Advanced optoacoustic methods for multiscale imaging of in vivo dynamics

Visualization of dynamic functional and molecular events in an unperturbed in vivo environment is essential for understanding the complex biology of living organisms and of disease state and progression. To this end, optoacoustic (photoacoustic) sensing and imaging have demonstrated the exclusive capacity to maintain excellent optical contrast and high resolution in deep-tissue observations, far beyond the penetration limits of modern microscopy. Yet, the time domain is paramount for the observation and study of complex biological interactions that may be invisible in single snapshots of living systems. This review focuses on the recent advances in optoacoustic imaging assisted by smart molecular labeling and dynamic contrast enhancement approaches that enable new types of multiscale dynamic observations not attainable with other bio-imaging modalities. A wealth of investigated new research topics and clinical applications is further discussed, including imaging of large-scale brain activity patterns, volumetric visualization of moving organs and contrast agent kinetics, molecular imaging using targeted and genetically expressed labels, as well as three-dimensional handheld diagnostics of human subjects.


Introduction
Living systems exhibit complex, multi-level processes whose behaviour is difficult to predict or understand by making observations at a single spatial or temporal scale. Diseases are often manifested via anatomical alterations or functional failures at the organ or whole-body level, yet their precursors are most efficiently detected by specific molecular targeting and imaging

S. Gottschalk
Sven Gottschalk holds a diploma in chemistry and a PhD degree from the University of Bremen, Germany. After researching on contrast agent development for magnetic resonance imaging at the Max Planck Institute for Biological Cybernetics in Tübingen, Germany, he joined the Institute for Biological and Medical Imaging (IBMI) in Munich. At IBMI, his research focus is on the development of biomedical applications of state-of-the-art multi-spectral optoacoustic tomography and microscopy systems. He is currently interested in functional optoacoustic neuroimaging and the application of novel imaging probes and contrast agents for optoacoustic imaging.
at the cellular or sub-cellular scales. Similarly, many biological processes are manifested at multiple temporal scales, e.g. local neural activity occurring on a millisecond scale is closely linked to much slower cerebral hemodynamic changes through a mechanism known as neurovascular coupling.
In vivo imaging across multiple scales is commonly associated with challenging compromises between the achievable contrast, imaging speed and spatial resolution. 1 For example, ultrasound (US) imaging is capable of imaging whole mammalian organisms with high imaging speed accounting for fast motion and perfusion but it chiefly captures mechanical tissue properties or blood flow. 2 Conversely, optical imaging uses contrast mechanisms that offer a highly versatile ability to visualize biological processes at the cellular and molecular levels. By using powerful new classes of probes based on fluorescence dyes, reporter genes and nanoparticulate agents, previously invisible processes associated with tissue function, disease progression and treatment can now be sensed, both in real time and longitudinally. In particular, the Nobel prize winning discovery of fluorescent proteins led to reporter molecules that enable intrinsic tagging of cells, thereby facilitating the observation of cellular or subcellular activity, from gene expression and protein function to signaling pathways. While these breakthroughs provided new windows for microscopically interrogating systems level biology, 3,4 inherent optical limitations restrict the effective imaging depth of most optical microscopy techniques to below a millimeter in highly scattering tissues. 5 Macroscopic imaging approaches, such as fluorescence molecular tomography (FMT) 6 make use of the reduced haemoglobin absorption in the 650-950 nm spectral window to visualize optical contrast through several centimetres in highly vascularized mammalian tissues. Promising new developments include the introduction of near-infrared-shifted fluorescent molecules that can be used for in vivo labeling of deep tissue functional and molecular processes. [7][8][9] However, in-depth optical observations remain complicated due to intense photon scattering that contributes to a significant resolution loss and limited quantification capacity beyond a few hundred microns depth.
Optoacoustic (photoacoustic) imaging is increasingly attracting the attention of the biomedical research community due to the important new features that it added to the existing imaging toolset. The technique capitalizes on the inherent advantages of both optics and ultrasound as it uses short-pulsed light radiation as probing energy and detects ultrasound generated by photon absorption and thermoelastic expansion. 10 As a result, optoacoustics reports on the versatile optical absorption contrast but relative to other optical methods provides a sort of 'supervision' by exploiting the low scattering of ultrasound to break through the barriers imposed by optical diffusion. Furthermore, multi-spectral optoacoustic tomography (MSOT) readings based on multi-wavelength excitation enable identifying chemical

S. Shoham
Shy Shoham is Professor of Biomedical Engineering at the Technion -Israel Institute of Technology, and a Visiting Professor of Neuroscience at NYU Langone Medical Center. He heads the Technion's Neural Interface Engineering Lab (NIEL), which develops and applies photonic, acoustic and computational tools for spatiotemporal interfacing with neural circuits. Prof. Shoham holds a BSc in Physics from Tel Aviv University, a PhD in Bioengineering from the University of Utah, and was a Lewis-Thomas postdoctoral fellow at Princeton University. He is a member of the editorial boards of the Journal of Neural Engineering and of Translational Vision Science & Technology. composition of biological samples via spectroscopic analysis, and hence render additional information not captured by other modalities. 11 The label-free optical absorption contrast explored in biomedical optoacoustics readily provides 'illuminating' information regarding the presence of intrinsic tissue components such as oxy-and deoxy-haemoglobin, melanin, bilirubin, lipids and water. 12 The strong optical absorption of haemoglobin allows the visualization of vascular structures and hemodynamic responses, maintaining sub-millimeter resolutions at depths of several centimetres within highly scattering living tissues for nearinfrared (NIR) wavelengths. 13 Furthermore, bio-chromophores have specific spectral signatures that allow them to be distinguished from each other within an integrated absorption signal, with their relative signals contributing diverse information about function and/or pathological status of the tissue being examined. 14 Another important property of the optoacoustic methodology is the great assortment of exogenous compounds that can be explored for specifically enhancing the absorption contrast. Due to the versatility and wide availability of optical labeling approaches, optoacoustic imaging studies of e.g. gene expression or targeted bio-markers can be done in a similar way by resolving the accumulation of agents with specific spectral signatures. 12 These include fluorescent molecular probes, fluo-and chromo-proteins, quantum dots, gold-, carbon-, and polymer-based nanoparticles, porphyrins or even contrast agents used in other imaging modalities, such as microbubbles or iron oxide particles. Finally, optoacoustics provides a unique multiscale imaging capacity, allowing bridging the gap between the microscopic and macroscopic realms with the same type of contrast. 13 Taken together, these key enabling properties have prompted the development of high throughput optoacoustic systems for in vivo pre-clinical and clinical imaging, further providing high sensitivity and spatial resolution, portability, as well as real-time operation capability. The temporal dimension is of paramount importance in biological observations in allowing the study of complex interactions that are otherwise invisible in single snapshots of living systems. In this context, real-time optoacoustic imaging in two or three dimensions has been made possible by simultaneous detection of dense tomographic information around the imaged object for each excitation laser pulse. 15 The most recent efforts in the field of optoacoustic functional and molecular imaging have established new technological platforms employing spherical matrix arrays, parallel acquisition hardware, GPU-based data processing and fast laser tuning systems in order to enable acquisition and visualization of spectroscopic information from entire tissue volumes at video rates. This has set the stage for the so-called five dimensional (real-time threedimensional multi-spectral) optoacoustic imaging to emerge as a new tool poised to offer unprecedented insights in biological discovery. 11 This review highlights on the most recent advances that enable powerful new applications for visualizing multiscale in vivo dynamics, from neural activation and real-time kinetics at the organ level to whole-body longitudinal studies of tumour progression. Optoacoustic contrast agents are classified according to their in vivo stability, specificity, and dynamic contrast enhancement properties while multiple examples of novel applications involving visualization of multiscale in vivo dynamics are further introduced.

Optoacoustic approaches enabling imaging of biological dynamics
In this section, existing optoacoustic imaging approaches are reviewed, focusing on the temporal scales covered by each approach and the amount of information delivered. The principles of optoacoustic signal generation are first introduced along with the general limitations affecting the potentially achievable imaging rate. The dynamic capabilities of multi-dimensional imaging systems operating at microscopic to macroscopic spatial scales are then described.

Optoacoustic signal generation
Optoacoustic signals correspond to pressure waves generated thermoelastically by absorption of photons. Typically, a pulse of light gets absorbed in the tissue, depositing a tiny amount of heat in the optical absorption zone. The instantaneous thermal expansion causes an initial pressure increase and subsequent emission of stress (ultrasound) waves (Fig. 1).
If the characteristic thermal diffusion length of the medium is shorter than the spatial resolution of the imaging system during the laser pulse, the so-called thermal confinement regime is maintained. This corresponds to no leakage of energy out of the effective optical absorption zone, and maximal thermal energy densities are attained. Considering a typical thermal diffusivity of D B 0.1 mm 2 s À1 in soft biological tissues, this condition can be readily met for typical biological imaging applications that use nanosecond-duration laser pulses. 16 In addition, a stress confinement regime is fulfilled if the laser pulse duration is shorter than the time required for the generated stress wave to propagate out of the heated region defined by the effective spatial resolution. 16 For instance, given a common speed of sound of B1540 m s À1 in soft tissues, the stress confinement condition would be readily fulfilled for the laser pulse duration of B10 ns and a typical diffraction-limited spatial resolution of B100 mm of the imaging system (corresponding to ultrasonic detection bandwidth in the 10 MHz range). Fig. 1 Schematic illustration of the optoacoustic signal generation and detection. Short light pulses at selected optical wavelengths are absorbed by the tissue chromophores and contrast agents, leading to instantaneous heating and thermal expansion. As a result, ultrasound pressure waves are excited and measured around the imaged object.
Using pulse durations satisfying both thermal and stress confinement regimes, the initial local pressure rise p 0 induced by laser energy deposition and subsequent instantaneous heating can be simply expressed via 17 being m a and F the local optical absorption coefficient and light fluence, respectively. G is the dimensionless Grüneisen parameter, which lumps together the thermoelastic properties of the medium, i.e. 16 where b is the thermal expansion coefficient, c is the speed of sound and C p is the specific (per unit mass) heat capacity at constant pressure. In simple terms, the locally-induced optoacoustic pressure is proportional to the thermoelastic constant of the medium, the local optical absorption coefficient and the amount of light energy reaching this point, the latter being basically proportional to the per-pulse laser energy. In the case of fluorescent agents, the optoacoustic generation efficiency is reduced since only non-radiative relaxations directly contribute to the elastic wave conversion, in which case the effective thermoelastic conversion efficiency G eff can be expressed as 18,19 with QY representing the quantum yield of the molecule. Note that light absorption in solid nanoparticles is further accompanied by a different heat conduction mechanism than for small molecules dissolved in a liquid. Thereby, the thermoelastic conversion efficiency is further conditioned by the heat propagation rate and the thermal resistance at the surface of nanoparticles, which can significantly affect G eff . 20 Safe application of optoacoustic imaging to human subjects implies that the light fluence on the skin surface is maintained below the permissible exposure standards. ANSI safety limits establish that the per-pulse laser fluence must be kept below 20 mJ cm À2 for wavelengths below 700 nm, increasing towards 100 mJ cm À2 at 1050 nm. 21 Considering a typical value (G B 0.2) of the Grüneisen parameter in soft biological tissues and the permissible fluence levels, the maximum local optoacoustic pressure that can be induced in blood is approximately 738 kPa at 584 nm and 26 kPa at 797 nm (two isosbestic wavelengths of haemoglobin). ANSI safety standards further establish the maximum mean laser intensity for continuous exposure of the same area. This limit is 200 mW cm À2 for wavelengths below 700 nm, increasing to 1000 mW cm À2 at 1050 nm. 21 The ANSI limits result in a trade-off between the maximal per-pulse energy and pulse repetition rate of the laser that fulfill both safety criteria.
The last part of the optoacoustic imaging problem concerns the propagation of the generated pressure waves towards the US detectors (Fig. 1). The distance from the absorbers to the measurement location is encoded by the time-of-flight of the detected US signals. On the other hand, the actual size of the absorber is encoded by the duration of the detected timeresolved signal, or equivalently by its frequency spectrum.
In general, higher US frequencies are generated by smaller absorbers. Typical optoacoustic waveforms generated in highly heterogenous living tissues carry a mixture of signals generated by absorbers of different sizes. Thus, they generally contain a wide range of frequencies, from several tens of kHz up to several tens of MHz, which requires the use of wideband detectors in the imaging system.

