|Home | About | Journals | Submit | Contact Us | Français|
Photoacoustic tomography (PAT) is probably the fastest growing biomedical imaging technology owing to its capability of high-resolution sensing of rich optical contrast in vivo at depths beyond the optical transport mean free path (~1 mm in the skin). Existing high-resolution optical imaging technologies, such as confocal microscopy and two-photon microscopy, have fundamentally impacted biomedicine but cannot reach such depths. Taking advantage of low ultrasonic scattering, PAT indirectly improves tissue transparency by 100 to 1000 fold and consequently enables deeply penetrating functional and molecular imaging at high spatial resolution. Further, PAT holds the promise of in vivo imaging at multiple length scales ranging from subcellular organelles to organs with the same contrast origin, an important application in multiscale systems biology research.
Photoacoustic tomography (PAT),1–3 also referred to as optoacoustic tomography, is defined as cross-sectional or three-dimensional (3D) imaging of a material based on the photoacoustic effect. Planar imaging without depth resolution or one-dimensional (1D) depth-resolved imaging does not belong to tomography. In the photoacoustic phenomenon, light is absorbed by a material and converted to heat, and subsequently an acoustic wave is generated due to thermoelastic expansion. PAT—a hybrid of optical imaging and ultrasound imaging—combines rich optical contrast and high ultrasonic resolution in a single modality. PAT is capable of providing high-resolution structural, functional,4–7 and molecular8–13 imaging in vivo in optically scattering biological tissue at unprecedented depths. While structural imaging cannot differentiate live and dead tissues, functional imaging can do so by measuring physiological functions, such as the blood flow and oxygenation. In addition, molecular imaging can sense biomarkers to identify specific cancer cells or detect gene expression products to track gene activations.
The milestones of PAT and its precursor are briefly summarized here. One hundred years after Bell reported the photoacoustic effect in 1880,14 Bowen envisioned its potential for imaging with the excitation of ionizing radiation (high-energy electrons, x-ray photons, neutrons, and other charged particles) or non-ionizing radiation (radiowaves, microwaves, and ultrasonic waves).15 He illustrated 1D depth-resolved imaging using radiowave excitation but never mentioned the possibility of using optical excitation. Ten years later, laser-induced photoacoustic depth-resolved imaging was achieved initially with no or poor (e.g., ~6 mm) lateral resolution in the direction perpendicular to the acoustic axis.16–20 To acquire cross-sectional or 3D images that are both axially and laterally resolved, PAT based on an approximate reconstruction algorithm was developed;21 alternatively, PAT based on tightly focused ultrasonic detection was attained.22 In 2003, in vivo functional PAT was reported.4 The subsequent rapid growth of research on PAT is highlighted in this article.
Although many biomedical imaging modalities exist, their limitations call for new imaging technologies. Optical imaging is a strong contender primarily because of its rich contrast. From the safety perspective, light is nonionizing radiation with photon energy of the order of 2 electron volts, whereas x-ray has photon energy of thousands of electron volts and can consequently cause carcinogenesis. From the physics perspective, optical signals are related to the molecular constituents of tissue and hence can provide biochemical information. From the optics perspective, optical absorption is related to various intrinsic contrast origins, such as oxygenated and deoxygenated hemoglobin, melanin, lipids, and even water. As a result, optical imaging can provide functional imaging—mapping of physiological parameters. By using multiwavelength measurement, one can simultaneously quantify concentrations of multiple chromophores of different colors, such as oxygenated and deoxygenated hemoglobin molecules in red blood cells. Such quantification of hemoglobin can provide functional imaging of the concentration and the oxygen saturation of hemoglobin—both parameters are related to angiogenesis and hypermetabolism (hallmarks of cancer)—as well as brain activities or injuries.4,23,24 In addition, extrinsic optical absorption contrast agents can be used to provide molecular imaging of gene expressions or biomarkers.8–13
High-resolution optical microscopy faces two fundamental challenges: diffraction and diffusion. Diffraction limits the spatial resolution, whereas diffusion limits the penetration. The resolution of optical microscopy has been improved beyond the diffraction limit recently.25–27 This article focuses on breaking the diffusion limit.
