If you don't remember your password, you can reset it by entering your email address and clicking the Reset Password button. You will then receive an email that contains a secure link for resetting your password
If the address matches a valid account an email will be sent to __email__ with instructions for resetting your password
Department of Electrical Engineering and Computer Science, and Research Laboratory of Electronics, Massachusetts Institute of Technology, Cambridge, Massachusetts
Department of Electrical Engineering and Computer Science, and Research Laboratory of Electronics, Massachusetts Institute of Technology, Cambridge, Massachusetts
Department of Electrical Engineering and Computer Science, and Research Laboratory of Electronics, Massachusetts Institute of Technology, Cambridge, MassachusettsHealth Sciences and Technology, Harvard and Massachusetts Institute of Technology, Cambridge, Massachusetts
Department of Electrical Engineering and Computer Science, and Research Laboratory of Electronics, Massachusetts Institute of Technology, Cambridge, Massachusetts
Correspondence: James G. Fujimoto, PhD, Department of Electrical Engineering and Computer Science, and Research Laboratory of Electronics, Massachusetts Institute of Technology, 50 Vassar Street, Cambridge, MA 02139-4309.
Department of Electrical Engineering and Computer Science, and Research Laboratory of Electronics, Massachusetts Institute of Technology, Cambridge, Massachusetts
Ultrahigh resolution spectral domain-OCT (UHR SD-OCT) enables in vivo visualization of micrometric structural markers which differentially associate with normal aging versus age-related macular degeneration (AMD). This study explores the hypothesis that UHR SD-OCT can detect and quantify sub-retinal pigment epithelium (RPE) deposits in early AMD, separating AMD pathology from normal aging.
Design
Prospective cross-sectional study.
Participants
A total of 53 nonexudative (dry) AMD eyes from 39 patients, and 63 normal eyes from 39 subjects.
Methods
Clinical UHR SD-OCT scans were performed using a high-density protocol. Exemplary high-resolution histology and transmission electron microscopy images were obtained from archive donor eyes. Three trained readers evaluated and labeled outer retina morphological features, including the appearance of a hyporeflective split within the RPE–RPE basal lamina (RPE-BL)–Bruch’s membrane (BrM) complex on UHR brightness (B)-scans. A semi-automatic segmentation algorithm measured the thickness of the RPE-BL-BrM split/hyporeflective band.
Main Outcome Measures
Qualitative description of outer retinal morphological changes on UHR SD-OCT B-scans; the proportion of the RPE-BL-BrM complex with visible split (%) and the thickness of the resulting hyporeflective band (μm).
Results
In young normal eyes, UHR SD-OCT consistently revealed an RPE-BL-BrM split/hyporeflective band. Its visibility and thickness were less in eyes of advanced age. However, the split/hyporeflective band was again visible in early AMD eyes. Both qualitative reading and quantitative thickness measurements showed significantly elevated visibility and thickness of the RPE-BL-BrM split/hyporeflective in early AMD eyes compared to age-matched controls.
Conclusions
Our imaging results strongly support the hypothesis that appearance of the RPE-BL-BrM split/hyporeflective band in older subjects is dominated by the BL deposit, an indicator of early AMD well known from histology. Ultrahigh resolution SD-OCT can be used to investigate physiological aging as well as early AMD pathology in clinical imaging studies. Developing quantifiable markers associated with disease pathogenesis and progression can facilitate drug discovery, as well as reduce clinical trial times.
Financial Disclosure(s)
Proprietary or commercial disclosure may be found after the references.
Age-related macular degeneration affects multiple structures in the retina, particularly the retinal pigment epithelium (RPE), its basal lamina (BL), and Bruch’s membrane (BrM).
The name RPE-BL-BrM has been proposed to accommodate the appearance in AMD of new layers and sub-RPE-BL space, which contains drusen, type 1 macular neovascularization, and sequelae.
High resolution histology and transmission electron microscopy (TEM) have identified 2 distinctive basal deposits, basal laminar deposit (BLamD), and basal linear deposit (BLinD).
the pathognomonic extracellular deposit of AMD and its major risk factor. The relative roles of BLamD and BLinD in AMD progression have been challenging to determine in vivo, because standard clinical imaging modalities have insufficient resolution and contrast to detect them.
Using a prototype ultrahigh resolution spectral domain-OCT (UHR SD-OCT) instrument to image normal adult subjects and AMD patients,
we found that the RPE-BL-BrM complex, seen in commercial OCT as a single hyperreflective band, can be further resolved as 3 distinctive bands in healthy young adult eyes and in eyes with early AMD. The feature appears as a “split”, in which 2 hyperreflective bands are separated by a hyporeflective band. These 3 OCT bands become progressively nonuniform with normal aging, and the split is not resolvable in most normal eyes by the sixth to seventh decade of age. The visibility of an RPE-BL-BrM hyporeflective split strongly associates with age and AMD diagnosis, which suggests its potential to differentiate aging and AMD pathology.
Comparing UHR SD-OCT brightness (B)-scans from early AMD eyes (N = 19) and normal adult eyes across the life span (N = 63), and guided by histology and ultrastructural findings, this study explores the hypothesis that the RPE-BL-BrM split/hyporeflective band corresponds to the accumulation of BLamD in eyes diagnosed with AMD. While OCT imaging lacks molecular specificity to distinguish between BLamD and BLinD, BLamD is on average 3× thicker than BLinD, and therefore dominates measures of total thickness.
Our results also suggest that the RPE-BL-BrM split/hyporeflective band in young normal eyes may originate from the basal portion of the RPE, which includes contributions from its basal infoldings.