Optical-resolution microscopy
Since optoacoustics involves both optical excitation and acoustic detection, it offers a great diversity of possible embodiments of the imaging device. One common approach is based on selective excitation of a spot of the surface of the tissue by focusing the incident laser beam. This modality, often termed as opticalresolution photoacoustic microscopy (OR-PAM), can resolve optical absorption contrast in cellular and sub-cellular structures with spatial resolution limited by optical diffraction (Fig. 2). Note that point-by-point scanning of the tightly focused laser beam is necessary, which implies no continuous illumination of the same spot. Thereby, limits on the mean power are not applicable and only the peak fluence needs to be capped. As a result, pulse energies below 1 mJ are typically employed to avoid tissue damage. 10 The light beam is generally shaped so that the depth range of interest is covered within the optical depth-of-field. In biological tissues, however, the reachable depth is limited by light scattering, which prevents effective focusing beyond B1 mm depth, 5 similarly to the optical microscopy techniques. Yet, in optoacoustic microscopy, one-dimensional (1D) image profiles along the depth direction can be readily obtained from the optoacoustic signal collected by a coaxially aligned ultrasound transducer following a single laser pulse. 23,24 Two-(2D) or three-dimensional (3D) images are subsequently formed by raster-scanning the excitation beam along the region of interest and superimposing the 1D profiles acquired for each position.
The lateral resolution is given by the width of the focused light beam and is generally in the micrometer range. 13 In general, a tighter focus implies better lateral resolution at the expense of the depth of field. This can be achieved e.g. with a transillumination approach, where ultrasound signals are detected from the opposite side of the sample. Trans-illumination can be used to efficiently combine optical and optoacoustic microscopy into an integrated multi-modal system. 25,26 However, an epi-illumination approach is generally preferred as it produces less acoustic distortions, particularly for thick samples. To date, multiple epi-illumination-based systems have been suggested. [27][28][29][30][31] The lateral resolution can further be enhanced with superresolution approaches based on several non-linear mechanisms. [32][33][34][35][36][37] The axial (depth) resolution, typically on the order of tens of microns, 10 is determined by the frequency bandwidth of the collected optoacoustic signals. In principle, ultra-wideband detection (in the GHz range) of optoacoustic responses excited with very short (sub-nanosecond) laser pulses provides a way of bringing the axial resolution down to micron levels. 38 However, these ultrasound frequencies are strongly attenuated in tissues, which limits the effective penetration depth of this approach to a few tens of microns. 39 Much like in other raster-scan-based imaging methods, such as confocal or multi-photon microscopy, the time required to form an image is determined by the pulse repetition frequency of the laser and the number of scanning points. In contrast to those methods, no depth scanning is needed. However, the separation between subsequent laser pulses must be longer than the time-of-flight of ultrasound waves so that no overlap is generated between the acquired signals. For instance, it takes about 0.67 ms for the acoustic wave to travel a distance of 1 mm in soft tissues, so that the laser pulse repetition rate must be kept below 1.5 MHz to avoid overlap. Thus, in an ideal case scenario, a typical raster-scan of 500 Â 500 points would require about 0.17 s if no signal averaging is applied. Faster imaging can be achieved if the scanning is limited to a single axis for rendering cross-sectional (B-mode) images, which has been used e.g. to track individual red blood cells in capillaries. 40,41 Several approaches were suggested to further increase the imaging speed. For example, by combining multifocal illumination with a microlens array and parallel signal detection with an US array, the imaging speed was enhanced by a factor of 3 to 4. 42 A different class of approaches consist in randomaccess scanning with a digital micro-mirror device (DMD), where only selected points in a 2D region are acquired, 43 or in mechanical scanning along an arbitrary 3D trajectory. 44

Scanning-based acoustic-resolution approaches
Despite its excellent spatial resolution performance, the effective penetration depth of OR-PAM is severely limited by photon scattering, similarly to other optical microscopy techniques. For depths beyond B1 mm, progressive randomizations of the propagation directions of photons prevent light focusing with standard optical elements. Optoacoustic imaging at greater depths can be instead achieved with acoustic resolution via raster-scanning of focused ultrasound detectors. 45 For this, higher pulse energies, typically in the millijoule range, 10 are required to compensate for light scattering and attenuation in deep tissues. When using spherically-focused US detectors, 3D images can be formed in the same manner as in OR-PAM, i.e. by simply stacking depth profiles acquired from individual laser shots. Optoacoustically-generated waves in tissues typically exhibit an ultrawide bandwidth, thus cannot be efficiently focused, especially in the lower frequency range. Thereby, more sophisticated image formation approaches based on properly modelling the frequency-dependent sensitivity field of the transducer are generally preferred for optimizing image quality. [46][47][48] Resolution and imaging depth can be easily scaled by properly selecting the effective detection bandwidth of the ultrasound detector. Indeed, it is possible to operate at acoustic resolution covering spatial scales ranging from major vessels 49 to capillaries 50 all the way to individual cells. 51 Both axial and lateral resolutions are determined ultrasonically and hence can be enhanced using higher frequency detectors. Higher resolution however comes to the detriment of the achievable depth due to the reduction of the depth-of-field of the transducer and the increase in acoustic attenuation. A hybrid optical-acoustic resolution optoacoustic microscopy approach has been realized using coaxial illumination and detection design. 52 This ascertains smooth transition between optical resolution in superficial microscopic imaging into ultrasonic resolution when imaging at greater depths within intensely scattering tissue layers.
Alternative scanning approaches based on signal acquisition at a set of projections around a biological sample have been suggested. This imaging approach, referred to as optoacoustic tomography, essentially replaces spherically focused detection with tomographic scanning of unfocused or cylindrically-focused detectors. Here image formation is based on reconstruction algorithms, such as e.g. back-projection or model-based inversions, [54][55][56][57] analogous to those used in other tomographic imaging modalities. In fact, the first in vivo optoacoustic images from small animals were obtained with a tomographic optoacoustic system based on a cylindrically-focused transducer scanned around the head of a rat. 58 The same tomographic scanning geometry has also been used e.g. for imaging small animals of different sizes [59][60][61] or human fingers. 62 Other tomographic imaging approaches based on cylindrical 63 or spherical 64 trajectories have further been suggested. High energies per pulse, typically tens of millijoules, 10 and large illumination areas are generally employed in optoacoustic tomography as unfocused detection is commonly used to reach deeper areas. An all-optical tomographic optoacoustic imaging system has been implemented which uses, in addition to the optoacoustic excitation light, a focused interrogation laser beam that is scanned along the surface of a Fabry-Pérot polymer film sensor that changes its local thickness in response to the impinging optoacoustic waves. 65 In this way, tomographic information along a planar surface is collected using unfocused (point) detection, resulting in accurate 3D reconstructions (Fig. 3). In some cases, the distinction between optoacoustic microscopy and optoacoustic tomography is rather vague. For example, a combination of raster-scans with a focused transducer at multiple orientations has also been suggested for high resolution imaging of zebrafish larvae. 66 Analogously to optoacoustic microscopy, the image acquisition speed in scanning optoacoustic tomography is determined by the number of measurement locations. Data acquisition can be accelerated by using compressed sensing schemes. [67][68][69] Parallelization of an all-optical detection approach was proposed by detecting optical integrating line detectors with a CCD camera. 70 Additional acceleration can be achieved by scanning ultrasound arrays instead of single detectors. Various scanning geometries have been proposed using linear and curved arrays in order to render whole-body 3D images from small animals [71][72][73] or from the human breast. 74

Cross-sectional array-based imaging
The tomographic optoacoustic methods described above are clearly fundamentally constrained by the need for mechanical scanning. This constraint can be completely or partially removed when rendering images using parallel optoacoustic data acquisition from multiple locations after every laser pulse. This is similar to B-mode ultrasonography based on linear arrays, which does not require mechanical scanning to form 2D images in real time. Cross-sectional optoacoustic imaging based on the same configuration has been developed, 75,76 where the illumination of the tissue is provided from the lateral sides of the transducer array via fibre bundles 76 or via an integrated laser diode module. 77 If all signals are simultaneously collected without multiplexing, the imaging frame rate is determined by the pulse repetition rate of the laser. This essentially implies that the temporal resolution is theoretically limited by the time-of-flight of pressure waves across the imaged region. 78 The main advantage of this approach is that it can be readily integrated into multi-modal imaging systems rendering co-registered ultrasound and optoacoustic images. [79][80][81] Furthermore, it is compatible with handheld operation, facilitating clinical translation. Conversely, while linear arrays are convenient for ultrasonography, they are highly suboptimal for optoacoustic imaging as the tomographic reconstructions manifest severe limited-view artefacts, which limit visibility of tissue morphology. 82 Indeed, it has been shown that accurate optoacoustic reconstructions imply collection of signals from a large tomographic view surrounding the imaged object. 83 For instance, MSOT small animal scanners use tomographic data collection with partial-or full-ring concave transducer arrays to render crosssectional reconstructions in real time (Fig. 4). 84,85 The technique uses advanced lasers with fast wavelength tuning capacities in order to acquire tomographic information at multiple illumination wavelengths. 86 The distribution of the intrinsic tissue chromophores and extrinsically-administered contrast agents are subsequently mapped using spectral unmixing approaches. 87 3D images of the entire animal can further be obtained by scanning the array along the elevational dimension. 88 The important advantages of this approach related to real-time image acquisition have led to its widespread use in biological research. [89][90][91][92] For example, dynamic imaging of fast biological events is possible at video rate of 50 frames per second or faster, essentially limited by the pulse repetition frequency of the laser. 15 In this configuration, imaging is facilitated by horizontal placement of the mouse on a membrane surface without direct contact with water, thus assuring ease of handling and highthroughput performance. 88 Handheld optoacoustic probes based on similar cylindrically-focused concave arrays have further been developed (Fig. 4), outperforming linear-array-based optoacoustic imaging in terms of quantification performance and image quality. 86,93

