous and normal datasets are randomly chosen to train a SVM classifier with polynomial kernel and the remainder is used for testing. With 50 times data resampling, the SVM classifier, for the training group, gives 100% sensitivity and 100% specificity. For the testing group, it gives 89.68± 6.37% sensitivity and 93.16± 3.70% specificity. These results are superior to those obtained earlier by our group using features extracted from photoacoustic raw data or image statistics only.
To overcome the intensive light scattering in biological tissue, diffuse optical tomography (DOT) in the near-infrared range for breast lesion detection is usually combined with other imaging modalities, such as ultrasound, x-ray, and magnetic resonance imaging, to provide guidance. However, these guiding imaging modalities may depend on different contrast mechanisms compared to the optical contrast in the DOT. As a result, they cannot provide reliable guidance for DOT because some lesions may not be detectable by a nonoptical modality but may have a high optical contrast. An imaging modality that relies on optical contrast to provide guidance is desirable for DOT. We present a system that combines a frequency-domain DOT and real-time photoacoustic tomography (PAT) systems to detect and characterize deeply seated targets embedded in a turbid medium. To further improve the contrast, the exogenous contrast agent, indocyanine green (ICG), is used. Our experimental results show that the combined system can detect a tumor-mimicking phantom, which is immersed in intralipid solution with the concentrations ranging from 100 to 10 μM and with the dimensions of 0.8 cm×0.8 cm×0.6 cm , up to 2.5 cm in depth. Mice experiments also confirmed that the combined system can detect tumors and monitor the ICG uptake and washout in the tumor region. This method can potentially improve the accuracy to detect small breast lesions as well as lesions that are sensitive to background tissue changes, such as the lesions located just above the chest wall.
A photoacoustic contrast agent that is based on bis-carboxylic acid derivative of indocyanine green (ICG) covalently conjugated to single-wall carbon nanotubes (ICG/SWCNT) is presented. Covalently attaching ICG to the functionalized SWCNT provides a more robust system that delivers much more ICG to the tumor site. The detection sensitivity of the new contrast agent in a mouse tumor model is demonstrated in vivo by our custom-built photoacoustic imaging system. The summation of the photoacoustic tomography (PAT) beam envelope, referred to as the “PAT summation,” is used to demonstrate the postinjection light absorption of tumor areas in ICG- and ICG/SWCNT-injected mice. It is shown that ICG is able to provide 33% enhancement at approximately 20 min peak response time with reference to the preinjection PAT level, while ICG/SWCNT provides 128% enhancement at 80 min and even higher enhancement of 196% at the end point of experiments (120 min on average). Additionally, the ICG/SWCNT enhancement was mainly observed at the tumor periphery, which was confirmed by fluorescence images of the tumor samples. This feature is highly valuable in guiding surgeons to assess tumor boundaries and dimensions in vivo and to achieve clean tumor margins to improve surgical resection of tumors.
In this study, we present a novel photoacoustic contrast agent which is based on bis-carboxylic acid
derivative of Indocyanine green (ICG) covalently conjugated to single-wall carbon nanotubes
(ICG/SWCNT). Covalently attaching ICG to the functionalized SWCNT provides a more robust system
that delivers much more ICG to the tumor site. The detection sensitivity of the new contrast agent in
mouse tumor model is demonstrated in vivo by our custom built photoacoustic imaging system. PAT
summation signal is defined to show the long-term light absorption of tumor areas in ICG injected mice and
ICG/SWCNT injected mice. It is shown that ICG is able to provide 33% enhancement at approximately 20
minutes peak response time referred to pre-injection PAT summation level, while ICG/SWCNT provides
128% enhancement at 80 minutes and even higher enhancement of 196% at the end point of experiments
(120 minutes on average). Additionally, the ICG/SWCNT enhancement was mainly observed at the tumor
periphery as confirmed by fluorescence images of the tumor samples. This feature is highly valuable in
guiding surgeons to assess tumor boundaries and dimensions in vivo and improve surgical resection of
tumors for achieving clean tumor margins.
