Fig. 11.1
Examples of AO retinal images. (a) Confocal image of photoreceptors at the edge of the foveal avascular zone. (b) Confocal image of the nerve fiber layer and vessels. (c) Motion contrast image of capillaries at the edge of the foveal avascular zone [11]. (d) Confocal image of hard exudates among cone photoreceptors in a patient with diabetic retinopathy. (e) Split detection image of cone photoreceptor inner segments. (f) Motion contrast image of microaneurysms and capillaries in a patient with diabetic retinopathy [39]. Images shown in (a) and (b) were acquired using an adaptive optics instrument manufactured by Canon, Inc. (Tokyo, Japan), similar to [51]. Images shown in (c), (d), and (f) were acquired using a custom-built adaptive optics instrument described in [48]. The image shown in (e) was acquired using a custom-built adaptive optics instrument described in [50]
The first report of AO utilization in a clinical setting was in 2000 in a single patient with cone-rod dystrophy [32]. Since then, the application of AO in clinical settings has rapidly expanded [33] such that AO has been used to investigate a large number of conditions including age-related macular degeneration [16, 34–36], diabetic retinopathy [37–39], glaucoma [23, 25], and inherited retinal degenerations [40–45].
This chapter is intended to bridge the gap between research and clinical applications by providing an overview of the current clinical capabilities of AO, practical considerations before pursuing an endeavor in AO retinal imaging, and the potential for using AO in inflammatory eye conditions.
Basic Components of an Adaptive Optics Instrument
All hardware-based AO instruments contain three key components which are integrated into an underlying imaging platform:
Wavefront sensor. The human eye has natural aberrations that limit the ability of imaging instruments to capture detailed images of the retina [46]. These aberrations can be measured using a wavefront sensor. In AO retinal imaging, the most popular wavefront sensor in use is the Shack-Hartmann wavefront sensor, which generates a characteristic array of spots (Fig. 11.2).
Wavefront corrector. Once the aberrations are measured, they need to be corrected by the system. In AO retinal imaging, both deformable mirrors and liquid crystal on silicone spatial light modulators have been used. Any wavefront corrector that is used will have its limits in terms of the amount and types of wavefront errors that it is able to correct.
Control system. A software algorithm is needed to link the wavefront sensor with its corrector.
Fig. 11.2
Examples of Shack-Hartmann wavefront sensor spots . (a) Ideal spot pattern from an unaberrated eye shows uniform spacing between each spot. (b) Spot pattern resulting from astigmatism which shows nonuniform spacing in one direction. Overall, the spots are closer together along one diagonal when compared to the other diagonal. (c) Spot pattern with a large number of distorted spots present which could affect the performance of the AO corrector. Care should be taken to insure that high-quality spots are obtained for the AO correction
A practical understanding of these three components will help to troubleshoot challenging imaging sessions. For most users, the biggest gain in image quality will be realized making sure that the wavefront sensor is able to capture clean spots. Every AO correction begins with the wavefront sensor. If a clean spot pattern is not acquired, then the wavefront corrector will be unable to generate an accurate correction (Fig. 11.2c). In such a case, the resulting image with AO correction may or may not be better than the starting image without AO correction.
Basic Principles
There are some important concepts that should be taken into consideration before deciding to embark on a journey using AO in the clinic. These basic principles outline some of the important parameters behind successful AO imaging and help to set general expectations and lead to better overall success and satisfaction when using AO imaging.
- 1.
AO by itself is not an imaging technique, but rather, a module that must be combined with an underlying imaging platform. The main purpose of AO is to enhance the resolution of its underlying imaging platform. As such, it is critical that the underlying imaging platform be an instrument with very good performance. Close examination of published images with and without AO correction will reveal that it is possible to see some photoreceptor structure in a healthy subject even without AO [1, 47, 48].
Examples of imaging platforms with which AO has been combined include:
- (a)
- (b)
- (c)
- (d)
Other systems: AO has also been implemented on combined systems that can simultaneously perform scanning-based imaging and OCT [57–59]. There are also examples of AO being applied to instruments from microscopy, astronomy, and other fields. These illustrate the diversity of AO instruments that are now available.This chapter will focus primarily on the combination of hardware-based AO with scanning-based (“SLO”) systems.
- (a)
- 2.
AO retinal imaging works best for larger pupils. The improvement from using AO is more noticeable when imaging through a larger pupil as opposed to a smaller pupil. As a rule of thumb, the larger the pupil, and the shorter the axial length, the better the achievable resolution.