Volumetric 4D and 5D imaging
The capability for real-time imaging introduced through the development of cross-sectional (2D) imaging systems represents a central milestone in dynamic optoacoustic imaging, yet does not address the requirement by many applications for rapid imaging across entire volumes. Cross-sectional imaging also has some associated drawbacks, like the need for focusing along the elevational dimension, which is responsible for highly anisotropic resolution on all axes and other artifacts associated with out-of-plane signals. 95,96 It has been long recognized that the most efficient implementation of optoacoustic imaging implies simultaneous collection of time-resolved signals in 3D from as many locations (projections) around the imaged object as possible, thus avoiding limited-view effects and attaining excellent image quality. 71 Thereby, attempts to achieve high fidelity volumetric imaging were based on rotating (scanning) a set of unfocused detectors or sparse spherical detection arrays around the imaged object so that signals at a large number of locations are acquired for optimal 3D tomographic reconstruction. 71,74 However, these scanning-based approaches hindered imaging of fast biological processes and were prone to image artifacts caused by in vivo motion, e.g. breathing or heartbeat. In response, four-dimensional (4D) optoacoustic tomography approaches have been recently developed that employ matrix arrays with a large number of unfocused detection elements distributed on a spherical aperture. [97][98][99] This configuration, having relatively large and densely-packed detection elements, further ensures optimal signal-to-noise ratio performance and ability to collect high quality real-time volumetric data for each illumination pulse without signal averaging (Fig. 5a). Furthermore, the particular orientation of the detection elements provides good sensitivity in the imaged region of interest located around the geometric centre of the spherical detection geometry. The 4D optoacoustic imaging approach has enabled tracking of contrast agent perfusion in entire organs, 97 beat-by-beat imaging of fast beating murine heart 100 or monitoring the distribution of epileptic seizures in whole mouse brains. 101 The same configuration is further suitable for handheld operation mode, and human angiographic imaging at centimetre-scale depths was demonstrated attaining nearly isotropic spatial resolution on the order of 200 mm. 102 Hybrid combination of 4D optoacoustic and ultrasound imaging was further showcased by introducing passive absorbing elements. 103 Imaging across multiple spatial and temporal scales can be achieved by scanning the spherical array along a helical trajectory, a technique termed spiral volumetric optoacoustic tomography (SVOT). 73,104 In vivo experiments in mice demonstrated a wide  range of dynamic imaging capabilities for this method, from 3D high-frame-rate visualization of moving organs and contrast agent kinetics in selected areas to whole-body longitudinal studies with unprecedented image quality (Fig. 5b). The particular ability to deliver high resolution images at the whole-body scale while preserving ultrafast 3D imaging capability in smaller regions with the same type of contrast makes the SVOT technique unique among the existing pre-clinical imaging modalities.
In general, the biological information provided by optoacoustic imaging systems is encoded in five independent dimensions, namely, the three spatial dimensions, time and the optical excitation spectrum (wavelength). Multi-spectral 3D imaging is possible if a tunable laser or multiple laser sources operating at different wavelengths are used, while the temporal information can further be conveyed by subsequent acquisitions of the multi-spectral volumetric recordings. Recently, fast multi-spectral imaging capacity has been enabled with the introduction of fasttunable lasers capable of changing the emitted wavelength on a per-pulse basis. 11,86 By facilitating real-time multispectral 3D imaging, 11 all the five imaging dimensions can be simultaneously accessed (Fig. 6).
The effective multi-spectral imaging frame rate is then determined by the pulse repetition rate of the laser and the number of wavelengths considered. State-of-the-art five-dimensional (5D) optoacoustic systems support acquisition of spectrally-enriched 3D information at a volumetric frame rate of 20 Hz (considering imaging with five different wavelengths), providing unprecedented capabilities for imaging in vivo functional and molecular dynamics, also in a handheld configuration suitable for clinical application. 105,106 It should be noted that 5D data acquisition can be compromised due to presence of significant motion between images acquired at the different wavelengths, e.g. due to heartbeat, fast perfusion, or when operating the scanner in a handheld mode. To mitigate this problem, one may attempt using higher laser pulse repetition rates, which however may not be feasible due to data acquisition limitations or the safety limits on the average power exposure. An alternative solution consists in using multiple laser sources in a burst mode with a delay of a few microseconds. 107 The later approach enables functional optoacoustic imaging in the presence of very fast motion of up to 2 m s À1 maintaining relatively low laser pulse repetition frequencies so that the maximum permissible laser exposure levels are not exceeded.
Overall, the capacity to perform simultaneous imaging along all five dimensions comprises a unique enabling feature of the optoacoustic modality, poised to offer new insights into the workings of living organisms, particularly at the functional and molecular levels, as also elaborated in Section 4 of this review.

Dynamic contrast enhancement approaches
Parallel to the rapid technological developments, substantial efforts have been devoted to the engineering and optimization of functional and molecular contrast agents for optoacoustic imaging. In general, the optical absorption mechanism underlying optoacoustic signal generation offers high versatility in mapping the distribution of endogenous or exogenous light absorbing substances and excellent reviews covering optoacoustic contrast agents are available. 12,18,[108][109][110][111][112] Rather than distinguishing the agents based on their static chemical composition, herein we classify the contrast enhancement approaches according to the particular delivery method used, in vivo kinetics and biodistribution, as well as dynamic contrast modulation and molecular sensing properties.

Bio-marker delivery methods
3.1.1. Endogenous chromophores. Biological tissues contain a variety of endogenous chromophores with distinct absorption spectra that can be exploited for label-free structural and functional imaging with optoacoustics. In the visible and NIR ranges, light is mainly absorbed in living mammalian tissues by haemoglobin, melanin, lipids and water (Fig. 7). Differences in the absorption spectrum of haemoglobin associated to oxygen binding encode important information related to physiological activity. Since many diseases undergo structural changes at time scales ranging from days to weeks and months, imaging can be used to visualize and quantify these changes. For example, vascular structures can be mapped for depths of millimetres to centimetres within mammalian tissues 13 and accurate estimation of oxygen saturation is possible with proper models of light attenuation. 113 Due to the strong intrinsic haemoglobin contrast, optoacoustics represents a valuable tool to study the evolution of important hallmarks of cancer such as angiogenesis 53,114 and hypermetabolism. 115,116 Imaging the vascular fat deposition in atherosclerotic plaques is also possible at infrared wavelengths, 117,118 while the strong absorption of melanin can be exploited to characterize skin melanomas 49 and metastatic melanoma cells. 106 Measurable endogenous changes may also occur on significantly shorter time scales. For example, neuronal activity leads to complex hemodynamic responses in the brain via neurovascular coupling. Changes in blood oxygen saturation, total haemoglobin, blood volume, oxygenized and deoxygenized forms of haemoglobin can be readily monitored with optoacoustic systems operating at sub-second temporal resolutions. [119][120][121] Note that the presence of strong endogenous (background) contrast may on the other hand represent a problem when considering enhancing the contrast extrinsically as carried out in molecular imaging applications. With a typical 2 mM concentration of haemoglobin in blood, 128 vascular structures may conceal e.g. signals from genetically-expressed labels that can typically attain in vivo concentrations only in the order of mM, even at high expression levels. 129 Thereby, signal amplification approaches based on multispectral unmixing or dynamic contrast enhancement become essential in order to efficiently map the distribution of relatively low concentrations of contrast agents. 87,130 Background absorption along with optical scattering are also responsible for the strong attenuation of light in biological tissues. Optical attenuation is significantly stronger as compared with acoustic attenuation for frequencies below 20-30 MHz, thus it represents the main limiting factor for deep tissue imaging. 39 Light penetration is maximized in the so-called nearinfrared (NIR) window between 650 and 1350 nm, 127 while it is significantly aggravated by strong absorption of blood at shorter wavelengths and water at longer wavelengths ( Fig. 7). For deeptissue imaging purposes, excitation optical wavelengths within this range are therefore commonly selected, where optoacoustic imaging with centimetre-scale penetration and beyond is enabled. 102,131 The wavelength dependence of optical attenuation further contributes to the distortion (spectral colouring) of the optoacoustic signals originating from deep locations. 132 Thereby, the performance of the particularly employed unmixing approach, rather than the signal-to-noise performance of the imaging system, generally determines the minimum detectable concentration of extrinsically administered contrast agents.

