Quantified elasticity mapping of retinal layers using synchronized acoustic radiation force optical coherence elastography

Age-related macular degeneration (AMD) is the leading cause of blindness in the elderly (over the age of 60 years) in western countries. In the early stages of the disease, structural changes may be subtle and cannot be detected. Recently it has been postulated that the mechanical properties of the retina may change with the onset of AMD. In this manuscript, we present a novel, non-invasive means that utilizes synchronized acoustic radiation force optical coherence elastography (ARF-OCE) to measure and estimate the elasticity of cadaver porcine retina. Both regions near the optic nerve and in the peripheral retina were studied. An acoustic force is exerted on the tissue for excitation and the resulting tissue vibrations, often in the nanometer scale, are detected with high-resolution optical methods. Segmentation has been performed to isolate individual layers and the Young’s modulus has been estimated for each. The results have been successfully compared and mapped to corresponding histological results using H&E staining. Finally, 64 elastograms of the retina were analyzed, as well as the elastic properties, with stiffness ranging from 1.3 to 25.9 kPa in the ganglion to the photoreceptor sides respectively. ARF-OCE allows for the elasticity mapping of anatomical retinal layers. This imaging approach needs further evaluation but has the potential to allow physicians to gain a better understanding of the elasticity of retinal layers in retinal diseases such as AMD.

have been many segmentation methods that have been used to separate the different layers within the retina [4][5][6]. Many of these features are built into commercial imaging systems, and automated detection is possible. The thickness of the layers and other anatomical structures can be analyzed in-vivo and correlated with pathology. Although, this structural information is very helpful in mid to late stage AMD, often it is not sufficient for very early diagnosis of AMD.
As a precursor to neovascularization and disease progression, the elasticity of the tissue will change [7][8][9][10]. Friedman et al. demonstrated that increased scleral rigidity might be a precursor to the onset of AMD, which brought to attention the possibility of using mechanical properties as a diagnostic tool [8]. Shahbazi et al. used ultrasound to show that the posterior ocular elasticity of patients with AMD is indeed different than that of healthy humans, but did not offer retinal layer analysis [9]. Recently, Chen et al. presented the possibility of retinal elastin decreases during the early onset of AMD [10]. Since these cellular level changes are expected to precede structural abnormalities that can be detected with current imaging techniques, we believe that there is significance in studying the mechanical properties of the retina. In addition, elasticity changes in the retina and choroid can also occur when the microvasculature changes or when drusen forms. Optical coherence elastography (OCE) may offer an alternate method to detect the environmental changes that occur when alteration of cellular properties begin to take place. Measuring the elasticity will give scientists and physicians a better understanding on the disease mechanism in the early stages, and possibly provide a powerful diagnosis tool.
In order to study the mechanical structure of the retina, several studies attempted to provide elastic properties by performing mechanical strain testing in vitro [11][12]. However, strain testing is not possible for tissues in vivo. In addition, the entire retina is extracted as a single unit, which means that individual layer information cannot be obtained. Although mathematical modeling of the retina to determine the Young's modulus has been reported, in vivo determination remains a challenge [13]. In order to perform in vivo imaging with information of different layers, a high resolution system functional imaging system is necessary.
OCE is a relatively new method of providing elasticity mapping with high resolution and sensitivity [14]. Several different applications have been studied using this technique, including the study of corneal elasticity [15][16][17][18]. However, the mechanical properties of the retina are still not well defined since the retina is inaccessible to many elastography methods or is limited by low resolution. We recently reported on an acoustic radiation force (ARF) OCE method that can map out the elasticity of the cornea both axially and laterally with high resolution [19].
In this manuscript, we present the quantified retinal layered elasticity map for the first time in a porcine model. The instrumentation has been updated, including a mechanical stage to increase the field of view and synchronization to generate a spatial elasticity map. In this way, the phase cycle of the tissue vibration is uniform in the direction parallel to the mechanical stage. First, structural optical coherence tomography (OCT) and functional OCE imaging were performed on a healthy pig retina near the optic disc and in the periphery of the retina after isolation of the posterior portion of the eye. Then the OCE phase information was analyzed along the axial direction. Then segmentation was performed on the retinal layers using OCT and the corresponding layer was matched on the OCE and histology. Nanometer displacement differences were observed between the layers. Finally, the relative stiffness was analyzed over 64 B-scan samples and statistical analysis was performed. The Young's moduli are estimated for each layer using the experimental stiffness ratios and average elasticity obtained from literature.