Based on these observations, we believe that UHR SD-OCT can visualize and measure outer retina alterations at micrometer scale as a marker for normal aging vs. early AMD. Ultrahigh resolution SD-OCT can provide a clinical imaging tool enabling in vivo and longitudinal assessment of key AMD pathological features, which are typically accessible only ex vivo. Ultrahigh resolution SD-OCT imaging of the RPE-BL-BrM promises to elucidate normal aging as well as AMD pathogenesis and provide a potential marker for drug development and therapeutic trials.
Methods
Study Design
This cross-sectional, observational study investigated whether UHR SD-OCT can detect and quantify outer retinal changes for differentiating aging and AMD pathology. Participants were prospectively enrolled from the retina clinic at the New England Eye Center. Normal diagnosis is based on medical history and normal dilated fundus examination by the recruiting ophthalmologist. Age-related macular degeneration diagnosis and severity are based on the clinical classification definition using color fundus photography assessed by a retinal specialist (N.K.W., C.R.B., A.W., and M.C.L.).
The study protocol was approved by institutional review boards at the Tufts Medical Center and Massachusetts Institute of Technology. All study procedures adhere to the tenets of the Declaration of Helsinki and comply with the Health Insurance Portability and Accountability Act of 1966. Written informed consent was obtained after explaining the purpose, procedures, benefits, and risks of the study.
The normal cohort includes subjects with no known retinal pathology or history of pathology and normal examination. A subset of eyes included in the normal cohort was from subjects who are followed in the clinic for a retinal pathology in the contralateral eye that is not known to involve the macula of the normal study eye. The AMD cohort includes both unilateral and bilateral nonexudative (dry) AMD pathology. The fellow eyes of exudative AMD were included if there was no evidence of any neovascularization in the study eye. Exclusion criteria include the presence of neovascular activity, evidence of ocular pathology other than AMD, and high refractive error/significant opacity that prevents satisfactory OCT imaging.
Histological Preparation and Imaging of Donor Eyes
Exemplary high-resolution histology and ultrastructure images were obtained from archived human donor eye specimens. Detailed procedures for fixation and staining are documented in previous publications.
Briefly, donor eyes were preserved within 4 hours of death by immersion (after removal of the cornea) in 0.1 mol/L phosphate-buffered 1% paraformaldehyde and 2.5% glutaraldehyde. Full thickness tissue punches containing the fovea were post-fixed with 2% osmium and embedded in epoxy resin. For light microscopic overview, tissues were sectioned at 0.8 μm thickness, stained with 2% toluidine blue in 2% sodium borate, and imaged with a slide scanning system and 60 × oil objective (numerical aperture 1.4; VSI 120, Olympus). For TEM, gold sections (nominally 90 nm thick) were mounted on grids and post-stained with mixed lanthanides (Formvar Carbon Support Film on Specimen Grid and UranyLess, both from Electron Microscopy Sciences). Transmission electron microscopy images were acquired at original magnifications of up to 6500× (Tecnai 120kv TEM, FEI; BioSprint 29 Megapixel CCD camera, AMT). Images were adjusted to maximize the intensity histogram for contrast and white balance (Photoshop, Adobe). To delineate the anatomical landmarks in aged and AMD eyes, color overlays were drawn over electron micrographs using a pen display (Cintiq, Wacom) and graphics software (Illustrator, Adobe).
UHR Prototype OCT Instrument and OCT Imaging
The design and specifications of the UHR SD-OCT prototype instrument are documented in our previous publication.
Briefly, the UHR-OCT achieves 2.7 μm axial resolution and 128 kHz amplitude (A)-scan rate, representing ∼1.5 to 2 times improvements over commercial instruments. We designed a high density (HD) raster scan protocol, using 9-mm long horizontal B-scans with 1800 A-scans (5 μm A-scan spacing), imaging a 6-mm vertical field with 241 B-scans (25 μm spacing between B-scans). This protocol optimizes visualization of fine outer retina features and detection of small drusen, while maintaining an acceptable total scan time (∼3.8 seconds).
We designed the acquisition, processing, and analysis methods in order to reduce artifacts and potential interpretation errors. Most commercial instruments average repeated B-scans to increase visibility of retinal features. However, small axial or transverse eye motion during or between B-scans can blur micrometric features and cause image interpretation errors. Therefore, to increase feature visibility, we performed HD B-scans with large numbers of A-scans and then analyzed individual HD B-scans without image averaging. OCT images were read by displaying the OCT signal using a linear scale instead of a logarithmic scale; the latter is used in most commercial instruments. Logarithmic display scale compresses dynamic range, enabling features with weak signals, such as retinal nuclear layers, to be seen in the same display as the RPE, which is strongly scattering. However, the compression also broadens both the axial and transverse extent of small features, causing a loss of resolution. Linear display preserves resolution, but parameters such as black and white levels must be adjusted in order to optimize visualization of features of interest, analogous to viewing X-ray computed tomography. Images were read in electronic form using custom software where readers could adjust display parameters (e.g., the OCT levels are scaled to fill the full display range of the monitor with minimum saturation) as well as pan and zoom to evaluate outer retinal features, similar to image viewers used in radiology.
The quality of the OCT images was benchmarked using contrast-to-noise ratio. We calculated contrast-to-noise ratio using the OCT amplitude difference between median signal level within the RPE-BL-BrM band and the background, normalized by the noise standard deviation. Contrast-to-noise ratio is analogous to signal strength index or quality index reported by commercial OCT instruments.
In a subset of normal subjects, we performed OCT angiography using the same UHR SD-OCT instrument. The OCT angiography scan protocol acquired 5 repeated B-scans, with 400 A-scans per B-scan and 5 × 400 B-scans in a 3 × 3 mm2 macular field (scan time ∼7.8 seconds). Vascular contrast was generated using amplitude decorrelation
and overlaid with structural image. The intrinsically co-registered and depth resolved OCT and OCT angiography provided additional insight into the anatomical origin of the outer retinal bands, enabling a more precise determination of the BrM position.