Intravenous injection.
Intravenous injection is the most common approach to administer probes for enhancing the absorption-based contrast. Ideally, exogenous contrast agents optimized for deep tissue in vivo optoacoustic imaging should have (1) high absorption per molar concentration or per unit mass, preferably in the NIR window; (2) sharply peaked absorption spectra or other distinctive features enabling unambiguous differentiation from background absorption; (3) high photostability under biologically relevant energy levels of the excitation light and (4) low toxicity and viable clearance from the body. Other desirable features are application-dependent. For example, exogenous probes can greatly enhance optoacoustic contrast, particularly in the NIR window, where the relatively weak absorption of haemoglobin hampers visibility of deep-seated vasculature. The highest contrast enhancement is achieved a few seconds after the intravenous injection, when the local concentration of the undiluted bolus is maximal. 11 A real-time optoacoustic imaging system is needed to accurately capture such perfusion transients. 85,97 For unspecific vascular or organ perfusion studies, agents with rapid renal clearance are then preferable as this generally guarantees low toxicity. However, targeted agents are less likely to be delivered to the target tissue or cells if they are rapidly cleared, and a longer circulation time is desirable. The latter is affected by different physicochemical properties, such as the size, shape, chemical composition or surface modifications of the injected molecules or particles. The size or, alternatively, the hydrodynamic diameter (HD) is a parameter of particular relevance as it directly influences the clearance mechanism. For example, B6 nm is a known threshold for efficient renal clearance. 133 The circulation time for several optoacoustic contrast agents of different HDs is shown in Fig. 8a. Fig. 8b shows an example of the clearance of a smallmolecule-based agent (Alexa-Fluor 750) through the renal system as visualized with SVOT. The size of the agent is also important to overcome important circulatory barriers. For instance, capillary pores in normal tissues have typical apertures of 5-10 nm, thus generally only small molecules are able to extravasate healthy vessels. 12,134 Even more restrictive is the blood brain barrier (BBB), which enables only the passage of molecules smaller than 400 Da (o1 nm) with high lipid solubility. 135 Permeability is enhanced in leaky neovasculature associated with several diseases, where larger fenestrations in the capillaries allow the passage of agents up to approximately B100 nm in diameter. 136 Two distinct types of compounds used as extrinsically administered optoacoustic contrast agents are small-molecule dyes and nanoparticles. Many commercially-available biocompatible dyes, such as cyanine, squaraine, porphyrin derivatives, or boron-dipyrromethene (BODIPY) analogues, have been shown to absorb light at visible and NIR wavelengths. 141 Due to their small size (B1 nm), they can be fully and rapidly cleared from the body. 138 In fact, a number of organic dyes are FDA-approved and can be readily used in clinical trials involving optoacoustic imaging. Optical absorption spectra of major endogenous chromophores at typical concentrations occurring in living mammalian tissues. Melanin spectrum (brown) is shown for typical concentrations in the skin; 122 haemoglobin (red -oxygenated, blue -deoxygenated) for typical concentrations in whole blood (150 g l À1 -continuous lines) and average soft tissues (15 g l À1 -dashed lines); 123 water (cyan) for a typical concentration of 80% by volume in soft tissues; 124 lipids (yellow) for a concentration of 20% by volume. 125,126 The first (NIR -I) and second (NIR -II) windows, 127 where optical absorption is minimized, are indicated.
These include methylene blue (absorption peak at B664 nm), 142 Evans blue (absorption peak at B620 nm) 143 and indocyanine green (ICG, absorption peak at B808 nm in plasma). 106 Their small size and biocompatibility makes organic dyes adequate for extravascular and intracellular targeting. 144 Many dyes are also fluorescent and hence suitable for multimodal fluorescenceoptoacoustic imaging. Note that fluorescence is optimized either with a high extinction coefficient e or a high quantum yield QY, whereas the optoacoustic signal intensity is proportional to e(1 À QY). As a result, high extinction coefficient benefits both modalities whereas high quantum yield is less optimal for optoacoustic imaging. In fact, most organic fluorescent dyes, especially those with peak absorption in the NIR, usually have a low QY, making them highly suitable for constrast-enhanced optoacoustic imaging. On the other hand, many smallmolecule dyes have a relatively low molar extinction coefficient (Fig. 9a) and many of them suffer from high plasma protein binding and undesired aggregation, 141 which prevents rapid renal clearance. 138 To date, a myriad of different nanoparticles (NPs) have been used as optoacoustic contrast agents. [146][147][148][149][150][151] NPs refer to any structure with a size from 1 up to hundreds of nanometres that behaves as a unit. NPs have several advantages over smallmolecule-based contrast agents. For example, their size, shape and chemical composition influence their functional properties and can be tailored according to specific needs. 12 Due to prolonged circulation time, NPs smaller than 100 nm are particularly suitable for passive targeting based on enhanced permeability and retention (EPR). 152 Furthermore, their large surface allows attaching a large number of targeting moieties, thus increasing efficiency of binding to specific receptors. Some types of NPs can be further loaded with drugs or provide photothermal properties to be used as theranostic agents. 153 NPs provide significantly higher absorption per unit particle as compared with smallmolecule dyes, although the absorption per mass is generally similar (Fig. 9a). Also important is the fact that the optoacoustic signal generation mechanism in NPs is significantly affected by the heat propagation rate to the surrounding medium. Thereby, the relative optoacoustic signal intensities generated by different NPs can significantly differ from the corresponding relative optical absorbances ( Fig. 9a and b).
Biocompatible NPs are classified into organic and inorganic compounds based on their chemical composition 146,148,149 and into plasmonic and non-plasmonic particles based on the optical absorption mechanism. 109,110,112 Organic NPs are composed of organic molecules grouped together in nanostructures or  73 ICG 137 (note that ICG small molecules bind to albumin in blood, resulting in a hydrodynamic diameter of 11 nm that prevents kidney clearance 138 ); gold nanorods (AuNR); 139 liposomal ICG (Lipo-ICG) 137 and single-walled carbon nanotubes (SWNT). 140   combined through chemical bonds. For example, organic dyes can be encapsulated in nanodroplets, 154 liposomes 155 or virusmimicking nano-constructs. 156 These nanostructures generally provide improved photostability and prolonged circulation time with respect to the free dyes. 110 Self-quenching effects can also be produced when fluorescent molecules are clustered, which can enhance the generated optoacoustic signal. 157 On the other hand, aggregation of fluorescent molecules was reported to induce widening of the absorption spectrum towards the NIR region. 158 Another type of organic nanoparticles are semiconductor polymer nanoparticles (SPNs). 159 Recently, a new class of NIR SPNs has been shown to outperform highly-absorbing inorganic nanoparticles in terms of optoacoustic signal per unit mass (Fig. 9b) whilst providing high photostability. 145 Light absorbing porphyrin-lipid building blocks can self-assemble to form nano-vesicle structures (porphysomes) that can generate optoacoustic signals. 160 Other examples of organic NPs consisting e.g. of cellulose 161 or melanin 162,163 have been also shown to create usable optoacoustic contrast.
Plasmonic particles are a particularly favourable type of NPs as they are known to generate the strongest optoacoustic signal on a per particle basis. 12 The surface plasmon resonance (SPR) effect in noble metal NPs (e.g. gold or silver) enables strong and tunable optical absorption that is four to six orders of magnitude higher than that of single organic molecules. 164 Although silver NPs have been used as optoacoustic contrast agents, 165,166 gold NPs are considered preferable for in vivo optoacoustic imaging due to their better wavelength tunability properties and supposedly lower toxicity. 167,168 The absorption spectrum of gold NPs can be tuned by controlling their size and shape during synthesis. 169,170 Apart from standard nanospheres 171 and nanorods, 139,172 more complex gold NPs, such as nanocages, 173,174 nanoshells, 175,176 nanostars, 177,178 nanoprisms 179,180 or nanovesicles, 181 have been explored as optoacoustic contrast agents.
Another key advantage of gold NPs is the ability to chemically modify their surface in order to achieve better biocompatibility and functionalization properties. 182 Another major type of inorganic NPs is based on different carbon-based structures, e.g. carbon nanotubes, graphene nanomaterials and nanodiamonds. Single-walled carbon nanotubes (SWNT) are the most widespread carbon NPs, [183][184][185] with diameters of 1-3 nm and variable lengths, from nanometers to millimetres or even centimetres. They absorb light over a very broad spectrum and are also able to provide thermoacoustic contrast for electromagnetic waves in the GHz range, 186 although the per-particle optoacoustic signal strength is lower in SWNT than in gold NPs (Fig. 9b). NIR dyes or a thin layer of gold can be attached to their surface to enhance the generated optoacoustic responses and make the absorption spectra more distinct. [187][188][189] Other moieties can also be conjugated to SWNT for active molecular and cellular targeting. [190][191][192] Grapheneand graphene-oxide-based nanosheets have also been used as optoacoustic contrast agents, [193][194][195] demonstrating better dispersibility in biological systems. 108 Carbon nanodiamonds are also efficient light absorbers. By introducing neutral vacancies into their crystal lattice, an even higher optoacoustic signal than for gold NPs or SWNT can be generated for a similar particle size. 196 Other types of inorganic NPs have been used as optoacoustic contrast agents. In particular, nanostructures providing high absorption in the NIR range include copper sulphide NPs, [197][198][199] palladium nanosheets 200 or upconversion NPs. [201][202][203] Quantum dots are semiconductor nanocrystals with excellent fluorescence brightness that can also be used as optoacoustic contrast agents, 204,205 while carbon dots further represent a less toxic alternative with similar absorption characteristics. 206 Many NPs originally designed for non-optical imaging modalities may still have prominent light absorption properties, suggesting their use as multimodal contrast agents. Superparamagnetic iron oxide NPs are FDA-approved MRI contrast agents that were explored as standalone or combined agents in molecular optoacoustic imaging. [207][208][209][210] Microbubbles used in ultrasonography may additionally incorporate absorbing agents to change their optoacoustic contrast. [211][212][213][214][215] Porphysomes can further encapsulate fluorinated gases to be used as multimodal contrast agents for optoacoustics, fluorescence and ultrasound. 216 Similarly, encapsulation of gold NPs in liposomal or other types of nanostructures paves the way for devising contrast agents with multimodal and theranostic properties. 217 3.1.3. Interstitial delivery. Despite the general prevalence of the intravenous delivery methods, many probes are unable to reach the target when administered into the blood circulation, regardless of the level of chemical affinity. Depending on the investigated disease model or organ of interest, alternative administration routes may include subcutaneous, intratumoral or intracranial injections. The latter is important in brain imaging applications to label neurons or other cells that are otherwise unreachable with intravenous administration due to the high restrictiveness of the BBB. 218 The BBB can be overcome via an invasive procedure in which a small hole is drilled into the skull. By slowly inserting a small diameter injection needle or a pulled glass capillary (with tip-diameters of around 10-20 mm), contrast agents can be directly injected into the brain region of interest. 219 The main limitation of intracranial injections is the small volume that can be delivered (usually less than 1 ml in mice). Distribution of the probes into the brain tissue occurs only by diffusion. Thereby, the labelled or affected area may remain rather limited. 220 On the other hand, the injected probes have a very slow clearance rate, which enables longer effective imaging windows. Other less invasive methodologies have been reported for the purpose of temporary BBB disruption, e.g. burst-mode focused ultrasound. 221 Interstitial delivery has also been used in the field of cancer research, where the incompetent tumor vasculature often impedes efficient intravenous delivery of therapeutic and imaging agents. Intratumoral injection can be done by delivering the probe directly into the tumor microenvironment e.g. for targeted imaging or treatment monitoring purposes. 222,223 In addition, subcutaneous or intradermal injections of contrast agents can be of interest for imaging lymphatic vessels and nodes. For instance, the FDAapproved ICG dye has been shown to improve sentinel lymph node biopsy of axillary lymph nodes in a rat model. 224 A theranostic agent consisting of encapsulated conjugated oligomer nanoparticles was shown to accumulate in the sentinel lymph node after intradermal injection in the forepaw pad of a mouse, allowing for an accurate delineation of the lymph node margins. 225 Contrast-enhanced optoacoustic imaging of lymphatic vessels has also been performed after injection of Evans blue in the mouse tail. 226 MSOT in combination with a subcutaneous injection of ICG has proven to be valuable for the diagnosis of the metastatic state of sentinel lymph nodes in human melanomas, 106 representing a promising clinical application of the method.
3.1.4. Genetic reporters. Genetically-encoded expression provides arguably the most versatile approach for specific labeling of cells within a living organism. 3 Reporter genes enable optoacoustic detection and quantification of gene expression by either coding for light absorbing proteins or for enzymes that convert clear substrates into chromophoric products. Since fluorescent substances undergo photothermal conversion in non-radiative relaxations, highly-optimized fluorescent proteins (FP) can further provide optoacoustic contrast (Fig. 10a). The initial feasibility to image FPs with optoacoustics has been demonstrated with small organisms such as Drosophila pupa or adult zebrafish expressing proteins from the family of the green fluorescent protein (GFP). 59 However, the peak absorption of the GFP-based FPs has been so far only shifted to a maximum wavelength of 611 nm, 227 hence their application in mammal organisms is still hampered by the high background haemoglobin absorption at wavelengths below 650 nm (Fig. 7). More recently, far-red-and NIR-shifted fluorescent proteins have been engineered to provide peak absorption above 680 nm. 228 In particular, the infra-red fluorescent proteins (iRFP) are derived from bacterial phytochromes. 229 In contrast to GFP, which only requires oxygen for chromophore maturation, phytochromederived proteins incorporate biliverdin as the chromophore. Optoacoustic detection of iRFP has been demonstrated in living mice 129,230 and further validated against fluorescent measurements. 19 Due to their limited optoacoustic generation efficiency, detection of iRFPs has only been reported at shallow depths of several millimetres. Specifically-designed screening platforms are being used to optimize the performance of existing FPs and chromoproteins, 231 facilitating the engineering of novel genetic reporters for optoacoustic imaging and sensing. 232 Enzymatic reporter genes have also been explored as a means for generating optical absorption contrast (Fig. 10b-d). This approach was first demonstrated by imaging expression of the lacZ gene, 233 which encodes b-galactosidase converting 5-Bromo-4-chloro-3-indolyl b-D-galactopyranoside (X-gal) into galactose and the stable insoluble blue product 5,5 0 -dibromo-4,4 0 -dichloro-indigo (Fig. 10b). This highly absorbing product was shown to generate a measurable optoacoustic signal at centimetre-scale depths within ex vivo biological tissues. 234 However, its in vivo use is limited due to toxicity associated with administration of the X-gal chromogenic substrate, which may cause skin irritation and inflammation in rodents. The most widely employed enzymatic reporter for generating optoacoustic contrast is tyrosinase (Tyr) (Fig. 10c), a key enzyme in melanogenesis. 235 Tyr alone is sufficient for producing melanin in non-melanogenic cells from tyrosine without necessitating an extrinsic agent. 236 The tyrosinase homologue melA that produces melanin in bacteria has also been suggested. 237 Indeed, melanin is a dark and strongly absorbing pigment easily detectable with optoacoustic imaging systems. Melanin can further provide contrast in other modalities such as magnetic resonance imaging (MRI) 235,238 or positron emission tomography (PET) 239 and has also been suggested as a potential thermal therapy agent. 235 Yet, its relatively flat absorption spectrum makes it difficult to distinguish from blood via spectral unmixing. 240 In addition, although it has been possible to monitor the growth of Tyrexpressing tumours over periods of several weeks, 241 extensive melanin production is generally associated with increased cytotoxicity. 242 A viable alternative is the tetracycline-regulated inducible system, which further enables versatile control of gene expression. 243,244 Another promising enzymatic product that can generate optoacoustic signals is the bacterial pigment violacein, 245 which results from the multi-enzyme-based oxidation pathway of L-tryptophan (Fig. 10d). The particular benefits of violacein are its good photobleaching resistance and the highly distinct absorption spectrum extending above 650 nm. 245 In general, enzymatic reporter genes benefit from inherent signal amplification as each genetically-expressed enzyme can generate multiple products. On the other hand, quantification of gene expression may be hindered by dependence on the availability of the substrate.