Despite light diffusion due to tissue scattering, a quick experiment by transmitting flash light through the palm in the dark would prove the penetration capability of scattered light in biological tissue. Although the flash light is broadband, only red light transmits substantially. Unlike x-ray photons, optical photons do not cast a shadow of the bone. Such an observation is well-understood in biomedical optics. Because hemoglobin—the primary chromophore in biological tissue—absorbs light strongly in the green and blue spectral region, red light propagates further before being absorbed than shorter wavelength light. Further, strong light scattering by tissue components such as cell nuclei and mitochondria blurs any shadows of internal structures. In fact, a red photon may experience hundreds to thousands of scattering events before being absorbed. The reciprocal of the mean distance between equivalent isotropic scattering events—called transport mean free path—is on the order of 1 mm in the visible and near-infrared spectral range, around which photon propagation transitions from the ballistic regime into the diffusive regime.28 Because optical focusing is ineffective in the diffusive regime, the transport mean free path is referred to as the soft depth limit for high-resolution optical imaging. Beyond the soft depth limit, no pure optical imaging technologies have been able to achieve high spatial resolution. Although scattering per se does not eliminate photons, tortuous elongated photon paths boost the chance of tissue absorption. Even diffuse photons are too few to be practically useful beyond a depth of ~50 mm, which is referred to as the hard depth limit for optical imaging.
PAT overcomes the soft depth limit and achieves deep imaging at high resolution. Ultrasonic scattering coefficient in tissue is 2–3 orders of magnitude less than the optical counterpart. As a result, PAT can provide high spatial resolution by detecting ultrasonic waves induced by diffuse photons. In most cases, the center frequency and the bandwidth of the ultrasonic detection system predominantly determine the spatial resolution; the greater they are, the better the spatial resolution is, but the worse the ultrasonic penetration becomes. Such scalability empowers multiscale imaging. It is worth mentioning that lower-frequency electromagnetic waves including microwaves and radiowaves have been adopted for photoacoustic excitation to overcome the optical hard depth limit.29,30
Ultrasound imaging can provide high spatial resolution although ultrasound images appear grainy owing to the speckle artifacts. The relative spatial resolution—defined as the ratio of the penetration limit to the depth resolution (i.e., the pixel count in the depth direction without artificial interpolation)—is more than 100, which is considered high enough to provide good-quality images. From ultrasound imaging, PAT inherits the attribute of high spatial resolution but rejects the trait of speckle formation. The speckle-free nature of PAT results from strong correlations among the photoacoustic waves from the absorbers because all initial photoacoustic pressure rises are positive. Consequently, prominent boundaries always build up in photoacoustic images and suppress the interior fluctuations.31
Simply put, in PAT, light goes in and ultrasound comes out. For efficient ultrasound generation, a short-pulsed laser is the usual choice. The strength of the photoacoustic pressure depends on the optical energy deposition as well as the thermal and mechanical properties of the tissue. Because either unscattered or scattered photons can produce photoacoustic signals as long as photons are converted into heat, PAT can function in the optical diffusive regime. The image formation is essentially triangulation of the photoacoustic sources according to the time-of-flight signals recorded at multiple locations outside the tissue.2,28
PAT manifests in two major modes. In the mode of focused scanning PAT, such as dark-field confocal photoacoustic microscopy (PAM),32,33 a focused ultrasonic transducer is mechanically scanned.29 At each scanned position, the acoustic focusing yields lateral resolution while the acoustic time-of-flight provides axial resolution. In the mode of photoacoustic computed tomography, an array of unfocused ultrasonic transducers detects photoacoustic waves in parallel. Therefore, high frame rates—such as 50 Hz—can be attained.3,4,21,34–46 However, an inverse reconstruction algorithm is required to reconstruct a tomographic image. The early reconstruction was based on the assumption that each ultrasonic transducer receives signals simultaneously from photoacoustic sources on a plane instead of a spherical surface,21 which can be difficult to satisfy in practice. Later, various groups developed accurate reconstruction algorithms.3 More recently, an exact closed-form reconstruction formula was found for three detection geometries: planar, cylindrical, and spherical configurations.41 Interestingly, one may think of a focused ultrasonic transducer used in the first mode as an analog computer that performs approximately the inverse algorithm in the second mode. While hardware focusing does not present discretization artifacts, the focal position cannot be adjusted as readily as in photoacoustic computed tomography. As a result, the optimal lateral resolution is achievable only within the focal zone unless a virtual-detector-based synthetic aperture algorithm is used.