Unique features in co-registered ultrasound and photoacoustic images of ex vivo ovarian tissue are introduced, along with
the hypotheses of how these features may relate to the physiology of tumors. The images are compressed with wavelet
transform, after which the mean Radon transform of the photoacoustic image is computed and fitted with a Gaussian
function to find the centroid of the suspicious area for shift-invariant recognition process. In the next step, 24 features
are extracted from a training set of images by several methods; including features from the Fourier domain, image
statistics, and the outputs of different composite filters constructed from the joint frequency response of different
cancerous images. The features were chosen from more than 400 training images obtained from 33 ex vivo ovaries of 24
patients, and used to train a support vector machine (SVM) structure. The SVM classifier was able to exclusively
separate the cancerous from the non-cancerous cases with 100% sensitivity and specificity. At the end, the classifier was
used to test 95 new images, obtained from 37 ovaries of 20 additional patients. The SVM classifier achieved 76.92%
sensitivity and 95.12% specificity. Furthermore, if we assume that recognizing one image as a cancerous case is
sufficient to consider the ovary as malignant, then the SVM classifier achieves 100% sensitivity and 87.88% specificity.
This paper presents a real-time transvaginal photoacoustic imaging probe for imaging human ovaries in vivo. The probe
consists of a high-throughput (up to 80%) fiber-optic 1 x 19 beamsplitters, a commercial array ultrasound transducer,
and a fiber protective sheath. The beamsplitter has a 940-micron core diameter input fiber and 240-micron core diameter
output fibers numbering 36. The 36 small-core output fibers surround the ultrasound transducer and delivers light to the
tissue during imaging. A protective sheath, modeled in the form of the transducer using a 3-D printer, encloses the
transducer with array of fibers. A real-time image acquisition system collects and processes the photoacoustic RF signals
from the transducer, and displays the images formed on a monitor in real time. Additionally, the system is capable of
coregistered pulse-echo ultrasound imaging. In this way, we obtain both morphological and functional information from
the ovarian tissue. Photoacousitc images of malignant human ovaries taken ex vivo with the probe revealed blood
vascular and networks that was distinguishable from normal ovaries, making the probe potential useful for characterizing
ovarian tissue.
In this paper, we present the construction of an optical-resolution photoacoustic microscopy (OR-PAM) system and
studies done on the characterization of human ovarian tissue with malignant and benign features ex vivo. PAM images of
the ovaries showed more detailed blood vessel distributions with much higher resolution compared with conventional
photoacoustic images obtained with array transducers. In all, 29 PAM images (20 from normal ovaries and 9 from
malignant ovaries) were studied. Eight different features were extracted quantitatively from the PAM images, and a
generalized linear model (GLM) was used to classify the ovaries as normal or malignant. By using the GLM, a
specificity of 100% and a sensitivity of 100% were obtained for the training set. These preliminary results demonstrate
the feasibility of our PAM system in mapping microvasculature networks, as well as characterizing the ovarian tissue,
and could be extremely valuable in assisting surgeons for in vivo evaluation of ovarian tissue.
KEYWORDS: Imaging systems, Acquisition tracking and pointing, Ultrasonography, Field programmable gate arrays, Ovary, Tumors, Tissue optics, In vivo imaging, Pulsed laser operation, Ovarian cancer
In this paper, we report an ultrafast co-registered ultrasound and photoacoustic imaging system based on FPGA parallel
processing. The system features 128-channel parallel acquisition and digitization, along with FPGA-based reconfigurable
processing for real-time co-registered imaging of up to 15 frames per second that is only limited by the laser pulse
repetition frequency of 15 Hz. We demonstrated the imaging capability of the system by live imaging of a mouse tumor
model in vivo, and imaging of human ovaries ex vivo. A compact transvaginal probe that includes the PAT illumination
using a fiber-optic assembly was used for this purpose. The system has the potential ability to assist a clinician to
perform transvaginal ultrasound scanning and to localize the ovarian mass, while simultaneously mapping the light
absorption of the ultrasound detected mass to reveal its vasculature using the co-registered PAT.
Unique features and the underlining hypotheses of how these features may relate to the tumor physiology in coregistered ultrasound and photoacoustic images of ex vivo ovarian tissue are introduced. The images were first compressed with wavelet transform. The mean Radon transform of photoacoustic images was then computed and fitted with a Gaussian function to find the centroid of a suspicious area for shift-invariant recognition process. Twenty-four features were extracted from a training set by several methods, including Fourier transform, image statistics, and different composite filters. The features were chosen from more than 400 training images obtained from 33 ex vivo ovaries of 24 patients, and used to train three classifiers, including generalized linear model, neural network, and support vector machine (SVM). The SVM achieved the best training performance and was able to exclusively separate cancerous from non-cancerous cases with 100% sensitivity and specificity. At the end, the classifiers were used to test 95 new images obtained from 37 ovaries of 20 additional patients. The SVM classifier achieved 76.92% sensitivity and 95.12% specificity. Furthermore, if we assume that recognizing one image as a cancer is sufficient to consider an ovary as malignant, the SVM classifier achieves 100% sensitivity and 87.88% specificity.
