Photoacoustic Imaging of the Eye

2.1 PA imaging of the anterior segment vasculature of the eye

2.1.1 Normal iris vasculature

Anterior segment PA imaging of a normal eye receives signal from hemoglobin in the red blood cells of the iris microvasculature. Both AR-PAM and OR-PAM have been reported for iris vascular imaging. de la Zerda et al. [4] have used AR-PAM to qualitatively assess the rabbit eye in vivo. PA signals from the iris were detected, but the image of iris microvasculature was not clearly as the lateral resolution of the system was limited (around 200 μm). Hu et al. [13] have used a OR-PAM mechanical scanning system to get high resolution in vivo images of the iris microvascular in Swiss Webster albino mice (Figure 2a). The lateral resolution of the system was ~5 μm. The center frequency of the focused ultrasonic detector was 75 MHz. Hu et al. also have used two excitation wavelengths (570 and 578 nm) to estimate the concentrations of oxyhemoglobin (HbO₂) and deoxyhemoglobin (HbR), thereby have quantitatively measured sO₂ of the iris microvasculature. Dashed box in (Figure 2a) show the false-color map of the vasculature. With the limitation of mechanical scanning, the image acquisition took ~2 hours. Rao et al. [15] have used an OR-PAM hybrid-scanning system to image the major artery circle and the minor artery circle of the iris microvasculature in albino mice. Zhao et al. [14] have used a custom-built OR-PAM system to perform three-dimensional (3-D) vasculature imaging of rat iris in vivo with a lateral resolution of ~5 μm. They have extracted accurately vessel diameter, vascular density, and vascular tortuosity information (Figure 2b).

2.1.2 Corneal neovascularization

Normal corneal tissue is avascular and optically transparent. Corneal neovascularization is a pathologic condition that can lead to reduced vision characterized by the presence of blood vessels in the cornea. This pathophysiology has been observed in humans with inflammatory or autoimmune responses.

Liu et al. [16] have used a mechanical scanning OR-PAM system to image corneal neovascularization of alkali-burn-injured mouse eyes in vivo for the first time. The lateral resolution of the system is 2.76 μm, and the axial resolution is around 50 μm. OR-PAM can achieve a similar lateral resolution as confocal microscopy without a contrast agent. However, the imaging depth is limited (700 μm) and the depth of variation of corneal vessels can be more than 1.4 mm owing to the curvature of the mouse cornea. Jeon et al. [17] have demonstrated a new method to evaluate blood circulation in the eye by combining in vivo PAM imaging. By using a machine learning algorithm, random sample consensus algorithm (RANSAC), they could have visualize the PA anterior ocular vasculature image and demonstrate layer-by-layer analysis of corneal neovascularization on a mouse eye (Figure 3c–e). This OR-PAM system had a 3 μm lateral resolution. PAM provides the ability to delineate corneal neovascularization from iris vessels and enables visualization of individual capillaries with high-resolution (Figure 3b). By stimulating corneal angiogenesis with an alkali burn in Tie2-GFP fluorescent-reporter mice, Kelly-Goss et al. [18] have used the PAM system to analyze functional hemodynamic changes in microvessels over time during adult corneal angiogenesis. These functional measurements of the microvasculature supplement traditional metrics of vascular architecture, including blood flow velocity, hemoglobin content, and sO₂. The study is the first time to measure the functional information on corneal neovascularization.

Several of the above studies of the anterior segment vasculature images have utilized mechanical scanning OR-PAM system. However, mechanical scanning OR-PAM has limitations: the imaging speed is slow and careful coupling medium is required. Recently, a non-interferometric photoacoustic remote sensing microscopy system has been developed with a high scan rate and an improved signal-to-noise (SNR).

2.2 PA imaging of the posterior segment vasculature of the eye

2.2.1 Retinal vasculature

Posterior segment PAI of a normal eye images red blood cells in the retinal and choroidal microvasculature. To obtain the images, both AR-PAM and OR-PAM systems have been used, along with mechanical scanning and optical scanning systems.