- 3.
AO retinal imaging works best for small fields-of-views (FOV). Although it would be desirable to have a large FOV across which cellular level resolution could be achievable, currently, it is only possible to have high-quality AO images when using smaller FOVs (approximately 1 mm × 1 mm or smaller). For studying large clinical lesions, multiple, overlapping images need to be acquired and then stitched together in order to generate an AO image on a larger portion of the retina.
- 4.
AO retinal imaging is currently incompatible with some patients. Unfortunately, not all patients are suitable for AO retinal imaging with the current technology standards. Although improvements in technology are helping to mitigate these issues, there are several key criteria that contribute strongly to the success rate of AO imaging:
- (a)
Tear film. Patients should have a healthy tear film. Dry eye or contact lens used immediately prior to imaging may affect the tear film and therefore the quality of AO retinal imaging. Patients should be advised not to wear contact lenses on the day of imaging.
- (b)
Media opacity. The presence of any cataract or other media opacities will diminish the quality of the AO retinal image that can be acquired. Unfortunately, cataract surgery may not improve the chances of success. Most AO retinal imaging instruments are designed to perform well only when imaging through the natural crystalline lens; the image quality when attempting to image pseudophakic eyes tends to be poor.
- (c)
Fixation. Extremely good fixation is highly desired. When using the small FOVs that are necessary for AO retinal imaging, any amount of unstable fixation will become readily apparent. If the eye motion is too large, then it may become difficult or impossible to perform averaging across multiple frames, which can diminish the final image quality.
- (d)
Other factors. There may be other unpredictable factors that can affect the success of AO retinal imaging. In some patients, refractive surgeries such as LASIK may generate new aberration profiles that are difficult for an AO instrument to correct (particularly when using larger pupil sizes and when there is a sharp change in the local corneal curvature due to the surgery). Sustained patient motivation, compliance, and cooperation are also essential as is a well-trained operator or team of operators. If a patient is unable to sit still without talking for sustained periods of time, while maintaining a steady fixation with their eyelids fully open, then they may not be the ideal candidate for AO retinal imaging.
Despite these challenges, there are many published examples of successful AO imaging in patients even when the conditions are not ideal. The combination of a high-quality AO instrument with a skilled AO operator or team and careful study design can maximize the probability for success.
- (a)
- 5.
AO retinal imaging is a time- and resource-intensive endeavor, with regard to data acquisition, processing, and interpretation. The length of an imaging session is entirely dependent on the number and type of videos that one wishes to acquire. Acquiring one image in each eye can be accomplished in a few minutes, but acquiring a large number of overlapping images in just one eye may take 2 h or more. However, in general, image acquisition is one small part of the overall time and resource involvement that is needed. Analyzing AO data is an especially time- and resource-intensive activity, which can be roughly divided into two main steps.
- (a)
Step 1: Convert a set of overlapping AO videos to a montage of averaged AO images. Each AO video will have eye-motion artifacts, which need to be corrected [60–62]. Once the eye motion has been corrected, averaging can be performed to enhance the image quality. Averaged images are then assembled into a larger montage to create a single, combined AO image. For one dataset from one AO imaging session, this step can take anywhere from hours to days.
- (b)
Step 2: Quantify AO metrics. Whereas the previous step generates a qualitative image that can be used for qualitative assessments, typically, most studies will need to have quantitative data from the AO images (Fig. 11.3). New metrics are currently being proposed and developed. Examples of existing quantitative metrics for AO images include photoreceptor density [64], photoreceptor spacing [41], photoreceptor reflectivity [65], foveal avascular zone size [11], capillary density [11, 66], capillary tortuosity [38], erythrocyte speed [6, 13, 67], leukocyte speed [68], capillary pulsatility [12], and others. The quantification of such metrics can take anywhere from days to weeks, per subject. An important prerequisite of quantification is identification or segmentation [63, 67–74], which may lead to grader-to-grader variations [75]. AO-based microperimetry may also provide functional metrics at the single-cone level [76, 77].