KEYWORDS: Field programmable gate arrays, Ultrasonography, Data acquisition, Imaging systems, Photoacoustic spectroscopy, Digital signal processing, Image processing, Signal processing, Pulsed laser operation, Ultrafast imaging
Co-registered Ultrasound and Photoacoustic images provide complimentary structure and functional information for
cancer diagnosis and assessment of therapy response. In SPIE Photonics West 2011, we reported a system that acquires
from 64 channels and displays up to 1 frame per second (fps) ultrasound pulse-echo images, 5 fps photoacoustic images,
and 0.5 fps co-registered images. In this year, we report an upgraded system which acquires from 128 channels and
displays up to 15 fps co-registered ultrasound and photoacoustic images limited by our laser pulse repetition rate. The
system architecture is novel and it provides real-time
co-registration of images, the ability of acquiring the channel RF
data for both modalities, and the flexibility of adjusting every parameter involved in the imaging process for both
modalities.
The digital signal processor board is upgraded to an FPGA-based PCIe board that collects the data from the
acquisition modules and transfers them to the PC memory at 2.5GT/s rate through an x8 DDR PCIe bus running at
100MHz clock frequency. The modules FPGA code is also upgraded to form a beam line in 90 microseconds and to
communicate through ultrafast differential tracks with the PCIe board. Furthermore, the printed circuit board (PCB)
design of the system was adjusted to provide a maximum of 80dB signal-to-noise ratio at 60dB gain, which is
comparable to some commercial ultrasound machines.
The real-time system allows capturing co-registered US/PAT images free of motion artifacts and also provides
ultrafast dynamic information when a contrast agent is used. The system is built for clinical use to assist the diagnosis of
ovarian cancer. However, the hardware is still under testing and evaluation stage, experimental and clinical results will
be reported later.
We present a photoacoustic tomography-guided diffuse optical tomography approach using a hand-held probe for detection and characterization of deeply-seated targets embedded in a turbid medium. Diffuse optical tomography guided by coregistered ultrasound, MRI, and x ray has demonstrated a great clinical potential to overcome lesion location uncertainty and to improve light quantification accuracy. However, due to the different contrast mechanisms, some lesions may not be detectable by a nonoptical modality but yet have high optical contrast. Photoacoustic tomography utilizes a short-pulsed laser beam to diffusively penetrate into tissue. Upon absorption of the light by the target, photoacoustic waves are generated and used to reconstruct, at ultrasound resolution, the optical absorption distribution that reveals optical contrast. However, the robustness of optical property quantification of targets by photoacoustic tomography is complicated because of the wide range of ultrasound transducer sensitivity, the orientation and shape of the targets relative to the ultrasound array, and the uniformity of the laser beam. We show in this paper that the relative optical absorption map provided by photoacoustic tomography can potentially guide the diffuse optical tomography to accurately reconstruct target absorption maps.
A handheld photoacoustic tomography-guided diffuse optical tomography system for imaging deeply-seated targets in
scattering media is presented. This hybrid imager consists of a probe with an ultrasound transducer in the center and
straddled by two optical fibers for taking photoacoustic images. The diffuse optical tomography component comprises of
9 light-source fibers for delivering light to the imaged tissue, and 14 detector fibers for collecting the light. Single- and
two-phantom targets of high and low optical contrasts were immersed in a scattering intralipid solution to depths of up to
3cm and imaged. The reconstructed absorption coefficients of the targets with guidance from photoacoustic tomography
were compared to those obtained with a-priori depth-only information, and no a-priori information. The reconstructed
absorption maps yielded as much as 2.6-fold improvement in the quantification accuracy compared to the cases with no
guidance from photoacoustic tomography.
We present a multi-wavelength DC system using Light Emitting Diode (LED) sources of four
wavelengths in the near infrared range. These LEDs are commercially available, much cheaper
than laser diodes, and have adequate power to probe deeply seated lesions. In our system, 8
groups of LEDs of four wavelengths were deployed on a hand-held probe and 10 PMT detectors
were fiber coupled to the probe. A co-registered ultrasound (US) array located in the middle of
the probe provided lesion location and morphology, which were used for assisting near infrared
imaging reconstruction. Experiments evaluated the performance of the LED based DC system
using phantom targets.