Qualitative Assessment of RPE-BL-BrM Split/Hyporeflective Band Visibility in Normal Aging and Dry AMD Eyes
We chose to have readers view and interpret the images as a first step in order to better understand UHR OCT features. To establish a baseline for interpreting pathological features, multiple known features in young normal eyes were identified, measured, and compared to histology of different eyes. For each eye, 5 B-scans from the HD raster protocol were read and labeled: the center horizontal B-scan across the fovea centralis, and horizontal B-scans 1-mm and 2-mm superior and inferior. Each B-scan was displayed in linear gray scale using a custom MATLAB (R2017a, MathWorks) program, which allowed the reader to adjust display dynamic range and axially enlarge the image to best evaluate outer retinal features.
As a next step, 3 expert readers (S.C., O.A.Q., and Y.H.) independently assessed the visibility of the RPE-BL-BrM split/hyporeflective band, because this feature is readily detected as a binary determination and has good interobserver consensus. The readers labeled the transverse regions within each B-scan: (1) where the RPE-BL-BrM split/hyporeflective band is visible; (2) where AMD-specific lesions, including drusen, subretinal drusenoid deposits and geographic atrophy (GA) are present; and (3) where low OCT signal prevents reliable discrimination; that is, due to excessive vascular shadowing or vignetting. The readers were masked to the age and diagnosis. However, AMD lesions are visible in the B-scans so the readers may be aware which eyes were diagnosed with AMD.
The readers read a training set consisting of representative cases (1 young, 2 mid-aged, 2 older age, 1 early AMD, 3 intermediate AMD, and 1 late dry AMD with GA) before reading the full data sets. Then the readers jointly reviewed each other’s labeling on the training set, as well as consulted published OCT literature on AMD lesions, to reach a consensus. Consensus was reached when the labeling agreement is > 80% on A-scan by A-scan basis. Finally, after a washout period, each reader was asked to independently label all datasets, including relabel of data used in training with previous labels removed.
We report the visibility of the RPE-BL-BrM split using a single percentage value, which is calculated as the ratio between the number of A-scans where the split is resolved (Label [1]) and the number of total A-scans in the specified topographical region, excluding those labeled as known AMD lesions and areas of low signal (Label [2] and [3]). To mitigate slight variations in imaging field, only the center 6 mm range of each B-scan is included. The averaged results from the 3 readers are reported. Agreement between the 3 independent readers is assessed using intraclass correlation (2-way mixed, single measures, and absolute agreement) and visualized on Bland–Altman plots.
To assess the global trend, all 5 B-scans are used. To investigate topographic distribution of features, 2 eccentric regions are defined using a modified ETDRS grid. The central macula (fovea and parafovea) is defined as eccentricity ≤ 1.5 mm of the foveal center (i.e., central and inner ETDRS subfields), while the perifovea is defined as between 1.5 mm and 3.0 mm eccentricity (i.e., outer ETDRS subfields).
Quantitative Assessment of RPE-BL-BrM Split/Hyporeflective Band Visibility in Normal Aging and dry AMD Eyes
We performed quantitative measurement of the RPE-BL-BrM split/hyporeflective band using a combination of reader feature detection and software refinement. The software algorithm refines the reader manual tracing of layer boundaries using a set criterion based on OCT signal amplitude, aiming at minimizing a priori bias. We chose not to develop fully automated segmentation software for these initial assessments because we wanted traceability of the measurements to features which are visible and interpretable by readers.
First, the operator manually traced (segmented) the anterior and posterior boundaries of the RPE and the centerline of BrM, on axially enlarged B-scans. In locations where the split was not observed, the 2 traces merged, both corresponding to the posterior boundary of the RPE-BL-BrM complex. These locations are later detected and recorded.
The software algorithm refines the manual segmentation by first searching for the OCT signal peak near the BrM tracing line. Then the software searches for the half-maximum position as the posterior boundary of the RPE-BL. If the 2 segmentation lines merge, or the separation between them is < 3.2 μm (3.5 OCT pixels), the RPE-BL-BrM split is considered not resolved. Otherwise, the thickness of the hyporeflective band is calculated as the distance between the 2 software segmentation lines. The hyporeflective band cross-sectional area is calculated by integrating over the transverse extent where the split is observed. Consistent with qualitative reader studies, both the band thickness and area are calculated using the center 6 mm range of each B-scan only.
An upper threshold of 14 μm, approximately the RPE cell body height reported in histology,
is used to exclude potential drusen areas. However, it should be noted that BLamD, considered as the dominating contributor of the RPE-BL-BrM hyporeflective band in AMD subjects, can reach > 20 μm in more severe pathology. Thus, our reported values should be considered as a lower bound for AMD eyes. This consideration does not invalidate the study conclusions.
Statistical Analysis
Statistical analysis was performed using Microsoft Excel (Professional Plus 2019, Microsoft) and the Statistics Toolbox in MATLAB (R2017a, MathWorks). We performed Kruskal–Wallis one-way analysis of variance to test if any of the age/diagnosis category has significant different OCT imaging quality (i.e., contrast-to-noise ratio). Wilcoxon rank sum U test was used to compare between early AMD and age matched normal eyes. A P value of < 0.05 was considered statistically significant. Least squares linear regression was used to assess correlation between the RPE-BL-BrM hyporeflective band thickness and area with respect to age in the normal cohort. We did not adjust for potential correlations between eyes included from the same subject, partly because of the small sample size, and because the purpose of the tests is to associate UHR OCT phenotype with clinical diagnosis rather than for disease detection.