PAT is exquisitely sensitive to optical absorption. The photoacoustic excitation converts a small change in optical absorption coefficient to an equal fractional change in ultrasound signal. For the same reason, PAT is also sensitive to any change in the Grueneisen coefficient—the conversion efficiency from optical energy deposition to pressure, which can also provide image contrast. In particular, the Grueneisen coefficient increases with the equilibrium temperature (e.g., ~5% per Kelvin in water), which enables PAT to image temperature with a sensitivity of 0.16 °C for monitoring of thermal therapy.47–49
Each mK of laser-induced temperature rise in soft tissue yields ~800 Pa initial pressure, which is on the order of the sensitivities of typical ultrasonic transducers. The conversion from temperature rise to pressure is related to two parameters: the volume expansion coefficient and the compressibility. It is fortunate that the two parameters conspire to enable PAT to be applicable in tissue with a safe temperature rise.
PAM can image optical-absorption contrast at depths beyond the optical transport mean free path with a ratio of penetration limit to depth resolution greater than 100 [Fig. 1(a)].32,33 For ease of system operation, an optical fiber is used to deliver the excitation laser light, which is monitored by a photodiode for energy calibration. A conical lens is used to form a donut-shaped beam (i.e., dark field) so as to minimize the effect of surface photoacoustic signals on the deeper signals. The beam is then weakly focused into the tissue while the ultrasonic transducer focuses coaxially into the same region for confocal excitation and detection. The photoacoustic time-of-flight signal is recorded at each location of the ultrasonic transducer. By assuming the speed of sound to be 1.54 mm/μs, the time-of-flight is converted into depth, and consequently a 1D depth-resolved image (A-line) is formed. Linear or raster scanning over the tissue produces 2D or 3D tomographic images. Here, the broadband ultrasonic detector has a numerical aperture of 0.44 and a center frequency of 50 MHz. As a result, the lateral resolution, determined by the focal diameter of the ultrasonic transducer at the center frequency, measured ~45 μm. The axial resolution, determined by the ultrasonic bandwidth, measured ~15 μm. The maximum imaging depth, determined by the ultrasonic attenuation in tissue, measured >3 mm.32,33 A representative PAM image is shown in Fig. 1(b).
The image resolution and the penetration limit are scalable with the ultrasonic frequency. For example, employing a 5-MHz ultrasonic transducer and an 804-nm near-infrared light source scaled the penetration to 30 mm, whereas the resolution is scaled up to 144 and 560 μm in the axial and lateral directions, respectively.50 The near-infrared window is conducive for deep imaging because of the low hemoglobin absorption. Here, photoacoustic microscopy turns into macroscopy, which can potentially provide penetration up to the optical hard depth limit (~50 mm) at <1 mm resolution. Such a system can potentially image deep human tissue, such as the sentinel lymph node (SLN), which has an average depth of 12±5 mm in ultrasound images.50 An in vivo image of a rat overlaid with additional tissue was acquired with this system [Fig. 1(c)], where the SLN accumulated injected dye through the lymphatics.50 This imaging capability followed by needle biopsy can potentially transform SLN mapping—used for breast cancer staging—from a surgical procedure into a minimally invasive procedure.
The spatial resolution of the original PAM system can also be scaled down in dimension. Focusing light through an objective lens with a numerical aperture of 0.1 yields a 5 μm lateral resolution, which is limited by the optical focal diameter. Such a resolution allows in vivo imaging of capillaries—the smallest blood vessels—as shown in Fig. 1(d), where single files of red blood cells are imaged.51 While light delivery and ultrasonic detection are combined through a beam combiner in this work,51 they were also combined earlier through a ring ultrasonic transducer.52 If the numerical aperture is increased, sub-μm resolution can potentially be achieved. Of course, optical focusing is effective only within the soft depth limit.28 Thus far, a 0.7 mm penetration has been achieved.
Multiscale systems biology, recognized as an important future direction of biomedical sciences, needs the support of multiscale imaging in vivo with common signal origins. To understand the workings of a whole biological system, biological components spanning multiple spatial scales—subcellular organelles (sub-μm scale), through cells (μm), to organs (cm)—must be integrated. As demonstrated above, photoacoustic microscopy and macroscopy are uniquely positioned for such multiscale imaging with the same contrast mechanism—optical absorption.