KEYWORDS: Tumors, Carbon nanotubes, Photoacoustic imaging, Photoacoustic spectroscopy, Absorption, Hypoxia, Near infrared, Single walled carbon nanotubes, In vivo imaging, Digital signal processing
Development of new and efficient contrast agents is of fundamental importance to improve detection sensitivity of
smaller lesions. Within the family of nanomaterials, carbon nanotubes (CNT) not only have emerged as a new
alternative and efficient transporter and translocater of therapeutic molecules but also as a photoacoustic molecular
imaging agent owing to its strong optical absorption in the near-infrared region. Drugs, Antibodies and nucleic acids
could functionalize the CNT and prepare an appropriate system for delivering the cargos to cells and organs. In this
work, we present a novel photoacoustic contrast agent which is based on a unique hypoxic marker in the near infrared
region, 2-nitroimidazole -bis carboxylic acid derivative of Indocyanine Green conjugated to single walled carbon
nanotube (SWCNT-2nitroimidazole-ICG). The 2-nitroimidazole-ICG has an absorption peak at 755 nm and an extinction
coefficient of 20,5222 M-1cm-1. The conjugation of this marker with SWCNT shows more than 25 times enhancement of
optical absorption of carbon nanotubes in the near infrared region. This new conjugate has been optically evaluated
and shows promising results for high contrast photoacoustic imaging of deeply located tumors. The conjugate
specifically targets tumor hypoxia, an important indicator of tumor metabolism and tumor therapeutic response. The
detection sensitivity of the new contrast agent has been evaluated in-vitro cell lines and with in-vivo tumors in mice.
Co-registering ultrasound (US) and photoacoustic (PA) imaging is a logical extension to conventional ultrasound
because both modalities provide complementary information of tumor morphology, tumor vasculature and hypoxia for
cancer detection and characterization. In addition, both modalities are capable of providing real-time images for clinical
applications. In this paper, a Field Programmable Gate Array (FPGA) and Digital Signal Processor (DSP) module-based
real-time US/PA imaging system is presented. The system provides real-time US/PA data acquisition and image display
for up to 5 fps* using the currently implemented DSP board. It can be upgraded to 15 fps, which is the maximum pulse
repetition rate of the used laser, by implementing an advanced DSP module. Additionally, the photoacoustic RF data for
each frame is saved for further off-line processing.
The system frontend consists of eight 16-channel modules made of commercial and customized circuits. Each
16-channel module consists of two commercial 8-channel receiving circuitry boards and one FPGA board from Analog
Devices. Each receiving board contains an IC† that combines.
8-channel low-noise amplifiers, variable-gain amplifiers,
anti-aliasing filters, and ADC's‡ in a single chip with sampling frequency of 40MHz. The FPGA board captures the
LVDSξ Double Data Rate (DDR) digital output of the receiving board and performs data conditioning and subbeamforming.
A customized 16-channel transmission circuitry is connected to the two receiving boards for US pulseecho
(PE) mode data acquisition. A DSP module uses External Memory Interface (EMIF) to interface with the eight
16-channel modules through a customized adaptor board. The DSP transfers either sub-beamformed data (US pulse-echo
mode or PAI imaging mode) or raw data from FPGA boards to its DDR-2 memory through the EMIF link, then it
performs additional processing, after that, it transfer the data to the PC** for further image processing. The PC code
performs image processing including demodulation, beam envelope detection and scan conversion. Additionally, the PC
code pre-calculates the delay coefficients used for transmission focusing and receiving dynamic focusing for different
types of transducers to speed up the imaging process. To further speed up the imaging process, a multi-threads technique
is implemented in order to allow formation of previous image frame data and acquisition of the next one simultaneously.
The system is also capable of doing semi-real-time automated SO2 imaging at 10 seconds per frame by changing the
wavelength knob of the laser automatically using a stepper motor controlled by the system.
Initial in vivo experiments were performed on animal tumors to map out its vasculature and hypoxia level, which
were superimposed on co-registered US images. The real-time system allows capturing co-registered US/PA images free
of motion artifacts and also provides dynamitic information when contrast agents are used.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
INSTITUTIONAL Select your institution to access the SPIE Digital Library.
PERSONAL Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.