Sensing mechanisms
Optoacoustic contrast agents can be designed to perform specific dynamic sensing tasks e.g. by altering their absorption properties in response to variations in physiological environment or due to binding to a certain molecular target. In fact, haemoglobin is an endogenously present optoacoustic sensor whose absorption spectrum is altered due to the oxygen binding, The heme group of the molecule undergoes conformational changes when oxygen binds to iron (Fig. 11a). Haemoglobin is thus frequently used as a powerful label-free optoacoustic sensing mechanism in brain imaging or cancer research.
Optoacoustic sensing of tissue hypoxia has been reported using methylene blue, whose lifetime is highly sensitive to the oxygen partial pressure. 142 The acidic pH is another important microenvironmental parameter that increases the potential of tumour migration and invasion, hampering efficacy of many drugs. 246 Optoacoustic sensing of pH was first suggested with the fluorescent pH indicator dye seminaphthrorhodafluors-5F. 247,248 Recently, other optoacoustic pH sensors e.g. based on micelles, 249 liposomes, 250 self-assembled nanoprobes, 251 semiconducting oligomer nanoparticles, 252 dextran-based nanoprobes 253 or gold nanoparticles 254 have been developed.
Reporter genes can also be turned into powerful environmental sensors to deliver information on the presence and concentration of certain compounds in real time. GCaMP proteins are genetically encoded calcium indicators (GECIs) that undergo conformational changes when calcium is present (Fig. 11b). GCaMP was shown to exhibit optoacoustically measurable variations in its extinction coefficient due to the influx of calcium ions into neurons. 255,358 The calcium-sensitive dye Arsenazo III has also been used to detect calcium concentration changes in other types of cells. 256 The temperature dependency of the Grüneisen parameter affects the optoacoustic generation efficiency. Hence, any endogenous or exogenous chromophore can potentially serve as a thermal sensor. Measuring temperature variations is relevant for the understanding of many biological processes. For example, cellular division, enzymatic reactions or gene expression are typically associated to intracellular temperature changes that can be monitored with optoacoustic microscopy. 257,258 Other types of optoacoustic sensing approaches have been reported. Protease-mediated fluorescent dyes are smart labels that can sense disease-related processes in cancer, rheumathoid arthritis or cardiovascular diseases. The commercially available MMPSense 680 probe, which is activated by metalloproteinases (MMPs), has shown promise in molecular optoacoustic imaging. 259 Smart probes based on activatable peptides attached to NIR chromophores can be tailored to target specific proteases and have shown to provide optoacoustic molecular sensing capacity. 260,261 Reactive oxygen species can also be detected with semiconducting polymer nanoparticles. 145 Other examples are sensors of hyaluronidase 262 or copper(II). 263

Contrast modulation approaches
Contrast modulation approaches are based on contrast agents whose optoacoustic signal generation is susceptible to external manipulation of the sample. Generally, it is preferable that the induced signal variations are reversible, although irreversible changes can still be used. Dynamic changes in the optoacoustic contrast can be exploited in order to enhance detection sensitivity of specific probes by removing contributions from timeinvariant background absorbers. A number of approaches for optoacoustic contrast modulation exist. As mentioned above, optoacoustic signals are temperature dependent. Thus, the signal amplitude can be modulated by changing the temperature in the sample. This effect has been exploited for non-invasive temperature monitoring in clinical procedures involving radiofrequency ablation, 264 high-intensity focused ultrasound (HIFU), 265 cryoablation 266 or laser-induced thermotherapy (LITT). 267 At the microscopic level, temperature changes in individual cells can be measured optoacoustically with an accuracy of 0.2 1C. 257 In principle, no exogenous agents are required for optoacoustic temperature mapping, although specifically-engineered probes with enhanced temperature sensitivity are also available, e.g. silica-coated gold nanorods. 268 Moreover, the absorption spectrum of the ordered Bchl-lipid dye aggregates can be reversibly switched by exceeding a certain temperature threshold. 269 Optoacoustic contrast can also be modulated with light. It was previously shown that pulsed light radiation may cause photobleaching of fluorescent probes and other photodegradation mechanisms for energy levels below safety standards. 270 Even though these effects are generally considered detrimental, they can be exploited for enhancing the resolution and contrast in optoacoustic microscopy by considering the signal difference before and after photodegradation. 34 Optoacoustic signal recovery after bleaching can be measured on a slower time scale, providing a better understanding of cellular dynamics. 271 Reversibly photoswitchable FPs are yet another versatile contrast modulation tool that involves no permanent photodegradation. The fluorescent protein Dronpa can be switched with light at different wavelengths between a cis conformation and a trans conformation having different absorption spectra (Fig. 11c). Detection sensitivity can be greatly enhanced with this kind of probes by identifying the changes in the optoacoustic signal strength for the activated and deactivated states. 130,272 This can be done by inducing time-varying optoacoustic responses, which can be temporally-unmixed from the time-invariant (constant) background absorption. 130 Furthermore, the optoacoustic signal decay rate is generally proportional to the local light fluence, a property that can be explored for estimating the light fluence distribution with photoswitchable agents. 273 Nonfluorescent particles have also been suggested as photoswitchable probes. 274 Optoacoustic contrast can be further modulated with a combination of pump and probe laser pulses. This approach was first suggested to capture transient optical absorption during the triplet state microsecond-level lifetime of methylene blue. 275 By using shorter delays between the pump and probe beams, the optoacoustic signal generated by fluorophores can be modulated due to stimulated emission. 276 Another promising source of energy for optoacoustic signal modulation is ultrasound. In particular, microbubbles and nanodroplets can be manipulated with ultrasound fields, leading to optoacoustic signal changes. It has been shown that the optoacoustic signal generated by microbubbles in a methylene blue solution is lower than that generated by the same concentration of bubble-free dye. Signal enhancement is thus achieved by bursting the microbubbles. 214 In addition, the signal generated by plasmonic particles conjugated to microbubbles is enhanced with respect to the unconjugated mixture or free nanoparticles. 215,277 Reversible phase transitions can be also induced in a nanoemulsion of perfluorohexane droplets coated with gold nanospheres by combining light and ultrasound pulses, 278,279 which leads to an enhanced signal as compared to laser-only excitation.

Photostability and toxicity
When selecting appropriate contrast agents for optoacoustic imaging, effects related to their photostability under pulsed nanosecond radiation, irreversible bleaching as well as in vivo toxicity have to be carefully considered. The light pulses employed in optoacoustic imaging induce photobleaching effects that considerably differ from those known from fluorescence imaging using single-or multi-photon microscopes. 270,280 It has also been long recognized that pulsed laser radiation has different biological effects from those observed under continuous wave exposure. 281 For example, effects related to photochemical destruction and melting under nanosecond laser radiation have been reported for inorganic NPs. 282 Shape alterations of gold NPs induced by prolonged nanosecond light exposure may lead to optoacoustic signal degradation. 283 In this regard, SWNT have significantly higher photostability than gold NPs 12 and SPNs can even outperform SWNT in long-lasting optoacoustic signal generation 145 (Fig. 12a). Many FPs are known for their lack of photostability, which can be seen in effects like dark states, blinking, transient absorption and photobleaching. 284 Furthermore, FPs are known to induce phototoxicity under certain excitation conditions. 285 While most of these effects are yet unexplored for realistic in vivo imaging conditions, optoacoustic signal degradation associated to photobleaching effects in FPs has already been observed under different imaging settings 270,280 (Fig. 12b). Photobleaching may even take place for safe excitation energy levels 270 (Fig. 12c). Of particular interest are non-fluorescent chromoproteins or enzymatically-amplified chromophores, which in comparison to FPs exhibit a higher optoacoustic efficiency due to the absence of radiative relaxation and ground state depopulation, and also higher photostability. 245,280 The use of exogenous contrast agents, especially inorganic NPs, is also associated with a number of biocompatibility, cytotoxicity and long term accumulation issues. The relatively large size of most NPs prevents efficient renal clearance. Instead, they are generally taken up by the reticuloendothelial system (RES) and thus may accumulate in the body for long periods of time. 138 Many types of particles are also non-biodegradable with their potential toxicity effects remaining unknown. For example, severe long term toxicity concerns exist for SWNT, 286 although it has been shown that they may be cleared through the kidneys due to their high aspect ratio. 140 On the other hand, a number of studies further support the low toxicity of carbon nanodiamonds. 287,288 Surface coating e.g. with a silica layer is known to enhance photostability and heat dissipation and hence optoacoustic signal amplitudes. 20,289 Furthermore, evidence exists that gold NPs accumulate in the RES and thus their long term toxicity remains a potential concern. 290 An exception to this are biodegradable nanoclusters encapsulating small (B5 nm) gold NPs, 291 which can be cleared through the kidneys.

Applications enabled by optoacoustic visualization of multiscale dynamics
Recent advances in optoacoustic imaging and sensing methods and in associated approaches have been accompanied by the advent of biomedical applications that benefit from the advantages of state-of-the-art systems, particularly for imaging dynamics. In this section, we survey novel applications involving imaging of multiscale dynamics.

Cell tracking
Tracking individual cells is important for understanding their longitudinal behaviour and response under different physiological and pathological conditions. For example, detecting and tracking the fate of circulating tumor cells in vivo is essential for early detection of metastatic spread, facilitating cancer treatment. Quantifying the vascular flow of blood cells is important for identification of microcirculation irregularities associated with cancer, diabetes or other conditions. Optoacoustic detection of circulating tumor cells labelled with magnetic nanoparticles was demonstrated using an external magnetic field. 294 Magnetic trapping was shown feasible for flow velocities up to 5 cm s À1 , so that it may represent a promising approach for microsurgical extraction or laser ablation of these cells. Cell trapping has also been achieved with gradient acoustic forces induced by ultrasound and optoacoustic waves 292 (Fig. 13a). The motion of a population of blood cells can be characterized with optoacoustic Doppler velocimetry by analysing the correlation of single optoacoustic signals generated by two laser pulses with a millisecond-level delay 293 (Fig. 13b). A relatively low concentration of cells is essential for accurate estimations with this method, although velocity measurements in whole blood have been achieved under certain conditions. 295 Individual optoacoustic signals may further contain information on cell morphology. Irregular shapes of red blood cells were identified by analysing the frequency content of individual optoacoustic signals. 296 Potentially, the same methodology can be applied for detection of other types of morphological abnormalities in red blood cells occurring e.g. in malaria or sickle cell disease.
Three-dimensional optoacoustic imaging of single cells in motion is challenged by the harsh requirements for superb sensitivity, spatial and temporal resolution. High-speed 2D imaging with cellular resolution is possible with OR-PAM. Specifically, B-mode optoacoustic images could be obtained at a rate of 100 frames per second, which is sufficient to track the flow of red blood cells in capillaries. 40 This method has been used to analyse the microcirculation in cuticle capillaries. By using dualwavelength excitation, it was possible to detect changes in oxygen saturation in flowing individual red blood cells 41 (Fig. 13c), which can shed light on the process of oxygen delivery to tissues in microvascular structures. Volumetric tracking of flowing particles in real time has also been achieved with four dimensional optoacoustic tomography, 297 which holds potential for in vivo applicability with properly labeled cells.

Perfusion and organ function
Organ support and maintenance is a complex and intricate process which crucially relies upon their blood supply. Blood constitutes an excellent endogenous contrast for optoacoustics due to its photostability and well-characterized dependence of its absorption spectrum on oxygen saturation. Thus, optoacoustics is clearly a highly suitable imaging technique for the characterization of organ perfusion, providing high spatial and temporal resolution for interrogation of vascular structures at different scales.
Dynamic label-free optoacoustic imaging of blood perfusion in subdermal vessels has been carried out completely noninvasively. In particular, the ability to image perfusion of blood flow in subdermal capillaries across a wide field of view and down to the resolution of single cells has been demonstrated 24 (Fig. 14a). At the macroscopic level, restriction of blood supply in ischemic lesions was characterized along with the consequent changes in blood volume and oxygen saturation. [298][299][300] Combined with other imaging modalities, optoacoustics can further serve as a complementary tool to characterize blood flow. 301 Optoacoustic imaging of organ perfusion can be greatly enhanced by using contrast agents. For instance, perfusion of  Evans blue was also used to characterize the functional state of capillaries early after the first developments in OR-PAM. 143 Video-rate visualization of kidney perfusion by intravenous injection of ICG was among the first applications demonstrated with cross-sectional real-time MSOT systems. 85 By using multiple light wavelengths spanning between 750 and 900 nm, the ICG distribution could be efficiently unmixed from the background blood signal in the entire kidney region and over time. In a follow-up study, the capacity to image blood vessels, the kidneys, the liver and the gall bladder with 150 mm spatial resolution through entire mouse cross-sections was demonstrated. 302 The removal of ICG from the systemic circulation and its timeresolved uptake in the liver and gallbladder was characterized (Fig. 14b). Furthermore, it was possible to track the uptake of a carboxylate dye in separate regions of the kidneys. Alternatively, negative contrast can also be used to visualize organ perfusion. A drop in the optoacoustic signal is generated by injecting a saline solution into organs, and the ensuing blood flow can then be readily visualised with the signal rapidly returning to its original level. 303 Arguably, the most complex challenge tackled in this area is imaging perfusion in the healthy or diseased heart in animal models, since this is a relatively large organ that moves at a high speed. Tackling this challenge using other whole-body tomographic imaging modalities necessitates gating the acquisitions but state-of-the-art high-frame-rate 4D optoacoustic imaging now enables imaging the heart in 3D on a beat-by-beat basis. 100 In this way, it was possible to visualize the flow of intravenously injected ICG from the right ventricle to the lungs and back into the left ventricle (Fig. 14c). Importantly, this image sequence provides an accurate measurement of the pulmonary transit time (PTT), a parameter that can serve as a functional indicator of ventricular dysfunction and which cannot be measured by other imaging methods relying on cardiac gating.

