|Home | About | Journals | Submit | Contact Us | Français|
Recent advances to the adaptive optics scanning light ophthalmoscope (AOSLO) have enabled finer in vivo assessment of the human retinal microvasculature. AOSLO confocal reflectance imaging has been coupled with oral fluorescein angiography (FA), enabling simultaneous acquisition of structural and perfusion images. AOSLO offset pinhole (OP) imaging combined with motion contrast post-processing techniques, are able to create a similar set of structural and perfusion images without the use of exogenous contrast agent. In this study, we evaluate the similarities and differences of the structural and perfusion images obtained by either method, in healthy control subjects and in patients with retinal vasculopathy including hypertensive retinopathy, diabetic retinopathy, and retinal vein occlusion. Our results show that AOSLO OP motion contrast provides perfusion maps comparable to those obtained with AOSLO FA, while AOSLO OP reflectance images provide additional information such as vessel wall fine structure not as readily visible in AOSLO confocal reflectance images. AOSLO OP offers a non-invasive alternative to AOSLO FA without the need for any exogenous contrast agent.
Intravenous (IV) fluorescein angiography (FA) is the clinical gold standard for assessing retinal vasculature, revealing the extent of disease and response to treatment . Despite its utility, the IV injection of fluorescein dye is considered invasive, and has been reported to cause rare but potentially severe systemic adverse reactions [2–6]. Furthermore, with the limitations in axial and lateral resolution of standard fundus cameras, IV FA provides an incomplete visualization and differentiation of the inner and outer capillary plexuses, as described in human retinal histology . Recent advances in adaptive optics imaging have been employed to overcome these limitations [8–17], and two major techniques are now being used to image the retinal microvasculature for the purposes of research and anticipated adaption into clinical practice [13, 14, 16, 18, 19]. Adaptive optics scanning light ophthalmoscope FA (AOSLO FA) is a confocal imaging technique that allows simultaneous acquisition of structural reflectance and functional perfusion images. The sensitivity of the technique permits the use of orally administered fluorescein which may have a safer side effect profile than conventional IV fluorescein [20, 21]. AOSLO offset pinhole (OP) imaging utilizes an off-center pinhole and has recently been described for the visualization of blood flow and microscopic features of the vascular wall [14, 16]. Coupled with motion contrast processing techniques [22–27], AOSLO OP imaging can reveal microvascular perfusion maps of the retinal vasculature, without the use of an exogenous contrast agent. In this study, we compare the information provided by AOSLO FA and AOSLO OP structural and perfusion maps, in imaging healthy and diseased retinal microvasculature.
This study adhered to the tenets of the Declaration of Helsinki and was approved by the Institutional Review Board of the New York Eye and Ear Infirmary. Three healthy male subjects (3 eyes; age range 23-25 yo) and five patients (5 eyes), each with a single retinal pathology were recruited for this study. Inclusion criteria included normal anterior segment with clear phakic lens, pupil dilation of at least 5 mm, clear media and good central fixation with a best-corrected visual acuity (BCVA) better than 20/80. Eyes had minimal or no foveal edema evident by spectral domain optical coherence tomography (Heidelberg Spectralis HRA + OCT, Heidelberg Engineering Inc., Heidelberg, Germany). All three healthy subjects were self-reported healthy individuals with no significant past medical history and no clinical signs or symptoms of ocular pathology. The pathologies of the patients were: hypertensive retinopathy (HR; RR0095; 47 yo male), diabetic retinopathy (DR; RR0167; 49 yo female), macular drusen (RR0160; 37 yo female), central retinal vein occlusion (CRVO; RR0151; 62 yo female), and branch retinal vein occlusion (BRVO; RR0129; 55 yo female). Written informed consent was obtained after the nature and potential risks of the procedure were explained. All tested eyes were imaged with AOSLO FA and OP during the same visit, except for the BRVO patient, who was imaged with AOSLO OP 4 months after the initial AOSLO FA imaging session. BCVA was recorded for all subjects and patients before AOSLO imaging. Pupils were dilated with 1 drop of 2.5% phenylephrine hydrochloride ophthalmic solution (Bausch & Lomb Inc., Tampa, FL) and 1% tropicamide ophthalmic solution (Akorn Inc., Lake Forest, IL).