System d
In this elastog a spectral dom optical signal synchronized Fig. 1

Retina a
A porcine ey removed so th one lateral dim on. This increa s more uniform an in these exp ultrasonic tra submerged in t avels through t m. In Fig. 1 eye was was then fixed using agar gel and mounted to the imaging stage in the water bath. The optical disc was identified on the retina by its diameter of approximately 1.7 mm. Using the synchronized system, a region of 3 mm by 3 mm was scanned near the optical disc. Previously [19], only the middle 500 μm by 500 μm region can be assumed to have uniform excitation, but with the addition of the mechanical stage, a region of 500 μm by 3 mm is able to achieve uniform force. This expands our imaging region by 6 fold in this figure.
The OCT cross-sectional image showing the optical disc region is displayed in Fig. 3(a), while the corresponding OCE phase map is displayed in Fig. 3(b). The B-scan shown is in the same direction as the movement of the mechanical stage. Within this B-scan, an accurate uniform acoustic field is guaranteed for elastography since synchronization allows for phase cycle uniformity in the direction perpendicular to the galvo scanning.
The OCT and OCE images of the peripheral retina are shown in Figs. 3(c) and 3(d), respectively. The same scanning mechanism was used over a 3 mm by 3 mm region after shifting the focus to the peripheral retina, and then extracting the phase information, , where σ is the stress and z is the axial depth. It is clear that the optic disc region had a much smaller vibrational response, indicating a stiffer tissue, than the peripheral retina and the retinal regions closer to the optic disc. The pocket structure in the middle of Fig. 3(b) had a very high vibrational response, which indicates a softer tissue that is concluded to be a collapsed blood vessel. This figure portrays the elasticity of the optic nerve relative to the peripheral retina and demonstrates the feasibility of using ARF-OCE in the mapping of retinal elasticity. It was also noted that the relative stiffness of the retina changes in the axial direction, suggesting the separation of different layers, with the top layer indicating the softest structure. Previously, we have shown that the ARF field is uniform across the imaging axial depth, and verified the ability of OCE to separate the stiffness of individual axial and lateral components in a connected agar phantom with inclusions [19]. OCE was able to extract the stiffness of the 2 layers of different elasticity accurately according to verification with compression testing methods [19]. In addition, we have also demonstrated that the vibrational response is uniform within the axial imaging range for a uniform phantom [20], which means that the material attenuation is minimal within the 1 mm imaging range. Therefore, according to the phase map, we conclude that the stiffness of the retina increases from the inner ganglion side (white arrow) to the outer photoreceptor side (yellow arrow).

Segmen
Since differen images, segm presented pre could be isola 4b was also s directly since between laye values, which retinal anatom results. There the imaging la cell layer, an probably cau collapsed blo approximately during storage The boun images, and t Within the in with a redder color or high connective bo possible to co The outer plex outer nuclear low scattering outermost lay but low displa μm for the po 3. Optical disc an phase cross-sectio cross-section of p ates ganglion side ntation and hi nt layers were mentation was eviously by ou ated as shown egmented usin e we have yet ers. Figure 4 [21].
in Fig. 4 to be 13 e using a dual layer elasticity, where the weight contribution was dependent on the thickness of the respective layers, and the mean was set to 13 kPa. The results are shown in Table 1, where the Young's modulus was estimated anywhere from 1 to 26 kPa. The outer nuclear layer was omitted from this estimation since scattering signal was too low for proper detection of OCE. According to previous literature, the bulk elasticity of the retina has been determined to be 20 kPa with mathematical modeling methods [24], while the layered elasticity has been shown to increase from the ganglion to the photoreceptor sides, ranging from approximately 10 to 30 kPa using shear wave OCE methods [25]. The results from the ARF-OCE study closely agree with the reported values, and the increasing elasticity trend from the ganglion layer has also been similarly observed. The advantages of this ARF-OCE system lies in the capability to noninvasively access the posterior eye and the high-speed mechanical mapping using a continuous modulated excitation. In addition, this ARF-OCE technology has nanometer scale displacement sensitivity, which allows us to minimize the ARF power [26]. According to the results for 4 retinal layers, it seems that the inside layer on top of the retina is much softer than the bottom layers, which is getting closer to the sclera. This is expected since the sclera, which is the protective outer layer, is much stiffer than the retina. A total of 64 B-scans were analyzed to correspond to 80% confidence level with 8% confidence interval. However, further analysis is necessary to accurately measure the retinal layers between several different samples. Since the ARF-OCE measurements are based on the sample displacement, it is less accurate for stiff samples that correspond to a low tissue displacement. The reason is that the measurements are often at the ten-micron level, meaning that a small amount of noise or motion will create a larger percentage of error. Therefore, in Table 1, the error increases with the stiffer layers of the retina as expected.