Results
Participants
The normal cohort consists of a total of 63 eyes from 39 subjects. The AMD cohort has 53 eyes from 39 subjects. Participant demographics are summarized in Table 1. Because there was limited number of older participants at our recruitment site, we included 3 subjects who had diabetes mellitus, but did not have clinical retinopathy. This is consistent with the stated inclusion/exclusion criteria of the protocol. There is currently no evidence that the inclusion of diabetic patients without diabetic retinopathy would affect our imaging results.
Table 1Demographics and OCT Imaging Quality of Study Subjects and Eyes
One subject each from the respective group has diabetes, but does not have clinical diabetic retinopathy as confirmed by color fundus photography. Under Kruskal–Wallis one-way analysis of variance, no statistically significant difference was found in OCT CNR between age or diagnostic categories (P = 0.176).
One subject each from the respective group has diabetes, but does not have clinical diabetic retinopathy as confirmed by color fundus photography. Under Kruskal–Wallis one-way analysis of variance, no statistically significant difference was found in OCT CNR between age or diagnostic categories (P = 0.176).
One subject each from the respective group has diabetes, but does not have clinical diabetic retinopathy as confirmed by color fundus photography. Under Kruskal–Wallis one-way analysis of variance, no statistically significant difference was found in OCT CNR between age or diagnostic categories (P = 0.176).
∗ One subject each from the respective group has diabetes, but does not have clinical diabetic retinopathy as confirmed by color fundus photography. Under Kruskal–Wallis one-way analysis of variance, no statistically significant difference was found in OCT CNR between age or diagnostic categories (P = 0.176).
Degenerative Changes of Outer Retinal Ultrastructure in Aging and Early AMD Eyes
Using high-resolution histology and TEM from 1 young normal eye and 1 older eye with early AMD, we show exemplary changes associated with aging and AMD (Fig 1). Photoreceptor outer segments, RPE apical processes, RPE cell bodies, BrM, and choriocapillaris (ChC) align in horizontal layers. In the context of AMD, BrM is best defined by its middle 3 components (i.e., inner collagenous, elastic, and outer collagenous layers).
The 3-layer BrM is bordered anteriorly by the RPE-BL, and posteriorly by the basal lamina of the ChC. In the RPE cells, melanosomes are optically dark and electron dense, and cluster in the apical cell body, notably in the older eye where they stand vertically. Lipofuscin and melanolipofuscin, which are also electron dense in this preparation, occupy the apical three fourths of the cell body. The cell nucleus is situated toward the posterior cell body, surrounded by numerous mitochondria. These are closely opposed to the basal infoldings, which are prominent in the younger eye. These membrane specializations for absorption and transport distribute in an elaborate labyrinth, extending 1.5 to 2 μm above the RPE-BL.
Figure 1Representative histology and ultrastructure images showing aging and early age-related macular degeneration (AMD) changes. Retinal pigment epithelium (RPE), RPE basal lamina (RPE-BL), Bruch’s membrane (BrM) and choriocapillaris (ChC) complex under the fovea of the left eye of a 32-year-old White female and the right eye of a 65-year-old White male with early AMD. Neurosensory retinas are artifactually detached and not shown. A, B, Toluidine blue-stained epoxy-resin sections. C, D, Transmission electron micrographs. E, F, Colorization of panels (C, D) shows tissue components (color key at bottom). A, C, and E, In a young adult, the 3-layer BrM consists of inner collagenous layer (ICL), elastic layer (EL), and outer collagenous layer (OCL). It is flanked by the basal laminas, of the RPE above and of the ChC below. RPE cell bodies extend processes between the basal portions of the cell body toward the basal lamina. The magnified inset shows mitochondria (m) at close proximity to an infolding (arrowheads). B, D, and F, The eye of an older adult exhibits the same layers of BrM with marked loss of RPE basal infoldings and appearance of basal laminar deposit (BLamD) in the same plane. The RPE cell body is indented by the deposit. Soft drusen material ("membranous debris" or "lipoprotein-derived debris") is found in the sub-RPE-BL space as basal linear deposit (BLinD) or crossing BLamD, where it may congregate as a basal mound (asterisk). Magnified inset shows BLamD and BLinD on either side of the RPE BL (blue) occupied by lipoprotein derived debris (arrowhead) and type VI collagen with a banded appearance. Note how the OCL of Bruch’s membrane extends further between the capillaries in the older adult (B, D, and F), compared to the young adult (A, C, and E). The ICL and EL remain flat between the capillaries. AP = apical process; L = lipofuscin; ML = melanolipofuscin; N = nucleus; OS = outer segment. Scale bars: Panels (A, B) = 10 μm; Panels (C–F) = 2 μm.
In early AMD (Fig 1, right column), the basal infoldings have disappeared and new layers have appeared. Basal laminar deposit is similar in electron density, texture, and composition to the RPE-BL
Thick, continuous layers of BLamD may act as a transport barrier between the RPE and the underlying ChC, contributing to the formation of high-risk soft drusen, BLinD, or both.
In this example a continuous BLamD, containing amorphous material and a notable component of fibrous long spacing collagen (type VI), can be seen atop the BrM, in the same plane previously occupied by the basal infoldings, having either replaced or incorporated these membranes. The RPE cell body is indented anteriorly by this deposit. A second new layer in this older eye is BLinD. Like soft drusen, BLinD is lipid-rich, biomechanically unstable, and molecularly proinflammatory, thus conferring risk for developing macular neovascularization.
Soft drusen material (membranous/lipoprotein-derived debris) can be found in both the sub RPE-BL space, in the form of BLinD, as well as crossing BLamD.