Prior to AOSLO imaging, color fundus photography (Topcon 3D OCT 2000, Topcon Corporation, Tokyo, Japan) was performed on all tested eyes. Conventional IV FA using either HRA-OCT (Heidelberg Spectralis, Heidelberg Engineering Inc., Heidelberg, Germany) or TRC 50IX Retinal Camera (Topcon Corporation, Tokyo, Japan) was performed on each of the five patients. No patient received both IV FA and AOSLO FA imaging on the same day. Regions of interest for AOSLO imaging were pre-identified on either the color fundus photographs or IV FAs. Axial length measurements of the tested eyes were obtained using an IOL Master (Carl Zeiss Meditec, Dublin, California). Individual retinal magnification factors were applied to the AOSLO images incorporating the measured axial lengths into the Emsley schematic eye model formula .
The AOSLO used in this experiment was a replica of the one described by Dubra and Sulai  with the visible channel modified for fluorescence imaging. In brief, three different light sources (488, 790 and 850 nm) which entered the eye as superimposed (coaxial), near-collimated 7.75 mm diameter beams were used for FA excitation, reflectance imaging and wavefront sensing respectively. All three beams localized onto a single spot on the fundus, which was rapidly scanned to form a 1, 1.5 or 1.75° square imaging raster, using a 15 KHz horizontal resonant optical scanner and a 16 Hz vertical optical scanner. Synchronous modulation of the light sources turned them on at the start of the imaging portion of each horizontal line used for imaging when scanning from left to right.
For AOSLO FA imaging, the fluorescence excitation light source used was a 488 nm diode laser (Lasos, Lasertechnick GmbH, Jena, Germany), with an interferometric band-pass optical filter centered at 525 nm and 45 nm bandwidth in front of the detector. This was addressed by using a 3.75x Airy disk diameter confocal pinhole. No attempt was made to compensate for longitudinal chromatic aberration variation among individuals. The reflectance channel light source was a superluminescent laser (SLD) with peak wavelength centered at 790 nm (Superlum Ireland, Carrigtwohill, County Cork, Ireland). The confocal pinhole size was approximately 1x and 16.67x the Airy disk diameter, as measured at the detector plane, for the AOSLO confocal reflectance and OP reflectance imaging, respectively.
The optical powers, without modulation, measured at the cornea, were 15 µW for the 850 nm wavefront sensing superluminescent diode (SLD), 100 µW for the 790 nm imaging SLD and 32 µW for the 488 nm fluorescein excitation diode laser. During imaging, the on/off modulation reduced the average powers delivered to approximately 25% of their original values. All sources were considered as lasers for the maximum permissible exposure (MPE) calculations. Light exposure calculations were determined to be 6 times below MPE according to the American National Standards Institute ANSI Z136 .
Each subject was stabilized using a dental impression on a bite bar (Splash! Putty, DenMat Holdings, LLC, Lompoc, California, USA). The bite bar was attached to a three axis translation stage for fine transverse and axial adjustment during imaging.
Subjects were instructed to direct their gaze either towards a green internal fixation target or the corners of the imaging raster. Because of light safety considerations, no retinal location was exposed to the combined light sources (488, 790, and 850 nm) for longer than 120 seconds. Coordinates of these fixation locations were recorded so that the corresponding retinal locations could be imaged accurately and repeatedly. Throughout the imaging session, subjects were encouraged to blink frequently to maintain their normal tear film and were provided with short breaks, at regular intervals or as needed.
For AOSLO FA, 20 mg/kg oral fluorescein (AK-FLUOR 10%, Akorn, Inc., Lake Forest, Illinois) was administered with orange juice to mask its taste. The use of oral fluorescein has recently been demonstrated to be effective and safe for AOSLO FA imaging in human eyes . Depending on the subject weight, 3-4 vials of fluorescein was administered orally with 50 mL of orange juice to mask its taste. An additional 50 mL of orange juice was provided to wash out any residual taste of the fluorescein. 15 minutes post-administration, simultaneous confocal reflectance and fluorescence image sequences were acquired at each region of interest as previously described , consisting of 125 frames at a frame rate of 15 Hz and using a 1.75° field of view (FOV). During AOSLO FA and AOSLO confocal imaging, the retina was simultaneously exposed to three different light sources (488, 790 and 850 nm). The overall imaging time was approximately 15 minutes including short breaks as needed.
AOSLO OP reflectance imaging of retinal microvasculature has been extensively described in previous reports [14, 16]. Briefly, the technique involves displacement of a larger confocal pinhole, located at the retinal plane conjugate in front of the detector. In this study, retinal microvasculature contrast was maximized by manually displacing the confocal pinhole along the x and/or y axis. Typically, displacing the confocal pinhole orthogonally to the orientation of the retinal vessel being imaged enhanced the visibility of the vascular wall structure and blood flow over a 1 or 1.5° FOV [14, 16]. During AOSLO OP reflectance imaging, the retina was simultaneously exposed to two different light sources (790 and 850nm). The overall imaging time was approximately 30 minutes including short breaks as needed.