Discussion and conclusion
This study using ARF-OCE technology is the first to visualize the mechanical properties of individual retinal layers, where 4 distinct layers were quantified. An ARF-OCE system was set up with synchronized excitation, detection, and scanning for better control of the modulation phase cycle. By analyzing the cross-sectional images perpendicular to the galvanometer scanning direction, it can be guaranteed that the B-scan shows a synchronized vibration in a uniform phase cycle. Using this approach, elasticity maps were obtained for the peripheral retina and the optic disc regions. Segmentation was performed on the OCT images and the corresponding layers were separated in the OCE, and both were matched to histology results using H&E staining. Further analysis was performed on 64 B-scans to estimate the thickness of the 5 layers as well as the relative displacement values. Using a weighted average method, the Young's moduli for 4 different retinal layers were estimated.
Since this is a feasibility study to demonstrate the ARF-OCE method to generate quantified elasticity maps, only results from 1 sample is analyzed, and the sample was not very fresh as can be seen by the collapse of the top three layers. We are currently doing work on more samples in order to study the consistency between different subjects. However, freshness of the samples remains a problem. Since the porcine eyeball deteriorates at a rapid rate, and retinal structure is highly correlated with perfusion and freshness, it is difficult to obtain ex-vivo data. In addition, the anterior of the eye is always clouded, making the removal of the cornea and lens a necessity. However, since the procedure itself is non-invasive, it has a very optimistic outlook in in-vivo animal studies. There are a few issues that must be addressed in advance.
The imaging region of the sample must be further expanded in order to decrease imaging time and complexity during the procedure. This can be solved by using an ultrasound transducer with a wider focus or no focus. Scanning can be performed over the entirety of the retina within 12.5 seconds. This also indirectly helps to address the problem of the safety of the acoustic excitation. In order to adhere to the mechanical index limit of 0.23 for human ocular tissue, it is necessary to lower the excitation power per unit area. The current MI was estimated to be around 0.9. Since OCE has nanometer displacement sensitivity, it would be sensitive enough to detect smaller vibrations that adhere to the MI.
Since ARF is used for excitation, a transmission medium is necessary for the propagation of the ultrasound force. While it is sufficient to use a PBS bath in ex vivo settings, the current setup would not be feasible for in vivo imaging. In clinical ultrasound for ophthalmic applications, ocular ultrasonic gel can be directly applied to the eye and serve as the medium between the probe and the eye. Another solution is a steridrape setup, which involves a transparent drape that comes into contact with the eye and the probe is submerged into fluid on the other side. Both these setups can easily be adapted for use with the ARF-OCE system.
Another observation we made was the blending of the OCE information at the boundaries between layers. This is caused by the bulk vibrations of the entire sample, where each of the connected layers affect each other. In addition to the visible layers, the sclera is relatively a much stiffer medium that can also affect the lower layers of the retina. If there is a stiff layer connected to a soft layer, the boundary shows a gradient [16]. This relationship must be taken into account when analyzing the mechanical properties of the layers. Modeling of the viscoelastic phenomenon can help us better understand this issue and make more accurate parameter estimations.
The ARF-OCE technology was used to quantify retinal layered elasticity and is adaptable to in-vivo applications. We believe that this initial demonstration in the porcine eye is a stepping stone to the translation of the technique, which can potentially provide a powerful tool for the clinical diagnostic management of AMD.