Photoacoustic sensing has recently been shown to detect single melanoma cells in lymph vessels in vivo in small animals.53 This technology, coined in vivo photoacoustic flow cytometry, has potential for in vivo detection of circulating tumor cells in humans. Cells flowing through a lymph valve are naturally focused into a single file, which facilitates single cell detection in a lymph vessel. The lymph transparency presents low background and consequently facilitates high-sensitivity detection. Melanin in the melanoma cells provides high optical absorption for photoacoustic detection without the use of exogenous labeling, which can be toxic. Nonetheless, exogenous contrast agents can be used to provide molecular specificity for targeted biomarkers.53 Several optical wavelengths can be used to detect multiple types of cells of different characteristic optical absorption spectra. Toward human applications, it was suggested that superficial lymphatics in the leg be first targeted.54 The concept can also be applied to blood vessels although background due to erythrocytes can compete with signals from other absorbers.55–58 Currently, the technology is limited to non-imaging sensing; however, PAM can be explored for imaging.
In circular-view photoacoustic computed tomography, ultrasonic detection follows a ring.4,59,60 A pulsed laser beam is expanded to illuminate the object. An ultrasonic transducer scans around the object in a circle to acquire the photoacoustic data. An inverse algorithm then converts the photoacoustic data to the initial photoacoustic pressure distribution, which is an image of the cross section containing the scanned circle. Representative functional images are shown in Fig. 2(a).4 The ultrasonic transducer has a broad bandwidth centered at 3.5 MHz, yielding an in-plane resolution of ~200 μm. The out-of-plane resolution, determined by the cylindrical focusing of the transducer, is ~1 mm. To accelerate the imaging speed, an array of ultrasonic transducers in a ring has been adopted.61
In planar-view photoacoustic computed tomography, ultrasonic detection follows a plane.34 Optical interferometry, instead of the piezoelectric effect, can be used to sense the ultrasonic displacement62–65 or pressure.66,67 When a solid planar Fabry-Perot interferometer was constructed,67,68 a dielectric or polymer spacer was sandwiched between a pair of mirrors. A focused laser beam was optically scanned across the surface of the interferometer. The detected power fluctuation of the reflected laser beam was converted to the optical thickness of the interferometer, which was further converted to pressure. Then, an inverse algorithm was used to reconstruct a photoacoustic image [Fig. 2(b)].68,69
Optical detection offers primarily three advantages.67 First, the optical detector can be made optically transparent, facilitating light delivery to the object for full-field imaging. Second, the spatial sampling frequency can be increased without sacrificing the sensitivity, allowing high-resolution imaging. Sensitivities of tens of Pascals have been achieved at 20-MHz bandwidth.67 Such sensitivities are comparable with those of 1-mm-diameter piezoelectric receivers and better than those of smaller-diameter ones. Third, the frequency response can be stretched to low frequencies, potentially enabling the detection of the interior of a large heterogeneity. Presently, the technology is limited by the system complexity and the read-out speed.
In spherical-view photoacoustic computed tomography, ultrasonic detection follows a truncated spherical surface.70 In a recently constructed small-animal whole-body photoacoustic tomography system,71 the animal, immersed in coupling liquid, is rotated inside a virtual spherical surface, on which a concave arc-shaped array of 64 piezo-composite ultrasonic transducers is situated. Two expanded counter-propagating laser beams illuminate the small animal orthogonally to the plane of the array. At 755 nm wavelength, the internal organs of a nude mouse were imaged [Fig. 2(c)].71
In linear-view photoacoustic computed tomography, ultrasonic detection follows a line.72–77 Linear ultrasound arrays in medical ultrasound imaging systems can be leveraged. The ultrasound probes are typically handheld and can therefore adapt to various anatomical sites conveniently. In ultrasound imaging, a subset of the array elements is switched on at a time for ultrasound transmission and detection. Shifting the selection of the subset across the array forms a two-dimensional image. In PAT, the entire array can receive ultrasonic signals simultaneously because photoacoustic excitation usually covers a large volume. Naturally, ultrasound and photoacoustic images can be displayed side-by-side to present complementary contrasts.