After AOSLO confocal reflectance and AOSLO FA imaging, the fluorescence sequences were co-registered with the confocal reflectance sequences using custom software. Respective images with high signal-to-noise ratio were then generated by averaging the 5-100 frames in each registered sequence , resulting in AOSLO confocal structural images and AOSLO FA perfusion maps. For AOSLO OP reflectance imaging, image sequences were registered and averaged, resulting in AOSLO OP structural images. As an additional step, AOSLO OP registered sequences (~100 frames per sequence) were used to create AOSLO OP perfusion maps using the motion contrast technique as described previously [13, 14]. Montages were created using Adobe Photoshop CS6 (Adobe Systems, Inc., San Jose, CA). The flow chart of the image processing for different imaging techniques is illustrated in Fig. 1 .
In order to compare AOSLO FA and OP imaging, three different retinal locations were chosen 1) the foveal capillary network, where the retina was thin (~250 µm), 2) 5° from the foveal center where the retina was relatively thicker (~300 µm), and 3) the peripapillary capillary network where the retina was thickest (~350 µm) and contained multiple capillary plexuses. The foveal avascular zone (FAZ) was delineated manually on a layer mask of both the AOSLO FA and AOSLO OP perfusion maps in Adobe Photoshop CS6 (Adobe Systems Inc, San Jose, CA). FAZ area in mm2 and effective diameter (diameter of a theoretical circle with the same FAZ area) were then computed based on these masks using MATLAB (The MathWorks, Inc., Natick, MA)
Additionally, any retinal pathologies evident during the sessions were imaged in order to highlight discrepancies between the two imaging techniques. Five prevalent retinal microangiopathies found in the patients were examined, including 1) microaneurysms; 2) vessel remodeling including vessel looping and collateral formations; 3) leakage; 4) non-perfused blood vessels; and, 5) macular drusen.
Although the light levels were maintained well below the ANSI standard during both imaging sessions, the additional short wavelength 488 nm used for AOSLO FA imaging posed a potential increased risk of photochemical damage to the retina [32, 33]. While oral administration of fluorescein has been shown to be safer than IV administration [20, 21], OP reflectance imaging eliminates any need for a contrast agent, further enhancing the safety of the procedure .
A well-defined FAZ was seen in all three healthy subjects on both AOSLO FA and AOSLO OP perfusion maps (Fig. 2 ). The mean ( ± SD) of FAZ area measured on the AOSLO FA and AOSLO OP perfusion maps was 0.297 mm2 ( ± 0.06) and 0.301 mm2 ( ± 0.06) respectively, resulting in a 1.3% difference in the mean FAZ area and 0.8% difference in mean FAZ effective diameter between the two imaging approaches. All FAZ area measurements were performed on the original perfusion maps. These FAZ areas are in close agreement with previous studies using conventional FA [34, 35] and AOSLO confocal motion contrast [10, 13] in healthy subjects. Figure 2 (right column) shows the superimposed AOSLO FA (green) and OP perfusion maps (red) after bidirectional elastic image registration using ImageJ plugin bUnwarpJ . All capillaries captured on the AOSLO OP perfusion maps were also identified on the AOSLO FA perfusion maps within a 2.4° diameter (~700 µm) centered at the fovea, and appear to be located in a single layer, in agreement with previous histological studies . These findings suggest that AOSLO OP is an effective noninvasive alternative to FA for assessing the perfused foveal capillary network.
Both AOSLO FA and AOSLO OP perfusion maps documented the inner capillary layer showing clear arteriolar capillary free zones (Figs. 3 and and44 ). In thicker regions of the retina, at a given axial focus, AOSLO FA perfusion map (Figs. 3(A) and 4(B)) is able to capture multiple layers of capillaries at a single time including the inner and outer retinal capillary layers compared to AOSLO OP perfusion map (Figs. 3(C) and 4(E)). This is probably due to the phenomenon that there is no structural background to the AOSLO FA images since only fluorescence is recorded. This in turn enhances the relatively lower contrast of capillaries located above and below the focal imaging plane. Unlike AOSLO FA image, images acquired by AOSLO OP reflectance have a uniform background, reducing the visibility of these capillaries. However, AOSLO OP perfusion map is able to capture multiple capillary networks by adjusting the focus to different depths while imaging and subsequently superimposing these serial images. Also note that the visualization of single file flow of red blood cells within peripapillary capillaries in Fig. 4(E) (Media 1).