Using an AR-PAM system, de la Zerda et al. [4] have used a pulsed laser (wavelength 740 nm) combined with a 25 MHz central frequency transducer to mechanically scan a living rabbit eye. They have obtained the blood distribution of the posterior eye. But owing to the system’s lateral resolution of 200 μm, limited features of the retinal or choroidal vasculature are visualized. Furthermore, it took 90 min to acquire the full 3-D image.

Jiao et al. [12] have developed an optical-scanning OR-PAM system for in vivo retinal imaging in 2010, which they have referred to as a photoacoustic ophthalmoscopy (PAOM). The pigmented rat fundus was imaging using the PAOM system (Figure 4a–c). In the PAOM system, the illumination source was a 532 nm pulse laser, the laser pulse energy was set lower than 40 nJ/pulse. The laser induced PA waves from the retina were detected by a custom-built needle ultrasonic transducer (30 MHz; bandwidth: 50%), which was placed in contact with the eyelid coupled by ultrasound gel. The lateral and axial resolution of the PAOM were ~20 and 23 μm, respectively. It only took 2.7 s for the volumetric PAOM images consisting of 256 B-scans. Multiple additional studies of imaging retinal vasculature by PAOM in small animals have been performed [10, 20, 21, 22]. PAOM as a non-invasive, 3-D microscopic imaging modality, has achieved high-speed, high-resolution in vivo imaging of the vasculature of the retina.

Before 2017, most reported PAM imaging of the posterior segment used small animals, like mice and rats. Mice have an axial length of ~3 mm and rats ~6 mm, which is much smaller that the human axial length of ~23 mm. In order to improve clinical translation of the technology, Tian et al. [19] have developed a novel PAM system to visualize rabbit chorioretinal vasculature. Rabbit eyes have an axial length of 18.1 mm [23], which is almost 80% of the axial length of human eyes. They have used a 570 nm pulsed laser. The excited PA signals have been captured by a custom-built needle-shaped ultrasonic transducer (27 MHz; bandwidth: 60%) placed in contact with the conjunctiva by ultrasound gel. The laser pulse energy on the rabbit cornea was 80 nJ or half the ANSI safety limit [24]. The lateral and axial resolution of the PAM system were 4.1 and 37 μm, respectively. This PAM system obtain high-resolution images of New Zealand rabbit retinal and choroidal vasculature (Figure 4d–f).

2.2.2 Choroidal vasculature

Wei et al. [25] have first applied an OCT-guided PAOM system to image the choroidal vessels in albino rats. The laser output optical wavelength was 578 nm, and the laser pulse energy was set below 40 nJ. The induced PA waves were detected by a custom-built, unfocused needle ultrasonic detector (35 MHz; bandwidth: 50%). The lateral resolution of PAOM was around 20 μm and the axial resolution of PAOM was 23 μm. Based on the OCT cross-sectional images, the distance from the retinal vessel layer to the choroid was found to be 200 μm [26]. Thus, the retinal and choroidal vessels were well resolved along the axial direction. Figure 5a–c show PAOM images of segmented retinal vessels and choroidal vessels. Song et al. [20, 27] have used an integrated PAOM and OCT system to detect in vivo posterior segment imaging of albino and pigmented rats. They have acquired both retinal and choroidal vessels from an albino rat by PAOM, but they were unable to visualize choroidal vessels from a pigmented rat due to melanin from the retinal pigment epithelium (RPE). Figure 5d and e show the contrast images. In addition, Figure 5f and g show the images of choroidal vessels in albino rabbit obtained by Tian et al. [19]. These experimental results have demonstrated that PAI can visualize the choroid and retinal vessels. However, PAM was limited to albino animals when imaging the choroid. The high melanin concentration of the RPE present in front of choroidal vasculature did not allow the illumination light with in visible spectrum range to penetrate through the RPE to generate sufficient PA signal from the choroid for imaging. Near infrared (NIR) light may be a solution to allow PAOM to image choroidal vessels in pigmented eyes.