UHR SD-OCT Imaging Improves Visualization of Outer Retinal Bands
The different optical properties of cellular components and interfaces contribute to the alternating hyperreflective and hyporeflective bands in OCT B-scans (Fig 2). Ultrahigh resolution SD-OCT visualization of the external limiting membrane (ELM), ellipsoid zone (EZ) of the photoreceptors, and cone interdigitation with the RPE apical processes (cone interdigitation zone [CIZ]) is consistent with prior literature.
The anatomical correlates of the EZ and CIZ bands are still being investigated, where researchers have proposed alternative interpretations as the photoreceptor inner segment/outer segment junction (IS/OS) and cone photoreceptor outer segment tips (COST), respectively.
Figure 2Histology and clinical OCT imaging of retinal layers. A, High-resolution toluidine blue histology section around the fovea of a 63-year-old female with a normal macula. In life, photoreceptor outer segments are straight and attached to the retinal pigment epithelium. Zigzag bending as in this specimen (compaction) often occurs in postmortem specimens even if attached. B, Ultrahigh resolution spectral domain-OCT (SD-OCT) B-scan from a 27-year-old White female. The left-hand side is displayed in logarithmic grayscale. OCT bands are labeled according to consensus nomenclature. Parentheses indicate alternative anatomical interpretation. The right-hand side is displayed in linear grayscale, which enhances visualization of thin features, but trades off dynamic range. Outer retinal bands are better visualized using UHR SD-OCT are sequentially labeled #i to #vii, since they are not included in the consensus nomenclature. AP = apical processes of RPE; ChC = choriocapillaris; CIZ = cone interdigitation zone with RPE; COST = cone outer segment tips; ELM = external limiting membrane; EZ = ellipsoid zone of the photoreceptors; GCL = ganglion cell layer; HFL = Henle fiber layer; ILM = inner limiting membrane; INL = inner nuclear layer; IPL = inner plexiform layer; IS/OS = photoreceptor inner segment/outer segment junction; ISel = XXX; ISmy = myoid zone of the photoreceptors; NFL = nerve fiber layer; ONL = outer nuclear layer; ONLc = cone nuclei in ONL; ONLr = rod nuclei in ONL; OPL = outer plexiform layer; OS = outer segments of the photoreceptors; RPE = retinal pigment epithelium; RPE-BL-BrM = RPE basal laminar-Bruch’s membrane complex. Scale bars = 200 μm.
Ultrahigh resolution SD-OCT enhances visualization of moderately reflective and hyporeflective bands of the outer retina, especially when the axial dimension is expanded and the B-scan is displayed in linear gray scale. These outer retinal bands are sequentially labeled from anterior to posterior as: a moderately reflective band immediately anterior to EZ (IS/OS) (#i); a moderately reflective band immediately anterior to CIZ (COST) (#ii), with higher visibility eccentric to the fovea; a thin hyporeflective band just beneath CIZ (COST) (#iii); a moderately reflective band anterior to the RPE-BL-BrM complex, visible only at > ∼0.5 mm eccentricity (#iv) (Fig 2B). We chose to label the features using numbers rather than by anatomical or histological structure because the interpretation is still under investigation.
A split RPE-BL-BrM complex can be consistently observed in young normal eyes. Consequently, in place of the single hyperreflective RPE-BL-BrM complex, a thick hyperreflective band (#v), a thin hypo-reflective band (#vi), and a thin hyper-reflective band (#vii), can be clearly resolved. Correlating OCT band positions and thicknesses with published histological measurements
provided strong supporting evidence that band #v corresponds to the apical RPE cell body, and band #vii to BrM (see Supplementary Text, Figs S3 and S4, and Table S2).
An avascular RPE-BL-BrM split was previously observed as an incidental finding in normal eyes using prototype OCT instruments, including adaptive optics OCT
Volumetric retinal imaging with ultrahigh-resolution spectral-domain optical coherence tomography and adaptive optics using two broadband light sources.
However, to our knowledge, this is the first time the RPE-BL-BrM split is investigated in a prospective cross-sectional study on normal and early AMD eyes.
Alterations of OCT Outer Retinal Bands in Normal Aging
The OCT outer retinal bands demonstrate subtle, but consistent, changes during normal aging. We describe observations using UHR SD-OCT B-scans from 1 young, 1 middle aged, and 1 older normal eye, which are representative of the larger cohort.
In the young normal eye (Fig 5A–C , 29-year-old Asian female), all hyperreflective outer retinal bands are clearly resolved, with clear boundaries visible across almost the entire 9 mm wide field of view. The visibility of CIZ (COST) decreases at larger eccentricities, corresponding to the decrease in cone density. The hyporeflective band #vi has a small interruption beneath the fovea center but is otherwise continuous throughout. The hyperreflective band #vii is thin and smooth, situated immediately anterior to the ChC/choroid.
Figure 5Ultrahigh resolution spectral domain-OCT in normal aging. A–C, A 29-year-old Asian female. En face OCT projection, center brightness (B)-scan in logarithmic grayscale, and flattened and axially expanded B-scan in linear grayscale. The retinal pigment epithelium-basal laminar-Bruch’s membrane complex (RPE-BL-BrM) split is evident. Hyperreflective bands #v and #vii, as well as the hyporeflective band #vi, are resolved across almost the entire field. D–F, A 57-year-old White female. Bands #v, #vi, and #vii have reduced visibility, but remain resolved temporally and near the optic nerve head (ONH) (yellow arrows). G–I, A 73-year-old White female. The RPE-BL-BrM split/hyporeflective band #vi is generally not visible, but a ∼1.2 mm segment can be resolved immediately bordering the ONH. Isolated, patchy RPE-BL-BrM splits are highlighted (yellow arrow). Scale bar in H applies to B and E, indicating twofold expansion of the axial dimension. Scale bar in I applies to C and F, indicating 10-fold expansion of the axial dimension.