Figure 3(B) and 3(D) show the same perifoveal arteriole with a 50 µm lumen diameter captured on AOSLO confocal structural and AOSLO OP structural images, respectively. Vascular wall fine structure including vascular mural cells are readily visualized by AOSLO OP structural image (Fig. 3(D), black arrow), but not on AOSLO confocal structural image (Fig. 3(B)). The enhanced visibility of the vascular wall fine structure is achieved because the off-centered pinhole simultaneously blocks the direct back scattered light from the overlaying retinal nerve fiber layer and collects multiply scattered light [14, 16].
A variety of retinal vascular abnormalities were studied using AOSLO confocal reflectance, FA, and OP reflectance imaging, including microaneurysms, vascular luminal clot, vessel looping, engorged and dilated blood vessels, vascular leakage, non-perfusion and capillary dropout, tortuous blood vessels, and irregular vessel lumen diameter.
Figure 5 shows a microaneurysm in the superior retina 6° above the fovea in a 49 year old female with DR. The AOSLO OP structural images and perfusion maps reveal a relatively normal inner capillary layer (Fig. 5(A) and 5(C)), while a 30 µm diameter microaneurysm can be seen by shifting the focus 120 µm deeper towards the outer retina (Fig. 5(B) and 5(D)). This is consistent with histological studies that localize most diabetic microaneurysms to the outer capillary plexus within the inner nuclear layer [38, 39]. Images with depth information can be created by superimposing serial AOSLO OP perfusion maps as illustrated in Fig. 5(F).
In AOSLO FA, fluorescein dye in the blood stream outlines the perfused retinal vasculature. AOSLO OP motion contrast image processing, on the other hand, employs motion of erythrocytes and/or leukocytes as the intrinsic marker of perfusion. Pixel intensity in the AOSLO FA perfusion map represents fluorescein dye accumulation, whereas the pixel intensity on the AOSLO OP perfusion map represents the blood flow. While the two perfusion maps appear similar, care must be exercised during interpretation or when comparisons are attempted for diagnostic or investigative purposes. An example demonstrating these differences is shown in Fig. 6 , in which various blood pooling and flow patterns of retinal microaneurysms were captured in the temporal retina in a 47 year old male with hypertensive retinopathy. Pooling of the blood is indicated by a relatively higher pixel intensity inside the microaneurysms than in the surrounding capillaries in the AOSLO FA perfusion map (Fig. 6(A), Media 2 and Media 3). A more complex structure of these microaneurysms is revealed by the AOSLO confocal structural and OP structural images, including vessel wall features and luminal clotting with infiltration, as demonstrated in Fig. 6(B) and 6(D). Unlike AOSLO FA, AOSLO OP perfusion maps (Fig. 6(C)) show relatively lower pixel intensity on the lower part of the microaneurysm than the upper part (Fig. 6(A) and 6(C), white arrows). This suggests that while a greater concentration of fluorescein dye was present in the lower part of this microaneurysm, the blood flow velocity was relatively slower when compared to the upper part. The registered AOSLO OP structural videos of the two corresponding microaneurysms (Fig. 6(D)) demonstrate a range of blood flow velocities (Media 4 and Media 5) and confirm the intensity discrepancy between AOSLO FA and AOSLO OP perfusion maps. It is anticipated that with the addition of a higher frame rate, AOSLO OP may provide a unique opportunity to study the flow dynamics of microaneurysms with different morphologies.
Vascular remodeling at various stages including vessel looping, capillary dropout and engorged/tortuous blood vessels were seen in all patients with vasculopathies. Figure 7 demonstrates these phenomena in a 55 year old female with BRVO. Both AOSLO confocal and OP structural images showed vessel looping of an arteriole (Fig. 7(A) and 7(C)). The corresponding AOSLO FA and OP perfusion maps revealed a capillary dropout region (Fig. 7(B) and 7(D), white arrows) and vessel looping (Fig. 7(B) and 7(D), white arrow heads). Figures 7(E) and 7(F) show a magnified structural image of the regions enclosed in the white boxes in Figs. 7(A) and 7(C). While the fine structure of blood vessel wall (Fig. 7(F), black arrows) could be readily distinguished on the AOSLO OP structural image, only limited vessel wall detail could be extracted from the AOSLO confocal structural image. Figure 7(G) shows a magnified AOSLO OP structural image of the region enclosed in the black box in Fig. 7(C), showing an uneven lumen diameter of the arteriole (large black arrow), vessel looping (white arrow), and non-perfused capillaries (small black arrows).