2.2.3 Retinal neovascularization

Retinal neovascularization (RNV) is the growth of abnormal new retinal blood vessels and represents a major cause of vision loss and blindness. RNV is a common complication of several retinal diseases, including proliferative diabetic retinopathy (PDR), retinal vein occlusions (RVO), sickle cell retinopathy, and retinopathy of prematurity (ROP).

Zhang et al. [11] have demonstrated an integrated PAM, OCT, and fluorescence microscopy (FM) multimodality system to evaluate RNV in vivo in rabbit eyes. The model of RNV was induced by intravitreous injection of vascular endothelial growth factor (VEGF), which mimics numerous retinal diseases, such as PDR. Using a similar PAM system as Tian et al. in 2017 but with a laser wavelength of 532 nm, this study has presented high-quality visualization of retinal vasculature and RNV in albino and pigmented rabbits in vivo. Figure 6 shows that the blood vessel diameter and blood flow in neovascularization are smaller than those of normal vessels, yet PAM can detect neovascularization and the small, irregular vascular network. At the same time, Nguyen et al. [28] have found that multi-wavelength PAM can selectively monitor and detect RVO and RNV in the rabbit retina. This has demonstrated that PAM could achieve label free imaging of microvasculature and RNV without administration of exogenous contrast agents. In addition, rabbit eyes have a similar size to human eyes, which paves the road to clinical translation of the technology.

2.2.4 Choroidal neovascularization

Choroidal neovascularization (CNV) is characterized by the abnormal growth of new vessels originating from the choroidal vasculature and their subsequent growth under the RPE, subretinal space, or a combination of both [29]. CNV most commonly occurs in neovascular age-related macular degeneration (AMD), which is a major cause of vision loss. Dai et al. have used PAM system to investigate laser-induced rat CNV evolution. For the PAM system, the laser output wavelength was 532 nm, and the pulse energy was 60 nJ. The induced PA signals from the retina were detected by a customized ultrasonic transducer (30 MHz; bandwidth: 15 MHz). the axial and lateral resolution of PAM were 50 and 20 μm, respectively. The new capillaries growing from the choroid up to subretinal space can be distinguished as shown in Figure 7.

2.3 PA imaging in retinal metabolism

Retinal oxygen metabolic rate (rMRO₂) is an essential parameter in the retina and can help further understanding of some blinding diseases, such as diabetic retinopathy [31, 32] and glaucoma [33, 34]. The precise measurement of rMRO₂ can be critical in investigating these blinding diseases. Song et al. [35] and Liu et al. [36] have successfully determined rMRO₂ in rats by integrating PAOM and OCT. Obtaining rMRO₂ measurements required measuring retinal blood flow and sO₂ together. They have quantified retinal blood flow by Doppler SD-OCT, and retinal sO₂ by PAOM at three wavelengths (570, 578, and 588 nm). Owing to the distinct light absorption spectrum between HbO₂ and HbR, multi-wavelength imaging can assess the sO₂ in retinal vessels. They have calculated total retinal blood flow as 7.43 ± 0.51 and 7.38 ± 0.78 μL/min within the venous and arterial systems, respectively. The sO₂ value in arterial and venous blood were 93.0 ± 3.5 and 77.3 ± 9.1%, respectively. In PAOM system, they used a tunable laser (output laser wavelength: 570, 578, and 588 nm), the laser energy was 40 nJ/pause. Acoustic waves were detected by an unfocused small-footprint ultrasonic transducer (40 MHz; bandwidth: 30 MHz). The lateral resolution was 20 μm, and the axial resolution was 23 μm. The study measured rMRO₂ in normal small animals. Hariri et al. [37] have demonstrated the use of PA ocular imaging (PAOI) in measuring chorioretinal oxygen saturation (CR-sO₂) gradients in New Zealand white rabbits with ocular ischemia model. The PAOI signal showed a sixfold decrease in CR-sO₂ after significant elevation of IOP during ischemia. In the PAOI, they used a tunable laser (680–970 nm), 750 and 850 nm were used to differentiate the HbO₂ and HbR. A linear array ultrasound transducer with a center frequency of 15 MHz was used to detect acoustic signals. The lateral and axial resolution were 580 and 290 μm, respectively. One limitation of the study was balancing special resolution and penetration depth. Although they could achieve much deeper penetration, they could not discriminate between the retina and the choroid, nor could they distinguish individual vessels.