Neuroimaging
The massive prospects and opportunities for imaging the central nervous system are only very partially addressed by existing neuroimaging methods despite decades of intense research and important advances. These methods either offer a very low spatio-temporal resolution or an extremely limited field of view and depth penetration. Current neuroimaging techniques suffer from additional major pitfalls: most rely on observing neural activity indirectly using effects mediated through complex neurovascular coupling processes that we are only beginning to truly understand, and they typically do not simultaneously assess all the relevant cerebral hemodynamic parameters. The vast advances in the development and application of optoacoustic imaging techniques now offer a viable alternative for dynamic high-resolution observations in the brain. The endogenous optoacoustic contrast based on spectrallydistinct haemoglobin measurements provides a remarkable performance for label-free visualization of cerebral hemodynamics under different stimuli. OR-PAM enables assessing all the relevant cerebral hemodynamic parameters like oxygenated (HbO) and deoxygenated (HbR) haemoglobin, blood oxygenation (SO 2 ), total haemoglobin (HbT), cerebral blood flow (CBF) and, derived from these parameters, the oxygen extraction function (OEF) and the cerebral metabolic rate of oxygen (CMRO 2 ). 304 The achievable spatial resolutions reach the capillary level and state-of-the-art systems offer temporal resolutions in the range of 100 kHz for 1D scans. The penetration depth is however limited to superficial cortical layers. 22 In one of its earliest embodiments, functional optoacoustic microscopy was able to provide an independent verification of the BOLD-fMRI 'initial dip' (Fig. 15a), demonstrating a decrease in blood oxygenation due to an increase in arteriole HbR in the first few 100s of milliseconds after an electrical paw stimulus. 305 While achieving adequate imaging performance with highresolution OR-PAM normally necessitates removal of the scalp, tomographic systems, such as those based on ring-shaped transducer arrays, are generally fully non-invasive and have been utilized to image resting state 121 (Fig. 15b) and stimulus-evoked 306 brain activity. A 512-element full-ring ultrasonic transducer array with an in-plane resolution of 100 mm was used to image glucose metabolism in the mouse cortex after intravenous injection of a glucose-analogue molecule providing optoacoustic contrast. 306 A lower resolution 64-element ring transducer array was used to image brain dynamics in awake, freely moving rats (Fig. 15c). 307 By directly mounting the array onto their head (with intact scalp and skull) longitudinal tracking of hemodynamic changes in a generalized epilepsy model was performed. Alternative real-time cross-sectional optoacoustic imaging systems were configured to acquire coronal sections of the brain of a mouse, which allows visualizing deep-seated structures. 308 As discussed above, the development of spherically-shaped transducer arrays with densely-packed elements has taken the tomographic concept a step further, allowing real-time recording of true 3D data. A 256-element detection array system, which can record data at rates up to 100 frames per s 119 and is capable of tuning the laser wavelength on a per-pulse basis, provides multispectral information for analysing hemodynamic changes deep inside the mouse brain. This has recently been demonstrated in a model of focal epilepsy in which the onset, spread and termination of 4-aminopyridine-induced epileptic seizures were visualized, with concurrent electroencephalographic (EEG) recording of the aberrant neural activity 101 (Fig. 15d).
Notwithstanding this important progress in advanced hemodynamic imaging, modern neuroimaging work has largely steered towards the development and application of indicators that allow a much more direct monitoring of neuronal activity, most notably calcium-and voltage-sensitive indicators. Genetically encoded calcium indicators (GECIs), and primarily the GCaMP family of indicators have become the workhorse tool for visualizing distributed activity across neuronal populations. In a very recent study, functional optoacoustic neuro-tomography (FONT) has successfully demonstrated imaging of calcium-sensitive protein GCaMP5G expressed in a transgenic zebrafish model. 255 The study established a very strong correlation between the fluorescence and optoacoustic activity-dependent signals of the indicator. It was demonstrated in a generalized seizure-model that calcium neuronal dynamics recorded with optoacoustics could be analysed deep inside the scattering adult zebrafish brain. Conversely, these dynamics were not observable with concurrent fluorescence imaging, which rendered a blurry appearance lacking the depth information (Fig. 16). The findings were recently confirmed in an adult zebrafish model in vivo. 358

Fast imaging based on photoswitchable probes
Several fluorescent proteins and organic dyes are known to change their photo-physical properties in response to optical excitation. For example, photoactivatable proteins can be activated from a non-fluorescent state and become fluorescent, while photoswitchable proteins can exhibit a change in their fluorescence emission and reversibly switchable proteins can be switched reversibly between a fluorescent and a non-fluorescent state. Photoswitchable probes play an essential role in super-resolution optical microscopy techniques such as photoactivation localization microscopy (PALM) or reversible saturable optical linear fluorescence transitions imaging (RESOLFT).
Detection of optoacoustic signal changes associated to activation and deactivation of photoswitchable probes was shown with the reversibly switchable FP Dronpa and its fast-switching fatigue resistant variant Dronpa-M159T. 130 By using a systematic sequence of laser pulses at 420 and 488 nm, it was possible to dynamically modify the amount of activated molecules and hence generate specific temporal profiles in the corresponding sequence of images acquired with 4D optoacoustic tomography (Fig. 17a). These temporal profiles can be used to isolate (unmix) the signal generated by the probe so that it can be accurately mapped. The advantage of temporal unmixing in comparison to the more commonly-used spectral unmixing is that it can be inferred from a sequence of single-wavelength images, i.e., no spectral coloring effects are produced. Thereby, it is easy to distinguish the signals generated by the probe from the baseline signals (constant temporal background). Interestingly, the number of activated molecules decays exponentially as a function of time with the decay constant proportional to the local light fluence. 273 The measured decay rates thus enable an accurate estimation of the light fluence distribution, facilitating extraction of quantitative information on the chromophore concentration (Fig. 17b). Phytochromes exhibiting isomeric states have further been used as optoacoustic photoswitchable probes for NIR wavelengths. 272,310 For example, the bacterial phytochrome BphP1 can be switched between two different states by light excitation at 630 and 780 nm. By calculating the difference between the images obtained for the two states, it was possible to detect BphP1-expressing U87 tumor cells at a depth of 8 mm with high sensitivity (Fig. 17c). Additionally, the reversible switching mechanism of BphP1 can be exploited for contrast enhancement in OR-PAM and to overcome the optical diffraction limit for super-resolution optoacoustic microscopy. 272

Molecular sensing
Optoacoustic imaging sensors are physical, chemical or biologically active probes that detect an environmental change and convert it into a signal or form of information which can be detected optoacoustically. This new layer of information can be readily coregistered with anatomical images to provide an accurate non-invasive mapping of distribution of molecular and functional markers in vivo.
Optoacoustic sensors have been used to probe tumour microenvironments and monitor treatments. A temperature sensitive nanoswitch probe was synthetized by intercalation of lightharvesting porphyrins within thermoresponsive nanovesicles. 269 The probe functions by absorbing light at two distinct wavelengths (680 and 824 nm) while its spectral features can be reversibly switched by exceeding a temperature threshold. Using this thermochromic property, localized temperature changes within tumour xenografts were determined in vivo and noninvasively (Fig. 18a).
Similar strategies may be applied to detect other stimuli such as pH and enzymatic activity. Due to their high metabolic rate, malignant tumours are commonly characterised by lower than normal pH levels. A theranostic nanovehicle for targeting pancreatic cancer was designed based on mesoporous silica nanoparticles (MSN) that encapsulate indocyanine green for enhancing optoacoustic contrast. 311 The tumour specificity was improved with the addition of both chitosan, targeting acidic pH, and urokinase plasminogen activator (UPA), targeting UPAR. Accordingly, the signal increase due to acidic pH conditions resulted in a 20-fold stronger optoacoustic response. In vivo, MSN-UPA particles demonstrated orthotopic pancreatic tumour specific accumulation compared to liver or kidney, as identified  View Article Online using real-time MSOT. By tracking in vivo nanoparticle biodistribution with MSOT, it was further confirmed that pH responsive, ligand targeted MSNs preferentially bind to pancreatic tumours for payload delivery (Fig. 18b). Optoacoustic imaging has also been found capable of imaging oedema, whose different variants include cerebral, pulmonary, macular or lymphatic fluid build-ups within the body. One suitable probe for this is a nanoparticle based on semiconducting polymer particles (SPN1). 145 It exhibits preferential narrowband absorption in the NIR and is resilient against photodegradation and oxidation. Most important is its ability to ratiometrically detect reactive oxygen species (ROS) at 700 and 820 nm (Fig. 18c). In fact, the increase in ROS has also been shown as a marker of apoptosis. In the same study, using a mouse model of acute oedema and zymosan to simulate ROS generation, SNP1 was successfully used to monitor the ROS regulation process intra-muscularly.
Another promising application of dynamic optoacoustic molecular sensing is the monitoring of thyroid-related disorders. The thyroid gland regulates many of the most important functions of the body, from metabolism to breathing. Thyroid cancer is rare but benign nodules on the gland are common. Follicular thyroid carcinomas are very similar to benign nodules yet differ in terms of environmental conditions. One noticeable difference  is the activity and presence of the matrix metalloproteinases (MMPs) and two members of this family (MMP-2 & MMP-9) were determined as biomarkers of malignant thyroid lesions. An MMP-activatable optoacoustic probe based on Alexa-Fluor 750 was successfully used to image FTC133 thyroid tumours subcutaneously implanted in nude mice in vivo. 261 Both MMP-2 & MMP-9 cleave the protein in a non-reversible fashion, which increases its optoacoustic and fluorescent signal.

Pharmacokinetic and bio-distribution analysis
The characterization of pharmacokinetic and biodistribution profiles is an essential step in the development process of new candidate drugs or imaging agents. In addition, targeted modifications of diagnostic and therapeutic probes can further facilitate selective probe delivery. The bio-distribution analysis is of particular importance in cancer research as angiogenesis and changes in vascular morphology significantly alter the probe uptake rates. The blood perfusion profiles are responsible for the establishment of multiple parameters within the tumour mass, such as hypoxic and nutrition gradients, cell viability, proliferation and drug response potentials. Tumour neovascularization typically exhibits leaky behaviour, which can be exploited for specific probe delivery. 4D and 5D optoacoustic tomography was reported to attain concurrent video-rate observations from entire tumour volumes. 312 Assisted with liposome particles encapsulating ICG (Lipo-ICG), simultaneous assessment of blood oxygenation gradients and vascularization in solid breast tumours was demonstrated further revealing different types of blood perfusion profiles in vivo. The method greatly benefited from its ability for high resolution in vivo assessment of 3D maps of blood oxygenation in and around the tumours. In this way, vascular tracts of various sizes as well as less oxygenated areas within tumours could be efficiently resolved. Furthermore, analysis of the perfusion kinetics of Lipo-ICG has allowed for an accurate classification of three different perfusion profiles in a subcutaneous tumour model of angiogenesis ( Fig. 19a and b).
Solid nanoparticles are known to provide excellent optoacoustic contrast but can potentially induce toxicity. One way to improve their biocompatibility consists in wrapping a toxic compound into a biocompatible and non-toxic compound, such as poly(D,L-lactide-co-glycolic acid) (PLGA), a biodegradable polymer commonly utilized for drug delivery. This however tends to change the kinetics and bio-distribution of the probe, so that whole-body small animal optoacoustic scanners are of great value for studying such novel constructs. A recent study explored the use of PLGA nanoparticles loaded with the 1,1 0 -dioctadecyltetramethyl indotricarbocyanine iodide (DiR) NIR-dye for optoacoustic contrast enhancement 313 (Fig. 19c). Differential kinetics and bio-distribution between an intravenous and intraperitoneal injection routes were studied. While in both cases the probe was taken up by the liver and spleen, a significantly higher amount was taken up into the spleen after intraperitoneal injection.
Surface modifications of nanoparticles may also assist in enhancing the optoacoustic contrast. For instance, in the case of gold nanorods, cellular uptake into the cells via endocytosis and high concentration in endosomes may lead to diminished optoacoustic signal intensities. However, a steric hindrance provided by silica coating of the nanorods leads to improved optoacoustic sensitivity. 314 In this study, the enhanced sensitivity allowed for monitoring labelled stem cells implanted into mice over 15 days with high spatial resolution, albeit dilutions of the nanorods in the cell cluster are produced with each cell division. The size of nanoparticles is an equally important parameter for determining the attained optoacoustic signal intensity. A particle size-dependent contrast enhancement effect was shown for copper sulfide nanoparticles with sizes between 3 and 7 nm. 315 While the in vitro contrast enhancement effect was proportional to the particle size, the latter has equally affected pharmacokinetics in vivo. After an intravenous injection, copper sulfide nanoparticles smaller than 5 nm presented higher contrast enhancement, especially at the tumour boundaries.
In a different application, the thermoelastic expansion associated with the optoacoustic effect was used to destroy intracellular vesicles. 316 The idea behind this approach is to prevent unwanted intracellular vesicular compartmentalization of anticancer drugs in cancer cells, which may lead to chemoresistance of tumours. To achieve the disruption of vesicles, a folate-modified SWNT was first trapped inside acidic vesicles after entering lung cancer cells. Irradiation with NIR pulsed laser then led to optoacoustically-induced shockwaves that broke the vesicles, resulting in the release of anticancer drugs. A 12-fold enhancement of antitumor therapeutic efficacy in tumour-bearing mice was achieved.