Photoacoustic computed tomography can be based on linear or planar ultrasonic transducers instead of point detectors.40,78–80 In this case, the detected signals represent an integration of pressure over cylindrical or planar surfaces with varying distances to each transducer. Planar ultrasonic transduction is of particular interest for primarily two reasons. First, the forward problem represents a standard Radon transformation used in x-ray computed tomography. Therefore, the associated inverse algorithms can be used directly for PAT. Second, the spatial resolution is limited only by the ultrasonic bandwidth. In conventional photoacoustic computed tomography, point ultrasonic detectors are approximated by finite-size detectors, resulting in a tradeoff between the detection sensitivity and the lateral resolution. If the detectors are too small, the sensitivity is compromised; if too large, the resolution becomes poor. Such a compromise is devoid in PAT based on planar ultrasonic transducers.
PAT lends itself to breast imaging. Breast tissue has relatively homogeneous (~±5%) speed of sound, and breast tumors tend to present higher total blood concentrations and lower oxygen saturations of hemoglobin than surrounding tissues. Figure 3(a) shows an in vivo photoacoustic image of a human breast,81 where the ultrasonic detection follows an arc.37 The imaged lesion measured ~18 mm and centered at 23 mm deep from the laser-illumination surface; it was confirmed by biopsy to be an invasive carcinoma. Figure 3(b) shows another in vivo photoacoustic image of a human breast,82 where the ultrasonic detection follows a plane.83 The lesion was identified as a ring-shaped structure having an average optical absorption contrast of 1.6 relative to the background. The lesion was estimated to be 30 mm in diameter, which matched the pathologically estimated size (26 mm).
PAT is well suited for molecular imaging. While functional imaging is based on endogenous contrast, molecular imaging is based on exogenous contrast—which essentially stains invisible biomarkers in vivo. Functional imaging maps physiological activities usually at the tissue or organ level, whereas molecular imaging measures biological and pathophysiological processes at the molecular level. Simultaneous functional and molecular PAT of glioblastoma brain tumors was demonstrated in live mice.13 A small-molecule organic dye conjugated with a cyclopeptide serves as the molecular imaging probe to target biomarker αvβ3 integrin, which tends to over-express in angiogenesis. The functional image shows tumor hypoxia, whereas the molecular image shows tumor over-expression of integrin.
PAT can track various nanoparticles, such as metallic nanoshells84, nanorods10,85–89, nanocages90, carbon nanotubes9, and dye-embedded nanoparticles91. The large optical absorption cross sections of nanoparticles92 provide excellent contrast for PAT. Unlike small-molecule organic dyes, nanoparticles resist chemical or thermal denaturation and photobleaching. However, further toxicity studies toward human applications are needed.
The surfaces of gold nanoparticles exhibit good biocompatibility and can accommodate functionalities by conjugating with targeting peptides or antibodies.11,87,93 So far, biomarkers targeted in PAT include EGFR11, HER-287 and CXCR-493, which have a propensity to overexpress in cancers such as the breast cancer, as well as ICAM-1 in inflammatory responses.88 In metallic nanoparticles, the optical absorption is based on surface plasmon resonance, and the peak absorption wavelength can be tuned throughout the visible and near-infrared spectral region by varying the physical dimensions.
Dye-embedded nanoparticles can also function as an effective contrast agent.91 Indocyanine green (ICG) is an organic dye that has been approved for human use. It has strong optical absorption around the 800-nm near-infrared wavelength. Although free-form ICG has been demonstrated to enhance photoacoustic contrast in vivo,94 the half-life of ICG in the blood stream is only several minutes, limiting the time window for sustained monitoring. Encapsulation of ICG in nanoparticles increases the circulation time significantly.91 In addition, encapsulation provides surfaces for adding functionalities, increases the dye concentration for greater optical absorption, and shields the dye molecules from destabilizing effects due to the biological environment. ICG-embedded nanoparticles conjugated with HER-2 antibodies have been shown to target breast cancer and prostate cancer cells in vitro.91
Molecular PAT of the lacZ reporter gene was demonstrated in living small animals.8 Reporter gene imaging, a form of molecular imaging, can facilitate understanding the nature of gene expression during in vivo development and pathogenesis. While conventional reporter gene imaging is based on fluorescent proteins such as the well-known green fluorescent protein, photoacoustic reporter gene imaging is based on optical absorption that is converted into heat. The lacZ gene—one of the most widely used reporter genes—encodes β-galactosidase. A sensitive chromogenic assay with X-gal—an optically transparent substrate—yields a stable dark blue product, which serves as an optical absorption contrast agent for PAT. Because the absorption of the blue product peaks around 650 nm, multi-millimeter penetration can be achieved. However, if multi-centimeter penetration is desired, new reporter genes that produce gene-expression products absorbing in the near-infrared spectral region need to be developed. By providing high spatial resolution at depths beyond the optical transport mean free path in vivo, PAT is expected to become a complementary tool to conventional optical imaging of reporter genes.