2.4 PA imaging of melanin of the eye

Melanin is naturally present in the eye within the iris, ciliary body, pigmented choroid, and RPE [38]. The RPE is a single layer of epithelial cells beneath the neurosensory retina and tightly adherent to the underlying choroid. The RPE plays a crucial role in the overall health of the retina: nourishing photoreceptors and disposing of retinal waste and metabolic end products. The melanin concentration in RPE can decrease over time due to light exposure and oxidative stress [39]. RPE melanin decrease is a sign of ocular senescence and is a risk factor and a pathognomonic sign of AMD [40, 41]. To further study melanin, PAI has been used to specifically detect and quantify melanin [42].

In 2010, Silverman et al. [43] for the first time have demonstrated that PAI can acquire images of melanin in the iris in ex vivo pig eye. Jiang et al. [44] have designed an adaptive optics PAM (AO-PAM) system to correct the wavefront errors of the illuminating light of PAM. The lateral resolution of PAM was improved to 2.5 μm with AO. They imaged the ciliary body and the RPE layer of an ex vivo pig eye. In addition, the single RPE cells were first resolved by AO-PAM system (Figure 8a and b). Jiao et al. [12] have successfully acquired images of the retinal vessels and RPE in pigmented rat eyes by using integrated PAOM with OCT system with a 23 μm axial resolution. The study was the first in vivo demonstration of PA imaging melanin of eye. Zhang et al. [46] have combined autofluorescence (AF) with PAOM system to image the RPE melanin and lipofuscin in albino and pigmented rat eyes. Song et al. have developed a multimodal system that contains PAOM, SLO, OCT, and FA to image the eye. They have demonstrated that melanin from the RPE and choroid was able to be imaged in vivo in the retinal of albino and pigmented rats, [20] as well as mice [21]. To demonstrate the feasibility of NIR-light PAOM in vivo, Liu et al. [47] have developed a dual-wavelength (532 and 1064 nm) PAOM and further integrated it with SD-OCT to image albino and pigmented rat eyes. They found that NIR light (1064 nm) PAOM system successfully imaged RPE/choroidal melanin with good signal-to-noise ratio (SNR) in pigmented rat eyes. After that, in 2015, Liu et al. [48] have demonstrated the feasibility of an OCT-PAM system working in the NIR-light. The OCT and PAM used the same broadband light source with a center wavelength of 800 nm. This system can provide a deep-penetration depth and melanin-specific images of the retina in pigmented rats.

The above papers are qualitative studies of eye melanin by PAI. PAI also has the capability of providing a quantitative reading of melanin concentration in the eye. Shu et al. [49] have performed a Monte Carlo (MC) stimulation to investigate light propagation and energy deposition in the eye and tested the feasibility of using PAOM to quantify the RPE melanin concentration. In this study, they used PA signals from retinal blood vessels as references to measure RPE melanin variation. However, the main challenge in accurate quantification of RPE melanin concentration is to distinguish the RPE from underlying pigmented choroid. The RPE is a 10 μm thick cell monolayer and is tightly attached to the choroid, and thus a higher axial resolution of PAOM was necessary. Shu et al. [45] then have used a micro-ring resonator (MRR) detector in their PAOM system to increase the axial resolution (<10 μm). They obtained images where the RPE and choroid can be distinguished in ex vivo porcine and human eyes and quantified the absolute melanin concentrations in the RPE and pigmented choroid, respectively (Figure 8c–h).