In the middle aged eye (Fig 5D–F, 57-year-old White female), the distinction and resolvability of ELM, EZ (IS/OS), and CIZ (COST) remain, although ELM has a thickened and irregular appearance. Most notably, the visibility of the hyporeflective band #vi is reduced. The band #vi is not clearly resolved underneath the fovea and appears patchy on the temporal side. The visibility is better on the nasal side near the optical nerve head.
Finally, older eyes showed progressive thickening and nonuniformity of outer retinal bands, and the hyporeflective band #vi visibility is greatly reduced. In an exemplary eye (Fig 5G–I, 73-year-old White female), band #vi is only visible in 1 patchy region on the temporal side and a small ∼1.2 mm wide segment bordering the optical nerve head.
Normal aging of the eye often reduces OCT signal and causes aberrations. However, these changes cannot fully explain the observed alterations. Specifically, the inner retinal layers, including the inner limiting membrane and the ganglion cell layer, were well resolved, indicating the OCT imaging quality was maintained. Thus, it is likely the observed changes in OCT morphology reflect true anatomical alterations with aging.
Alterations of OCT Outer Retinal Bands in Nonexudative AMD
To understand how the outer retinal bands are altered in AMD, we show representative cases of nonexudative (dry) AMD eyes with progressive disease severity. In a representative early AMD eye (Fig 6A–C , 71-year-old White male), the most noteworthy feature is a pronounced hyporeflective band #vi, which can be clearly resolved between the hyperreflective bands #v and #vii. This high visibility is present despite the older age. The B-scan also shows irregular elevation of EZ (IS/OS) and diminished contrast of the CIZ (COST). Interruptions, irregular thickening, and changes in intensities of the OCT outer retinal bands are common findings in our AMD cohort.
Figure 6Ultrahigh resolution spectral domain-OCT in nonexudative (dry) age-related macular degeneration (AMD). A–C, Early AMD eye from a 71-year-old White male. En face OCT projection, center brightness (B)-scan in logarithmic grayscale, and flattened and axially expanded B-scan in linear grayscale of an early AMD eye. The retinal pigment epithelium-basal laminar-Bruch’s membrane complex (RPE-BL-BrM) split/hyporeflective band #vi is well resolved despite the older age. Hyperreflective outer retinal bands have irregular elevations, as well as display reduced OCT intensity, nonuniform OCT intensity, or both. D–F, Early AMD eye with subretinal druseonid deposits from an 85-year-old White male. Outer retinal bands show similar alterations as the previous case. The split/band #vi is pronounced. G–I, Intermediate AMD eye from a 79-year-old White female. Hyperreflective outer retinal bands have nonuniform or thickened appearance. Hyporeflective band #vi appears continuous into the drusen (d) compartment. J–L, Late dry AMD eye with geographic atrophy from a 77-year-old White female. Band #vii has increased intensity in the atrophic region (g). Hyporeflective wedges (w) within the limits of Henle fiber layer are present on the atrophic borders. RPE-BL-BrM split/hyporeflective band #vi is resolved in region without drusen (yellow arrows). Hyperreflective band #vii is also visualized underneath drusen (blue arrowheads). Scale bar in H applies to B and E, indicating twofold expansion of the axial dimension. Scale bar in I applies to C and F, indicating 10-fold expansion of the axial dimension.
Similar observations were also found in another early AMD eye with subretinal drusenoid deposits (Fig 6D–F, 85-year-old White male). In addition to irregular thickness and elevation of the EZ (IS/OS) and CIZ (COST) bands, the hyporeflective band #vi is more pronounced than that of Figure 6C, despite being ∼14 years older.
Eyes with more advanced disease stages show lesions consistent with results from lower resolution commercial instruments. A cluster of medium-to-large sized drusen can be seen underneath the fovea in an intermediate AMD eye (Fig 6G–I, 79-year-old White female). Drusen interiors are diffusely and moderately scattering, with isolated hyperreflective and hyporeflective foci. The outer retinal bands frequently show disruptions and thickening; at some locations, especially above drusen, not all bands are clearly differentiable. Despite irregular thickening of bands, the hyporeflective band #vi is visible on the borders of the drusen. This band #vi appears to be continuous with the drusen compartment. The hyperreflective band #vii has a thickened, nonuniform appearance.
Finally, OCT B-scans display characteristic choroidal hypertransmission in areas of RPE atrophy in a late dry AMD eye (Fig 6J–L, 77-year-old White female). OCT morphological features of atrophy, including the ELM descent and a hyporeflective wedge posterior to the outer plexiform layer, can be appreciated; the latter feature is better seen in logarithmic scale B-scan because of the enhanced dynamic range. Within the region of GA, the hyperreflective band #vii is prominently visualized, because the RPE is absent, and therefore cannot shadow structures behind it. The hyperreflective band #vii is also continuous underneath drusen and surviving RPE, although the brightness is influenced by OCT beam attenuation. In the nonatrophic region, the hyporeflective band #vi is present adjacent to drusen.
Disruptions and altered outer retinal bands may affect OCT imaging performance, either highlighting or reducing the visibility of deeper layers. Notably, the thinning of the CIZ (COST) may improve the visualization of the underlying hyporeflective band #vi. However, after comparing its visibility across the macula and to aging changes from normal eyes, contradictory cases were found. Therefore, we do not believe that the thinning or reduced CIZ (COST) signal is solely responsible for the increased visibility of the hyporeflective band #vi.