Blood vessel leakage could only be detected on AOSLO FA perfusion maps. Figure 8 shows a foveal capillary network (~1.5° superior temporal) in a 62 year old female with CRVO. The AOSLO FA perfusion map shows focal leakage, indicated by the white arrow in Fig. 8(B). In contrast, AOSLO confocal and AOSLO OP structural images and AOSLO OP perfusion maps do not have the capability to show leakage (Fig. 8(A), 8(C), and 8(D)). Fluorescein leakage has been shown to be useful in assessing inner blood-retinal barrier function and retinal vascular endothelial integrity [40, 41], as well as for monitoring response to treatment. AOSLO OP structural images, however, were able to reveal non-perfused blood vessels (Fig. 8(C), black arrows), which were clearly absent on both AOSLO FA and OP perfusion maps. Furthermore, AOSLO FA perfusion maps may be more sensitive to detecting blood vessels with slower or weaker perfusion than AOSLO OP perfusion maps. Figure 8(B) (yellow arrow head) shows a weakly perfused capillary with lower FA signal intensity. The same capillary appears as non-perfused on the AOSLO OP perfusion map (Fig. 8(D), yellow arrow head), possibly due to a slower flow of blood cells.
Both the AOSLO FA imaging and AOSLO OP motion contrast processing techniques are subject to artifacts that may appear as vascular flow. AOSLO FA perfusion maps occasionally contain fluorescent structures which do not represent fluorescein movement. AOSLO OP perfusion maps may also show motion artifacts that does not appear to originate from blood cell motion.
Figure 9 shows a comparison of the AOSLO FA and OP perfusion maps at ~1° superior temporal to the fovea in a 37 year old female with macular drusen. While an intact foveal capillary network is clearly visualized in the AOSLO confocal structural image (Fig. 9(A)), the corresponding AOSLO FA perfusion map shows two diffuse hyperfluorescent regions appearing to emanate from the capillaries like microaneurysms (Fig. 9(B), white arrows). Since these structures do not appear on the AOSLO confocal structural images, the probable source of this hyperfluorescence is macular drusen and the autofluorescent property of RPE lipofuscin . Two corresponding hyper-reflective regions were captured on the AOSLO OP structural image (Fig. 9(C), black arrows). Compared to the AOSLO confocal structural image, the enhanced visualization of macular drusen on the AOSLO OP structural image suggests that the majority of the light returning from the drusen is multiply scattered light [43, 44]. This implies that some retinal features can be imaged better using an AOSLO tuned intentionally for collecting multiply scattered light instead of direct back scattered light. This also suggests a potential application of AOSLO OP reflectance imaging for studying age-related macular degeneration. The AOSLO OP perfusion map of the same region also shows an intact capillary network (Fig. 9(D)).
Clear motion artifacts, or a non-vascular origin, observed in some AOSLO OP perfusion maps (Fig. 3(C), black arrow & Fig. 7(D)) can be recognized by comparison with AOSLO FA perfusion maps. Artifacts in the AOSLO OP perfusion maps along the vessel wall suggest an imperfection of image registration causing non-blood flow related intensity variations. Further improvements to the AOSLO OP image registration technique may help eliminate these spurious motion artifacts. Table 1 summarizes the capabilities of the different techniques for retinal microvasculature imaging.
We have demonstrated that AOSLO OP provides retinal microvascular images comparable to those obtained with AOSLO FA with additional structural information such as vessel wall details. The non-invasive nature of AOSLO OP imaging makes it a very appealing imaging tool for screening and monitoring vasculopathies in the retina. AOSLO FA, however, uniquely demonstrates breakdown of the blood-retinal barrier sufficient to allow leakage and may be more sensitive for revealing marginally perfused capillaries with slower blood flow. The complementary information, these techniques provide, enhances our appreciation of the complexity of the retinal microvascular networks more than could be achieved using either technique alone. Moreover, both techniques offer the ability of detecting subclinical microvascular changes which will become increasingly valuable as we move towards earlier therapeutic management of these diseases.
Funding was provided by Marrus Family Foundation, Bendheim-Lowenstein Family Foundation, Wise Family Foundation, Chairman's Research Fund of the New York Eye, Glaucoma Research Foundation Catalyst for a Cure initiative and Ear Infirmary, an unrestricted departmental grant from Research to Prevent Blindness and the Glaucoma Research Foundation. NIH grants P30EY001931, UL1RR031973 and T32GM07356. Alfredo Dubra-Suarez is the recipient of a Career Development Award from Research to Prevent Blindness and a Career Award at the Scientific Interface from the Burroughs Wellcome Fund. This research was also supported by grants from the RD & Linda Peters Foundation and an unrestricted departmental grant from Research to Prevent Blindness. The authors would also like to thank Eric Cheang, Chun Lin Liu, Mathew Birnbaum, Lenny Rostomian, Ahmad Rehmani, Moataz Razeen, and Drew Scoles for their assistance with image processing.