PAI can obtain volumetric images of melanin in the eye and can quantify melanin in the RPE and the choroid. However, the technique has yet to be explored in eye disease models and the human eye, so it is unknown that how it can be used in eye research and clinic in the future.

2.5 Multimodal PA imaging system

PA imaging can obtain structural and functional imaging of the eye. PAI can also be combined with other ocular imaging devices, such as OCT, SLO, AF, AO, FA, and fundus photography. These combined multimodal platforms can compensate the weaknesses of each system and provide more comprehensive information of the eye. That would be very beneficial for investigating ocular pathology and detecting disease. For instance, with correcting wave-front errors with AO, the lateral resolution of PAM can be improved [44]. With PAM and AF combined, information on the distribution and concentration of retinal melanin and lipofuscin can be obtained in vivo simultaneously [46]. The most commonly integrated imaging systems combine PAI with OCT since this carries several advantages. OCT can not only acquire posterior segment anatomic information with high axial and lateral resolutions, but also can quantitatively measure retinal blood velocity and flow rate. In addition, OCT can also guide PAI, helping the PAI system image a specific region of interest in the posterior segment of the eye. Song et al. [20, 21] have developed a multimodality system which integrated PAOM, SD-OCT, AF-SLO, and FA to provide optical absorption, optical backscattering, and fluorescence properties of the retina in albino and pigmented rats and mice. Song and Liu et al. [35, 36] have quantified the rMRO₂ using PAOM and SD-OCT. PAOM measured the sO₂, and SD-OCT mapped the blood flow rate. Tian and Nguyen et al. [19, 28, 50, 51] have used a PAM and OCT dual-modality system to acquire chorioretinal imaging and retinal neovascularization imaging in living rabbit eyes. Zhang et al. [11] have described a multimodality imaging system that combines PAM, OCT, and fluorescence microscopy (FM) to evaluated retinal angiogenesis in albino and pigmented rabbit eyes. The multimodal images can be acquired from a single imaging system and co-registered on the same image plane. Thus, ocular diseases can precisely be detected and visualized at earlier stages, resulting in improved understanding of pathophysiology, improved management, and improved monitoring of retinal treatment to prevent vision loss.

3.1 Innovative laser sources for PA imaging

PA imaging can achieve anatomic information with monochromatic pulsed laser illumination. However, to acquire ocular functional information using PAI, multi-wavelength laser illuminations is required. An optical parametric oscillator (OPO) (for example, the Ekspla NT-242, pulse repetition rate 1 kHz, duration 3–6 ns, tunable wavelength range 405–2600 nm) can be used as a suitable laser source to obtain the above information. However, in multimodality systems, such as PA imaging and OCT, the two modalities can require different types of illuminations and thus separate light sources. Liu et al. [48] have used a single ultrafast laser source (pulse repetition rate 10 kHz, pulse duration 3 ns, center wavelength 800 nm; bandwidth 30 nm) to acquire melanin-specific images of in vivo rat retina simultaneously from PAM and OCT. However, this shared laser was limited to imaging melanin since the absorption of hemoglobin was relatively weak in the NIR range. Recently, pulse laser diodes (PLD) have been described as an alternative laser excitation source in mouse ear vascular structures [52]. In the future, one could envision PLD and other novel laser sources for PAI and multimodal imaging applied to ophthalmology.

3.2 Non-contact detection of PA signal

In reported studies to date, researchers either placed a water tank on the cornea or directly coupled an ultrasonic needle transducer with the eyelid or sclera using ultrasonic gel or balanced salt solution (BBS). Given the sensitivity of the human eye, this is not ideal for clinical application on patients. Non-contact remote sensing has been described where an interferometer is used to sense the surface displacement induced by PA ultrasound signals [53, 54]. Discovering a non-contact ultrasonic detection with both stability and sensitivity is a promising area of active investigation.