- (a)
Fig. 11.3
Examples of quantitative analysis. (a) Individual cone photoreceptors need to be identified for analysis. These cones, from Fig. 11.1a, were identified in a semiautomated manner as described in [63]. The extracted points can then be used to quantify metrics such as cone density or spacing. (b, c) The foveal avascular zone and parafoveal capillaries were extracted in a semiautomated manner as described in [11]. The extracted binary images can then be used to quantify metrics such as foveal avascular zone size or capillary density
Current Capabilities of AO Retinal Imaging
There is continuous development in the technical capabilities of AO instruments. Therefore, the list of items that can be observed using AO retinal imaging is constantly expanding. The following list highlights some of the capabilities that have been demonstrated using AO instruments. There are many options available for imaging, which are described below.
Although there are many possibilities, no single instrument can have every capability installed, due to differences in the design of the instrument. The list of features that are needed will depend on the specific needs of the clinic in which the AO instrument will be used.
Cone photoreceptors
Confocal images—By far, the most popular AO image in the literature is the confocal image of cone photoreceptors , which is generated from light that is reflected back from the photoreceptors (Fig. 11.1a). Since light has to enter and exit each photoreceptor and since individual photoreceptors are thought to exhibit wave-guiding properties [78], it is thought that the confocal image of photoreceptors represents the wave-guiding status of the cell.
The retinal eccentricity needs to be considered when imaging the photoreceptors, since the cone photoreceptors vary in size across the retina, with the smallest cones in the fovea and the largest cones in the periphery [79]. Due to their size, larger cones are more readily imaged by AO instruments, when compared to foveal cones.
Rod photoreceptors
Vasculature
Confocal images—The retinal vasculature can be imaged using confocal imaging alongside photoreceptors [80]. Since the inner retinal vessels reside in layers that are anterior to the photoreceptors, confocal images of the vasculature will either be in or out of focus. When the vasculature is in focus, a bright central reflex can be seen (Fig. 11.1b); when it is out of focus, a shadow-like image of the vessel is generated instead (Fig. 11.1a). The advantage of acquiring in the out-of-focus regime is that it is easier to visualize the motion of leukocytes [80]. One can also use image processing methods such as motion contrast enhancement to generate vascular perfusion maps (Fig. 11.1c, f) [11]. Motion contrast enhancement can also be used to enhance the contrast of individual blood constituents such as leukocytes and plasma gaps [12, 68].
Non-confocal reflectance images—When the pinhole that is placed in front of the detector is offset to one side, the type of light that is collected is changed. Light which is multiply scattered can be captured through a single offset pinhole (“offset pinhole”) [7]. A similar effect can be achieved by replacing the pinhole with a knife edge (“knife edge”) [81] . Split detection can also be used to acquire images of the vasculature [81]. These non-confocal reflectance methods can be used to enhance different features of the vasculature and are especially notable for imaging the vascular wall [8].
Fluorescence angiography using AO—AO instruments can also be outfitted with fluorescence imaging capabilities [9, 10, 18] to study clinical lesions such as microaneurysms [82]. However, when using injected dyes such as fluorescein with AO, this modality does not provide the capability to image the vessel wall [83].
Given the large number of possibilities for vascular imaging, the use of multiple approaches may provide complementary information.
Nerve Fiber Layer
Inner Retina
The inner retina is mostly transparent when imaged using AO instruments in confocal mode, with the exception of vessels and the nerve fiber layer. However, under certain conditions, it may be possible to visualize hyper-reflective lesions or features within the inner retina [26]. The use of non-confocal light also enables visualization of retinal ganglion cell soma [27].
Optic Nerve Head
Confocal images—The three dimensionality of the optic nerve head makes it a challenging structure to image. However, it is possible to image the lamina cribrosa in human subjects [29].
RPE
Confocal images—The RPE cannot be seen using confocal mode in most subjects. However, under special circumstances, when the photoreceptors are lost, it is possible to see the RPE using confocal mode [15].
Dark field—When all direct backscattered light is blocked and when some of the remaining light is used for detection, this type of modality is called dark field imaging. Dark field imaging was recently used to capture images of the RPE [17]. Unfortunately, dark field RPE imaging is not robust. Currently, dark field cannot be used to reliably image the RPE in patients.
Fluorescence—Autofluorescence can be combined with AO imaging to image the RPE [14, 20, 21]. However, due to light safety concerns [84], AO-based autofluorescence using visible wavelengths of light has not been routinely performed, and for most AO instruments, RPE imaging remains elusive. One of the main challenges is that the autofluorescence signal is very weak, even at light levels that are very close to the maximum permissible exposure. In the original implementations of visible-wavelength-based AO autofluorescence, 1000–1700 frames were summed in order to generate a single image of the RPE [14]. This weak signal also poses a challenge for setting the appropriate focal plane during imaging. With advanced focusing methods [16] and real-time eye tracking [62] it may be possible to image the RPE more robustly using safe light levels in patients. Finally, the use of nearinfrared wavelengths for AO autofluorescence [20–22] as well as imaging of indocyanine green using AO [18] are additional possibilities for visualizing the RPE.