The sensitivity of PAT is competitive among the various molecular imaging modalities although such a comparison depends on many parameters as well as the biomedical applications.95 When intrinsic biological contrasts due to such as red blood cells or melanoma cells are sensed, single-cell sensitivity has been achieved.51,55 When extrinsic small-molecule contrast agents are imaged, PAT can provide a sensitivity of the order of ~10 nmol/L or ~fmol/voxel at ~0.3 mm resolution.13,96 Pure optical fluorescence imaging can provide a sensitivity of the order of ~nmol/L or ~fmol/voxel at ~1 mm resolution,97 where the resolution is worse than that of PAT and the sensitivity is slightly better. Nuclear imaging such as PET can provide a sensitivity of the order of ~pmol/L or ~0.1 fmol/voxel at ~3 mm resolution, where the resolution is much worse than that of PAT and the sensitivity is better.98,99 However, the sensitivity of PAT can be improved potentially by increasing parameters such as the local optical fluence, the optical absorption cross section of the contrast agent, the number of signals for averaging, and the voxel size. For example, when extrinsic nanoshell contrast agents are imaged, PAT can provide a sensitivity of the order of ~10 pmol/L or ~0.0001 fmol/voxel at ~0.06 mm resolution.84 Because different modalities have dissimilar imaging capabilities, the sensitivity should not be the only parameter governing the selection of a tool for a particular application. In additional, nuclear imaging uses ionizing radiation, whereas PAT and optical fluorescence imaging use nonionizing radiation.
PAT has limitations as does any imaging technology. First, optical attenuation limits the penetration to ~5 cm when <1 mm resolution is desired in actual tissue.81,100,101 However, microwaves or radiowaves can be used for deeper excitation although the contrast origins differ. Second, ultrasound sustains strong reflection from gas–liquid or gas–solid interfaces due to the strong mismatch of acoustic impedances. Therefore, ultrasound signals cannot penetrate through gas cavities or lung tissues efficiently. For the same reason, ultrasonic detection requires direct contact between the ultrasonic transducers and the biological tissue. Usually, ultrasound coupling gel is applied to the tissue surface to avoid intervening air cavities. The development of noncontact optical detection of acoustic displacement can potentially overcome this problem.62–65 Third, ultrasound suffers significant attenuation and phase distortion in thick bones such as the human skull. Fortunately, unlike pulse-echo ultrasound imaging, PAT involves only one-way ultrasound attenuation through the skull. Sufficiently strong photoacoustic signals have been observed through Rhesus monkey skulls.102,103 The remaining challenge is to compensate for the phase distortion due to the skull.
The research field on PAT has gained significant momentum recently. The number of research articles on PAT has more than tripled since year 2003.104 PAT is expected to find broad applications in both biology and medicine. In the author’s opinion, the following in vivo applications hold particular promises and may come to fruition in the near future: melanoma detection,32,53,56,58 intravascular catheter imaging,105–107 endoscopic imaging, simultaneous functional and molecular imaging,13 reporter gene imaging,8 sentinel lymph node mapping,50,59 breast imaging,81,82,108–110 and multiscale imaging.
The author acknowledges the support by the National Institutes of Health Grants R01 EB000712, R01 NS046214, R01 EB008085, and U54 CA136398. Thanks to Song Hu for providing Fig. 1(d).
9. Competing interests statement
The author has a financial interest in Microphotoacoustics, Inc. and Endra, Inc., which however did not support this work.