3.3 Field of view of PA imaging

The field of view (FOV) of PA imaging is limited by the ultrasound transducer. The typical scan area of reported studies has been shown in Table 1. For most studies, the image area was no more than 3 × 3 mm. The human retina area is about 10 cm², so increasing the field of view would be beneficial, possibly through novel ultrasound transducer configurations or array patterns.

3.4 Speed of PA image acquisition

Several current PA imaging systems can acquire high-resolution images in <1 min. The imaging speed is limited by the laser pulse repetition rate. The image speed can be reduced until it depends on the ultrasound propagation time from the posterior eye. The advantage of improving imaging speed is that it will reduce motion artifacts and increase patient comfort. Ideally a clinical PA system would acquire images in less than a second to a couple seconds.

3.5 Lateral and axial resolution of PA

The lateral resolution of PA imaging is determined by the smallest achievable spot size of the illuminating light in the retina. Tian and Zhang et al. described the lateral resolution of their PAM system as 4.1 μm [11, 19]. AO allows one to correct for ocular aberrations and has been applied to ophthalmic imaging to improve resolution. A feasibility study was conducted on the integration of AO with PAM for ex vivo imaging of the eye, which have improved the lateral resolution of the PAM to 2.5 μm. If one adds AO into multimodality systems, it can further improve the lateral resolution of PA imaging.

The axial resolution of PA imaging is determined by the bandwidth of the ultrasound transducer. The larger ultrasonic bandwidth will allow higher axial resolution, but with less sensitivity. Therefore, a balance is required between axial resolution and sensitivity.

3.6 Real-time PA imaging

Currently many PA images require post-image processing to display images. Further investigation of an advanced ophthalmic PA imaging system is to obtain real-time imaging.

3.7 Imaging the choroidal vasculature in pigmented animals

PA imaging is often limited to albino animals when imaging the choroid due to the high PA signal of the RPE melanin in pigmented animals. Further methods need to be developed and refined to penetrate the RPE layer and allow improved choroidal visualization. NIR light has been proposed as a solution, but further testing is needed.

3.8 Imaging ocular disease animal models

PA imaging has been demonstrated in some animal eye models to date, including mouse corneal neovascularization model, rat CNV model, and rabbit retinal neovascularization model. PA imaging needs to be further explored in more eye disease models to evaluate how the information provided by PA imaging will be used clinically in the future.

3.9 Exogenous contrast agents and theranostics

To improve PA image sensitivity and specificity, exogenous contrast agents can be utilized. PA imaging with exogenous contrast agents also allows one to extend the imaging scope to molecular imaging [56, 57]. Exogenous contrast agents for PA imaging include organic and inorganic agents. However, each agent has advantages and limitations. Organic agents (e.g., ICG) can have a limited level of contrast enhancement but a more rapid path to clinical translation [58]. Inorganic agents (e.g., gold nanoparticle (AuNP)) can offer higher contrast but a less rapid path to clinical translation due to less long-term evidence of biosafety [59, 60]. Thus, exploring suitable exogenous contrast agents with safety and high contrast can be used for PA imaging will be meaningful in future research. In addition, theranostic agents that can be used both for diagnostic imaging and therapy, should be further refined and developed to allow for targeted therapy at the time of imaging.

3.10 Safety evaluations and clinical approval

Before PA imaging can be applied to human imaging, a thorough evaluation of laser safety is necessary. Although the reported systems compare their laser fluence to the ANSI laser safety regulations, [24] one must monitor the long-term effect of both single and multiple imaging sessions on the structure and function of the retina. In additional, regulatory approval should be sought for a clinical system so that PA can be applied to patients and diseases.