Qualitative Assessment of RPE-BL-BrM Split/Hyporeflective Band Visibility in Normal Aging and Dry AMD Eyes
To assess if the RPE-BL-BrM split/hyporeflective band #vi can be used as an imaging marker for aging and early AMD, we compared the visibility between the 2 categories. Intermediate and advanced dry AMD eyes were not included, because of the limited case number and the diverse phenotypes associated with later stage AMD which can potentially confound analysis. The early AMD cohort consists of 19 eyes from 13 subjects (Table 1).
The intraclass correlation coefficients for the 3 reads ranged from 0.81 to 0.84 for all 3 eccentricity ranges, indicating excellent agreement (Fig S7). The overall visibility percentage is plotted in Figure 8A with respect to age groups and AMD diagnosis. Additionally, a topographic analysis was performed in the region of the central macula–the B-scan segments within the central-inner ETDRS subfields (Fig 8B) and peripheral macula–and the B-scan segments within the outer ETDRS subfields (Fig 8C). We found that with normal aging, both eccentricities showed decreased split/band #vi visibility. Additionally, in normal subjects of aged ≥ 50 years, we noted a sharp decrease of visibility in the central macula, which is less pronounced in peripheral macular or globally. In all eccentricities analyzed, the split/band #vi visibility is significantly higher in AMD eyes compared to normal eyes of similar age (P < 0.026).
Figure 8Qualitative and quantitative assessment of retinal pigment epithelium-basal laminar-Bruch’s membrane complex (RPE-BL-BrM) split/hyporeflective band #vi in normal aging and early age-related macular degeneration (AMD). A–C, Global and eccentricity-based evaluation of percent of split/band #vi visibility, averaged from 3 independent readers. For each age/diagnosis category, N = 12, 9, 5, 11, 13, 13, and 19 from left to right, respectively. The central line indicates the median; the box indicates the 25th and 75th percentiles; the whiskers extend to the extremes. Values outside the 1.5 times interquartile range are denoted as outliers (+). Under Wilcoxon rank sum U test, P1 = 0.0049; P2 = 0.0045; P3 = 0.0056; P4 = 0.026; P5 = 0.0085; P6 = 0.0051, respectively. D, E, Semi-automatic measurement of hyporeflective band #vi thickness. Band #vi thickness decreases with age in normal eyes. Early AMD eyes have thicker band #vi compared to age matched normal eyes. F, G, Semi-automatic measurement of hyporeflective band #vi cross-sectional area. Band #vi area decreases with age in normal eyes. Early AMD eyes have increased band #vi area compared to age matched normal eyes.
Quantitative Assessment of RPE-BL-BrM Split/Hyporeflective Band Visibility in Normal Aging and Dry AMD Eyes
The semi-automatic segmentation algorithm refined the manual tracing based on a set of predefined criteria, reducing a priori bias and minimizing the impact of noise and speckle (Fig S9). The resolvability threshold of 3.2 μm (3.5 OCT pixel) was empirically chosen, based primarily on the axial resolution (∼2.7 μm). The reported RPE-BL-BrM split/hyporeflective band visibility using semi-automatic segmentation algorithm correlates strongly with human reading, though often being ∼20% higher (Fig S7D). The largest discrepancy occurs around 50% to 70% visibility. It is challenging for human readers to identify the RPE-BL-BrM split/hyporeflective band when its thickness approaches the OCT axial resolution, which may explain the disagreement among readers and versus the software in that range.
In normal eyes, the median thickness of band #vi decreased steadily, by −0.041 μm per year (95% confidence interval: −0.050 to −0.032, per linear regression, Fig 8D). Band #vi of early AMD eyes is significantly thicker compared to age matched groups (P < 0.001, Fig 8E). The hyporeflective band #vi area combines both thickness and coverage (i.e., linked to the visibility percentage). The cross-sectional hyporeflective band #vi area decreases in normal eyes (−380.3 μm2 per year, 95% confidence interval: −449.6 to −311.1, Fig 8F), but increases significantly in early AMD eyes (P < 0.001, Fig 8G).
Discussion
OCT is a pivotal tool for clinical diagnosis and management of AMD. New ophthalmic diagnosis and classification strategies are being proposed which utilize OCT features, either supplementing or replacing current standards.
Here, we report that UHR SD-OCT can reliably reveal fine features of outer retina in the clinic. To our knowledge, this is the first demonstration that an RPE-BL-BrM split can be consistently resolved in young, normal eyes, contrary to previous observations using standard resolution commercial OCT. Consequently, a hyporeflective band #vi can be identified within the RPE-BL-BrM complex, separating 2 hyperreflective bands #v and #vii. The visibility of RPE-BL-BrM split/hyporeflective band #vi decreases with normal aging but reappears and is significantly higher in early AMD.
The distinctive alterations in visibility suggest that the hyporeflective band #vi has different histologic and pathologic origins in young normal versus early AMD eyes. In young normal eyes, we hypothesize the band #vi corresponds to the basal RPE, with contributions from RPE basal infoldings. Correspondingly, the hyperreflective band #v maps to the organelle-rich RPE cell body, and hyperreflective band #vii to the BrM (see Supplementary Text). It is worth noting that when band #vi is not resolved, such as with lower resolution OCT instruments or in older normal eyes, the previous consensus interpretation of the RPE-BL-BrM complex is affirmed. While caution must be exercised when suggesting anatomic correlations based on observations from a limited number of subjects, our hypothesis is strongly supported by imaging, histology, and TEM evidence.
Specifically, the measured combined thickness of bands #v, #vi and #vii is 14.2 ± 1.1 μm in young normal eyes, consistent with reported RPE-BL-BrM thickness in histology (excluding RPE apical processes).