Treatment monitoring
Optoacoustic imaging based on endogenous or exogenous contrast can be used in a great variety of treatment monitoring applications. For instance, using endogenous tissue contrast and the temperature-dependence of the optoacoustic signal, it was possible to dynamically monitor tissue ablation procedures. In endovenous laser therapy (ELT), elimination of incompetent truncal veins is achieved via intravenous laser coagulation leading to cell destruction and collagen shrinkage, which was successfully monitored by 4D optoacoustics 267 (Fig. 20a). In this way, it was also possible to accurately follow the pull-back of the ablating laser catheter. Similarly, optoacoustics has also been used to monitor HIFU, 265 and radiofrequency ablation, 264 where multispectral imaging has also been used to identify spectral changes associated with tissue coagulation. Temperature monitoring during the cryoablation of a canine model of prostate cancer has been performed with an accuracy of 2 1C for a temperature range from 35 to À15 1C. 266 Endogenous haemoglobin optoacoustic contrast has also been used to follow up anti-angiogenic tumour therapy using acoustic-resolution optoacoustic microscopy. Here RNA silencing (siRNA) therapy was used to target tumour vasculature, whereas MSOT monitoring of the tumour response to anti-angiogenesis was done in a breast cancer mouse model 317 (Fig. 20b). The noninvasive imaging approach provided longitudinal monitoring of a successful therapy, seen e.g. by tumour shrinkage or reduced vessel density (over days to weeks).
Optoacoustics has also been used for monitoring brachytherapy in prostate cancer, which consists of implanting small sealed radiation sources, so-called seeds, inside or near the tumours. In this way, the radiation dose to cancerous tissue is maximized while sparing healthy tissue, yet factors such as seed migration, prostate motion, oedema, or implantation errors may lead to complications. In an in vivo canine study, seed locations were mapped with optoacoustic imaging using a transrectal ultrasound probe 318 (Fig. 20c). In this study, a new interstitial light delivery method and beamforming techniques were used to improve contrast from the metallic seeds. Another important diagnostic marker for cancer progression is the amount of cell death by tissue necrosis, which serves as a reliable indicator of rapid and aggressive tumour growth. In fact, some therapeutic approaches that are initially designed for inducing apoptosis may often result in the so-called secondary necrosis. An approach to assess the amount of necrosis during tumour progression or therapy is based on near infrared fluorescent carboxylated cyanine dyes with strong necrosis avidity. 319 By using noninvasive MSOT imaging in a mouse model, these dyes have been shown to accumulate in the deoxy-haemoglobin-rich tumour core (Fig. 20d). Ex vivo validation further confirmed the presence of tissue necrosis inside the tumours.

Cellular and sub-cellular function
Optical microscopy has become the workhorse of modern biological discovery. Despite its central role in nearly all biological and biochemical investigations, microscopic techniques have several significant drawbacks such as the need for labeling or the presence of unspecific endogenous autofluorescence. The fundamentally different intrinsic contrast explored in optoacoustic microscopy may thus turn advantageous for label-free observations. The high speed and subcellular resolution attained with state-of-the-art optoacoustic microscopy are thus poised to deliver new imaging capacities and biological insights at the cellular and sub-cellular levels.
Since optoacoustics is a truly spectroscopic imaging method, observations can be made by relying on numerous proteins and their subsidiary amino acids that absorb light at known wavelengths. For instance, DNA, cytochromes and mitochondria are known to have different absorption spectra in the 250 to 420 nm wavelength range. Optoacoustic microscopy studies of unstained fibroblasts were performed where the cytoplasm could be readily identified from the nucleus 320 (Fig. 21a).
Recent studies demonstrated 3D renderings of single blood cells from purely endogenous contrast 51 (Fig. 21b). The only comparable method with equivalent 3D imaging capacity at this level is atomic force microscopy, which is however much slower and requires arduous sample and instrument preparation. In yet another study based on a two-photon absorption mechanism in vivo label-free optoacoustic microscopy was used to observe non-fluorescent melanin distribution within mouse skin with high frame rates and sub-femtoliter 3D resolution 37 (Fig. 21c).
Finally, super resolution methods have undoubtedly been the largest contributor to the recent progress in light microscopy, and have also been broadly adopted in optoacoustic microscopy. Whilst fluorescence-based super resolution requires the use of stochastically emitting fluorescent proteins or high powered laser illumination for stimulation depletion (increased phototoxicity), optoacoustic microscopy can take advantage of endogenous contrast to produce comparable results. A third-order label-free optoacoustic nanoscopy technique was shown capable of visualizing mitochondria in fibroblasts down to a lateral resolution of 88.0 AE 12.5 nm 32 (Fig. 21d).

Developmental biology
Multiscale imaging is essential in developmental biology to understand the coordination of the multiple growth and development mechanisms at the molecular, cellular, organ and wholeorganism levels. Developmental biology and experimental genetics often rely on insects or vertebrate models such as Drosophila melanogaster or zebrafish due to their low breeding complexity, well documented genetics and good representation of several common human diseases. Zebrafish embryos have been imaged with OR-PAM. 25 A hybrid focus optoacoustic microscope operating both in the ballistic and diffuse regime of light was used for volumetric tracking of migratory melanophores in zebrafish at different developmental stages 321 (Fig. 22a). Three-dimensional visualization of melanophores is challenging using optical microscopy due to the lack of fluorescence contrast, yet mechanisms of specification and regeneration of melanophores are important for the understanding of melanomas and other related human disorders. Tomographic optoacoustic imaging has also been used for studying embryo development in mouse models 322 (Fig. 22b). Here the strong blood contrast has enabled accurate  Although several organisms can be efficiently visualized by optical microscopy due to their transparency at the larval and embryonic developmental stages, light scattering considerably increases as the organism grows and expands, making those organisms inaccessible by optical microscopy at the juvenile or adult stages. In response, hybrid optoacoustic and optical imaging methods were developed in order to expand the time window of developmental observations. 323 Examples of the recently showcased combinations include optoacoustic and confocal microscopy 25 as well as optoacoustic tomography and selective plane illumination microscopy. 324

Longitudinal dynamics and disease progression
Disease progression can be monitored with endogenous or exogenous optoacoustic contrast. Genetically encoded optoacoustic contrast mechanisms can help to visualize otherwise invisible non-vascularized tissues. For example, the tyrosinasebased genetic reporter system was used to obtain 3D optoacoustic images of xenografts in mice at 10 mm depths 241 (Fig. 23a). For this, melanin-producing cells were inoculated in a mouse, and optoacoustic imaging was capable of following the angiogenesis and tumour expansion over several weeks in high detail. In another study, neovascularization of a developing implanted tumour could be visualized with high resolution and over time using OR-PAM. 325 Here repeated imaging of the same skin area without motion artefacts was achieved by using a surgically implanted skin clamp. Major features of tumour angiogenesis, like the change of vascular tortuosity, dilation of vessel diameters and increase of blood supply could clearly be analysed.
Optoacoustics is also particularly suited for monitoring the evolution of skin cancers. In particular, melanoma is the deadliest type of skin cancer, known for the early occurrence of sentinel lymph node (SLN) metastases. The chief motivation is the current lack of diagnostic means to detect metastases in transit from the primary tumour site to the SLNs. Micrometastases in the identified SLN were detected and tracked in an in vivo mouse model using handheld 5D optoacoustic tomography 91 (Fig. 23b), which was shown to be superior to PET, especially in detecting in-transit metastases. MSOT also helped in the differential diagnosis of melanoma lymph node metastases versus other lymphadenopathies. Apart from cancer, other diseases can potentially be longitudinally monitored with optoacoustics. For example, diabetes, chronic renal failure, cardiovascular disease and neurological disorders are all associated with the accumulation of advanced glycation end-products (AGEs). AGEs are predictors of diseaseassociated long-term complications, thus their early non-invasive detection is of paramount importance. Spectral optoacoustic measurements of AGE in a porcine skin model demonstrated promising prospects for the detection and quantification of AGE accumulation. 327 Optoacoustic monitoring of stroke progression in the whole brain of living mice in a model of middle cerebral artery occlusion (MCAO) was also suggested. 299 Hypoxia resulting from post-stroke ischemia could be identified in vivo with MSOT by assessing asymmetric dynamics of deoxygenated haemoglobin in the brain. The area surrounding the stroke was further identified non-invasively as hemodynamicallycompromised. OR-PAM was further used for imaging amyloid plaques in an Alzheimer's transgenic mouse model 326 (Fig. 23c). Amyloid-beta plaques inside the brain are hallmarks of Alzheimer's disease that seem to play a major role in its development. While this application needed invasive imaging via a cranial window and a plaque-specific dye staining, plaques could readily be visualized in the presence of highly absorbing brain vasculature.