Potential Use of AO for Ocular Inflammatory Diseases
Given that the ideal AO patient has large pupils with clear media, it is not surprising that there are relatively fewer examples of AO applied to uveitis. The capabilities of AO have largely revolved around photoreceptors and the vasculature.
Photoreceptors—It is often unclear what the mechanism for vision loss is in uveitis, for example, whether it is due to dysfunction in the photoreceptor , inner retina, or RPE. AO retinal imaging provides an opportunity to start to explore the cellular basis of vision loss and can in particular be used to investigate the status of photoreceptor cells. To this end, one of the earliest examples of a custom-built AO instrument being used to study a patient with uveitis is a case series with four women with acute zonal occult outer retinopathy (AZOOR) [85]. Mkrtchyan and colleagues used AO to investigate the effect of AZOOR on cone photoreceptors and found evidence of focal cone loss in most patients, suggesting cone involvement. However, in one patient, they found an intact cone photoreceptor mosaic but evidence of inner retinal dysfunction. Therefore, both photoreceptors and inner retinal defects remain plausible as causes of vision loss in AZOOR. Recently, Nakao and colleagues reported a single case of AZOOR in which spontaneous remission suggested the possibility of reversible cone damage [86].
AO imaging may be useful for evaluating photoreceptor loss near chorioretinal lesions [87]. Direct evaluation of cellular structure within the eye after resolution of acute cases of uveitis may lead to new insights about how individual cells in the retina are able to recover from stress or disease.
Vasculature—Given the plethora of AO imaging strategies for the vasculature, AO vascular imaging could be particularly promising for investigations of inflammatory eye disease. Errera and colleagues used a flood-illuminated AO instrument to investigate three cases of retinal vasculitis and found evidence of focal perivascular opacification and signs of focal narrowing within venous lumen on AO images [88]. They suggest that AO might be useful for the early detection of subclinical perivascular infiltrates [89]. Offset pinhole, knife edge, and split detection AO imaging are all able to reveal details about the vessel walls which could be particularly interesting in cases of vasculitis. In particular, imaging the vessel wall could reveal evidence of infiltrating immune cells and serve as a biomarker for inflammation. Future advancements in AO technology could lead to improved single-blood-cell imaging capabilities. This could lead to the ability to noninvasively monitor the actual infiltrating immune cells.
RPE and choroid—Currently, the choroid is inaccessible in most patients by AO imaging, due to obscuration by the RPE . However, it may be possible to see lesions that affect the RPE and Choroid. Mrejen and colleagues report the presence of dark spots on AO imaging in a case report of acute posterior multifocal placoid pigment epitheliopathy (APMPPE) that seem to be distributed along choroidal vessels [90].
Recently, the combination of AO with fluorescein angiography has demonstrated the possibility for fluorescence imaging in patients. In the future, continued development of AO-enhanced indocyanine green imaging may provide a novel opportunity for visualizing blood flow in the choroid with greater detail [91].
Limitations of AO Imaging
Although AO is a very promising technology, there are some important limitations.
Clinical utility—Although AO adds great value to a clinic in terms of its research capabilities, one of the current limitations of AO is that it is not used in diagnosis or in the clinical management of disease. In the future, early detection of subclinical lesions or more precise monitoring of recovery after treatment may be important for establishing AO as an everyday tool in the clinic. From a translational perspective, AO may provide insight to pathology.
Interpretation of AO images—In diseased eyes, it is often very difficult to interpret the AO images. Although one may be tempted to interpret bright cones as healthy cones, there is no clear relationship between reflectivity and cone health [65]. Some cone photoreceptors may appear dark on one day and bright on another, while others may not change intensity at all. Often, the confocal AO image of an affected area is dark, with a few residual spots scattered throughout. Interpretation of these dark areas is difficult—one cannot conclude that a structure is absent just because it is not seen on AO—the structure could also be present but in a state that precludes it from imaging. In support of this notion, using AO, Wang and colleagues found evidence of cone function even in areas that were considered to be dark in patients with macular telangiectasia [77].
Stay updated, free articles. Join our Telegram channel
Full access? Get Clinical Tree