Clinical translation
The great potential of optoacoustic imaging showcased in preclinical research has encouraged the translation of this technology into the clinics with multiple applications envisioned, from intraoperative diagnostics to ophthalmology, dermatology and endoscopic imaging. One of the key advantages of optoacoustic imaging is its intrinsic potential to deliver complete volumetric tomographic datasets from the imaged object with a single interrogating laser pulse, a possibility that does not exist in other clinical imaging modalities. This capacity also comes with important clinical advantages, such as the ability to dynamically visualize the bio-distribution of contrast agents in 3D and reduce out-of-plane and motion artefacts, thus facilitating clinical observations. Dedicated handheld optoacoustic probes have recently been introduced for high performance imaging of human subjects. Those come with real-time tomographic imaging capacities in 2D 93,97 and 3D 329 as well as integrated pulse-echo ultrasonography capabilities. 103,330 One promising application is the assessment of the metastatic status of sentinel lymph nodes in human melanoma. In the first-in-human study, cross-sectional and volumetric MSOT were used to image SLNs ex vivo and in vivo in patients with melanoma. 106 In comparison with the conventional protocols for analysis of excised SLNs, it was possible to significantly improve the tumour metastasis detection rate from 214 melanoma patients using MSOT. When combining noninvasive MSOT imaging with subcutaneous injection of the ICG contrast agent, visualization of the SLNs in vivo in 20 patients was further performed with up to 5 cm effective imaging depth (Fig. 24a). MSOT identified cancer-free SLNs in vivo and ex vivo without a single false negative (189 total lymph nodes), with 100% sensitivity and 48 to 62% specificity. The handheld optoacoustic imaging technology was also used in a feasibility study assessing peripheral blood supply and vascular disease in human feet. Such diagnosis is important in the context of peripheral arterial diseases, diabetic foot, and (autoimmune) vasculitis. 331 In this study, optoacoustic imaging has been shown to be superior to conventional imaging methods (e.g. duplex ultrasonography) in terms of resolution (capillaries as small as 100 mm in diameter were resolved) and its intrinsic spectroscopic capacity to differentiate between arteries and veins. Also, due to its imaging speed, optoacoustic imaging allowed to identify pulsation in arteries.
One potential valuable application of optoacoustics is the diagnosis of breast lesions. It is a favoured application due to the generally low light attenuation in the breast as compared with other tissues, 128 which allows penetrating the entire human breast. 74 Indeed, breast cancer diagnosis was aimed with the very first implementations of optoacoustic tomography scanners, 332,333 which was recently followed by several designs optimized for deeper imaging and better detection sensitivity. 334 Several clinical breast imaging studies are ongoing and the method has shown potential for non-invasive detection of malignant lesions. In a clinical study on infiltrating ductal carcinoma, distinct optoacoustic patterns could be classified as mass-like, non-mass like and ringshaped 328 (Fig. 24b). Ultimately, the comparison of these image features with those from contrast enhanced MRI and with vascular stained histopathology has led to the attribution of optoacoustic signal intensities to the presence of vascularity. However, optoacoustic patterns of malignancy still need to be compared in larger patient cohorts before definite optoacoustic image descriptors can be developed as diagnostic criteria. Optoacoustic imaging is inherently limited by the effective depth of light penetration into living tissues, hindering whole-body human imaging. Instead, embodiments similar to endoscopic or intravascular ultrasound are being developed. 335 Spectroscopic intravascular optoacoustic imaging has shown promise to detect and distinguish lipids in atherosclerotic human plaques ex vivo. 336 Initial feasibility of in vivo intravascular optoacoustic imaging using an integrated ultrasound and optoacoustic imaging catheter was also demonstrated in a rabbit model. 337 In addition, many types of cancers in e.g. lungs, colon, pancreas or prostate are located inside (or in close proximity to) the respiratory system or gastrointestinal tract, where endoscopic procedures are applicable. This has prompted the development of the simultaneous multi-spectral optoacoustic and ultrasonic dual-mode endoscopy, which has shown its ability to image the trachea and the oesophagus of a living rabbit, thus also illustrating its potential clinical application. 338 In the field of dermatology, new non-invasive skin imaging tools are essential to aid real-time diagnosis of skin tumours, chronic inflammation, alopecia, scarring, burns etc., thus minimizing the need for invasive skin biopsy. Optoacoustics offers the unique capacity for high resolution 3D optical mapping of tissue by further delivering highly specific optical contrast from a depth of several millimetres to centimetres in living tissues. A recent study performed on human volunteers, has shown capacity for non-invasive structural and functional analysis of intact hair follicles and pilosebaceous units by volumetric handheld MSOT. 339 On-the-fly assessment of key morphometric parameters of follicles and lipid content as well as functional oxygenation parameters of the associated capillary bed was demonstrated with high spatial resolution below 70 mm.
One major limitation for clinical translation remains the lack of clinically-approved probes for contrast-enhanced optoacoustic imaging. ICG has been approved for human use since the 1960s. However, there are currently no other approved compounds that absorb light in the near infrared, which is essential for deep tissue imaging with optoacoustics. In addition, though regulatoryapproved preclinical devices have already reached the market, medical device permissions in Europe, United States and other parts of the world are still pending. Once approved, multicenter trials in a number of key clinical applications will facilitate the transition of optoacoustics from a highly potent research platform to an accepted clinical imaging modality.

Summary and outlook
The physical effect underlying optoacoustic signal generation, the so-called photophonic phenomenon, was discovered in 1880, 340 but only at the turn of 21st century has optoacoustic technology reached the level of maturity that allows it to become a well-accepted biomedical imaging modality. Indeed, it has taken the development of suitable laser sources, sensitive and broadband ultrasound detection technology, as well as fast data acquisition and processing capacities in order to realize efficient and practical imaging devices. 341 The rich endogenous optoacoustic contrast of biological tissues enables non-invasive readings of key functional parameters, such as hemodynamic changes, haemoglobin concentration, oxygen saturation or metabolic rates. Intrinsic sensitivity to other chromophores, such as melanin and lipids, endows optoacoustics with label-free specificity to early markers of cancer or atherosclerosis. Much like other optical imaging modalities, optoacoustics is safe for both small animals and clinical use as it utilizes non-ionizing radiation at the visible and NIR wavelengths. Technical progress in this area is mainly driven by a large array of unmet biological and medical needs that can be addressed by the unique contrast mechanisms and imaging performance available with optoacoustic methods. Pre-clinical applications have rapidly developed and both experimental and commercial imaging scanners can already be found in many laboratories around the globe. These systems are being used in studies spanning the entire palette of modern biology including cancer, cardiovascular diseases, neuroimaging, ophthalmology, immunology, diabetes and obesity, cell trafficking, and a multitude of other biological functions. The multi-disciplinary nature of optoacoustics is also evinced by the growing contribution from chemistry and nanotechnology, where innovations have taken place in areas of dedicated biomarker design, from nanoparticles and organic dyes, to targeted agents and genetically encoded markers. 12 Dynamic imaging at a high frame rate is of particular importance in neuroscience. The breakthrough multiscale imaging capabilities of optoacoustics can be better appreciated when comparing its dynamic imaging performance with other neuroimaging modalities (Fig. 25).
Optical microscopy can provide diffraction-limited or even subdiffraction spatial resolution on a sub-micron scale. 5 However, the imaging rate rendered with state-of-the-art volumetric microscopy methods is in the 1 mm 3 s À1 range. [342][343][344] Moreover, Fig. 25 Comparison of dynamic imaging capabilities of the various functional modalities used in small animal research and the clinics. Shown are: optical methods (violet) based on two photon microscopy (2P), 342 light-sheet microscopy (LSM) 343 and light field microscopy (LFM); 344 small animal 345 and human 346 functional magnetic resonance imaging (fMRI -orange); highdensity diffuse optical tomography (HD-DOT -gray); 347 functional ultrasound (fUS -green); 349 optical-resolution photoacoustic microscopy (OR-PAM); 22 4D and 5D optoacoustic tomography (4D-5D OAT) 255 (dots indicate three reported systems with isotropic resolution). photon scattering remains the fundamental physical limitation for those methods, thus the imaged volume cannot be extended beyond several 100s of microns in the depth direction, further impeding acquisition of dynamic information from large tissue volumes. Deep tissue optical imaging can be alternatively done by means of diffuse optical tomography techniques, which however suffer from severely impaired spatial resolution that degrades to about 5-10 mm at centimetre-scale depths, 347,348 depending on the type of imaged tissue. On the opposite edge of the performance scale are functional magnetic resonance imaging (fMRI) 345,346 or functional ultrasound (fUS). 349 These non-optical imaging methods are excellent in visualizing large tissue volumes, including the entire human brain. Yet the contrast is mainly representative of tissue morphology, its mechanical properties or hemodynamics. Despite significant efforts in the past decade, synthesis of targeted contrast agents remains difficult for those modalities, severely limiting their applicability in molecular or functional imaging studies. In addition, the relatively low temporal resolution in 3D together with susceptibility to motion artifacts often requires the use of gating approaches in other applications involving fast organ motion, such as cardiac imaging. Even the most advanced pulsed echo and Doppler US methods are prone to artefacts related to rapid motion, e.g. murine heart beat at a rate of 400-600 beats per minute, and fast dynamic imaging performance is still limited in these cases to cross-sectional (2D) scans.
Conventionally, a multi-modality approach has been employed to acquire information at multiple scales by e.g. combining optical microscopy with whole-body tomographic imaging modalities. However, the fundamentally different contrast mechanisms, sensitivity, and other metrics associated with different modalities often hamper efficient combination of the information obtained at several spatial or temporal scales. 350 Optoacoustic imaging is unique in its ability to bridge the gap between the microscopic and macroscopic realms with the same type of contrast, from microscopic observations at the single capillary and cell level to whole body imaging of small animals and deep tissue imaging of humans. By adjusting the parameters of optical illumination and/or ultrasonic detection, the spatial resolution and imaging depths can be flexibly adapted to the particular application. As a rule of thumb, depth-to-resolution ratio of approximately 200 can be achieved for in vivo applications. 13 Imaging of dynamic biological events is continuously applicable all the way from video-rate volumetric acquisitions to longitudinal studies of disease progression and treatment monitoring lasting for days or weeks. 73 As anticipated, the temporal resolution generally scales inversely with the effective field of view (Fig. 25). Generally, the optimum trade-off between spatial resolution, temporal resolution and field of view depends on the particular application considered and the optoacoustic imaging system must be designed accordingly in terms of its different technical parameters -pulse repetition rate of the excitation laser, size of the illumination spot, number and bandwidth of the ultrasound detectors, signal range and sampling parameters of the digital acquisition system.
The way ahead is fascinating. Given the current progress in nanoparticle research and the versatility of optoacoustic contrast mechanisms, it is anticipated that future developments in light absorbing agents would bring the dynamic contrast enhancement approaches to a new level of performance. Novel smart agents that selectively change their optoacoustic contrast with environmental changes or theranostic agents releasing drugs at specific targets may all lead to paradigm shifts in biomedical research. Ideally, agents need to be optimized for more efficient optoacoustic signal generation, both in terms of the overall generated signal strength and spectral response. In this regard, development of novel dyes and genetic labels with preferential absorption in the far-red or NIR ranges for deep tissue optoacoustic imaging represents a highly promising research direction. To this end, the majority of new optoacoustic contrast agents have only been tested at a proof-of-principle level or early stages of exploration, thus further validations for biocompatibility, toxicity and targeting efficiency are necessary to establish in vivo applicability. In particular, biocompatibility issues may limit applicability if the maximum allowable administered dose is insufficient for generating a distinctive signal. Other circulatory and physiological barriers exist for large-sized agents with sizes of hundreds of nanometers. 12 Given the relatively early stage of most optoacoustic contrast agent developments, reproducibility of the results can be greatly facilitated if the developed agents are efficiently disseminated among the research laboratories, assisted with standardized methods of chemical synthesis that reduce the production costs.
From the technical point of view, much like other imaging modalities, optoacoustics comes with its own set of limitations that call for development of advanced solutions. Naturally, imaging depth is restricted due to light attenuation in optically opaque tissues, which also affects the minimal detectable concentrations of intrinsic tissue chromophores and extrinsicallyadministered contrast agents. Some of it can be compensated by increasing the deposited laser energy. However, for in vivo applications, especially those involving human subjects, illumination on the skin surface is limited by the laser exposure safety standards. 21 As a result, the effective imaging depth is usually restricted to regions where the light fluence is sufficiently high to generate detectable pressure variations, typically up to a few centimetres in most soft tissues. Imaging in deeper areas can be alternatively done using intravascular or endoscopic approaches. To this end, in vivo sensitivities in the submicromolar (for organic dyes and proteins) and picomolar (for solid nanoparticles) ranges were reported. 351 In reality, sensitivity limits are not only affected by molecular weight of the probe employed but also by multiple additional factors, such as the total volume, the spectrum and absorption coefficient of the imaged chromophore, the noise equivalent pressure (NEP) of the detectors as well as the level and spectral dependence of background tissue absorption.
The highly heterogeneous nature of biological tissues may further lead to the appearance of significant image artifacts and compromise imaging performance and quantification. Related problems are acoustic mismatches between different soft tissues 352,353 as well as additional effects related to the presence of strongly reflecting and scattering tissues, 354 such as bones and lungs. In multispectral imaging applications, one further faces the challenge of the so-called 'spectral coloring'. Due to the non-local and non-linear dependence of the light fluence distribution on the optical properties of the object, 355 spectra of various tissue chromophores and agents, extracted by means of optoacoustics, might be corrupted. 113 For improving quantitative determination of chromophore concentrations, the light distribution in tissue needs to be accurately accounted for. 356 All those represent challenging problems in optoacoustic imaging and tomography and therefore very active research areas.
Quantitative rendering of chromophore concentration is perhaps among the most important yet challenging tasks of optoacoustic methods. Haemoglobin in its oxygenated and deoxygenated form is the most dominant intrinsic absorber in mammalian tissues in both the visible and most of the NIR range of the optical spectrum. On the one hand, this allows for the label-free quantitative determination of hemodynamics and blood oxygenation, which are related to a variety of physiological parameters. However, when imaging other chromophores or extrinsically-administered contrast agents, the strong blood background may hamper quantified extraction of biomarker concentrations. This prompts the development of new approaches for efficient separation/unmixing of the haemoglobin-related optoacoustic signals in the spectral, temporal or spatial domains.