The hyporeflective band #vi is believed to have a different origin in early AMD than in young normal eyes. With evidence from TEM and known AMD pathophysiology, we hypothesize that the hyporeflective band #vi corresponds to the accumulation of BLamD in early AMD eyes. The contribution of BLamD in AMD pathology was first described by S.H. Sarks
Ultrastructurally, early BLamD resembles RPE basement membrane material with additional components. While BLamD is present in older normal eyes, it is discontinuous, contained within small patches a few micrometers wide, and is predominately the early, thin type.
This is consistent with our observation that the visibility of the hyporeflective band #vi is low or patchy in these eyes. In AMD, however, BLamD is universally present, and is continuous, frequent, and thick.
When BLamD exceeds a certain thickness threshold in more severe disease stages, it can be visualized by standard resolution commercial SD-OCT as an avascular split of the RPE-BL-BrM complex.
With improved axial resolution and specialized examination and display methods, we believe UHR SD-OCT can now visualize BLamD in earlier disease stages. The accumulation of thick and late type BLamD is associated with vision loss,
Thus, the ability to resolve and quantify BLamD in earlier disease may enable investigation of its role in pathogenesis and progression from early to intermediate stages of AMD.
OCT imaging lacks molecular specificity; thus, we cannot distinguish BLamD from BLinD, which is another distinctive histopathological feature of AMD. Basal linear deposit is the same lipid-rich material as soft drusen in the same anatomical compartment, that is, between the RPE-BL and the inner collagenous layer of BrM. This location is consistent with the hyporeflective band #vi. The lipid-rich composition suggests a hyporeflective appearance in OCT. Compared to BLamD, BLinD is much thinner, with a reported median thickness around 2 μm in AMD eyes.
Given the ∼2.7 μm axial resolution of our prototype instrument, we do not expect BLinD to be routinely visible. However, visualization may be possible on the edges of soft drusen, where the dome of the druse slopes down and becomes continuous with BLinD. This might explain why the hyporeflective band #vi can be seen continuous into the drusen compartment in some B-scan images.
The RPE-BL-BrM split/hyporeflective band #vi may serve as an imaging marker for understanding the physiology of both normal aging and AMD. When analyzing reader results topographically, we observed variations in the ability to resolve the split in the central macula versus perifovea. Specifically, the visibility decreases faster with age in normal eyes within the central-inner ETDRS subfields as compared to the outer ETDRS subfields. However, some older eyes show elevated visibility in the central macula as compared to expected ranges, evident as outliers in Figure 5B. Adding to this trend, 4 of 19 early AMD eyes have > 90% visibility of the split in B-scan regions within the 1.5-mm eccentricity. In contrast, 0 eyes have > 90% visibility in B-scan regions corresponding to the 1.5-mm to 3.0-mm eccentricity. A group statistical test was not significant between the 2 eccentricities (P = 0.62), possibly because of the small sample size and high heterogeneity. Nevertheless, the observation suggests that BLamD first appears in the central versus peripheral macular subregions. This hypothesis is consistent with histological findings, where thick, late BLamD is more often encountered within the central macula.
Furthermore, BLamD is considered a prerequisite for soft drusen formation, the latter having a central concentration, associated with distribution of cone photoreceptors and their supporting Müller glia.
The strengths of our study include the use of a prototype UHR SD-OCT instrument and the inclusion of younger healthy adults for investigating the RPE-BL-BrM complex as a function of age. The UHR SD-OCT can achieve ∼1.5 to 2 times finer axial resolution and faster imaging speed compared to commercial instruments, which, combined with custom designed examination and display protocols, facilitates the visualization of micrometric morphological alterations. Aging is the largest risk factor for AMD,
and studies with a normative age reference database can be impactful for delineating and understanding pathological AMD alterations. We propose interpretations for the observed alterations based on physiological and AMD pathophysiological mechanisms, which are supported by histology, TEM, and current AMD literature. In this study, we focused on the hyporeflective band #vi because of its relevance to AMD; however, there are age-related changes in multiple other features which may also be associated with early AMD or eye health in general. Therefore, this data may provide a benchmark for future UHR OCT studies.
Study limitations include the relatively small sample size per age group and in early AMD. Clinicopathologic comparison of the UHR SD-OCT B-scans and TEM was not feasible, which prohibited establishing a more direct connection between OCT features and underlying structure. The current study used only a small portion of the B-scans from a volume dataset because our analysis approach required manual assisted reading and segmentation. Consequently, we could not investigate the topographical distribution of features in finer detail.
Future research directions include expanding the current analysis to the entire OCT data volume using automated image processing and pattern recognition software. The study used a fixed thickness threshold to determine RPE-BL-BrM split/hyporeflective band #vi visibility in software-based analysis. This empirical value can be optimized, potentially on an eccentricity basis, to improve interpretability as well as diagnosis sensitivity and specificity. Ultrahigh resolution 3-dimensional OCT is synergistic with ongoing studies using volume electron microscopy of outer retina
to bridge ultrastructural findings with clinical diagnostic markers. Longitudinal studies, especially on older eyes that develop AMD and early AMD eyes that progress to intermediate AMD, would be crucial to test the diagnostic and predictive performance of imaging markers, while further verifying their relationship with pathogenesis (e.g., NCT04112667).
The use of a prototype UHR SD-OCT instrument in this study may limit access to wider clinical communities. However, our study observations combined with the examination and display methods described in this study can facilitate interpretation of standard resolution OCT, and commercial manufacturers have begun developing higher resolution instruments. Measures such as thickness of the RPE-BL-BrM complex
might act as a surrogate marker for BLamD in older eyes. Finally, we should note that commercial instruments usually compress the OCT dynamic range and bit depth to allow the image to be displayed on monitor screens. Software modifications to commercial instruments that allow uncompressed OCT B-scans to be saved and analyzed will be important to enable future studies that utilize high resolution features.
In conclusion, this study demonstrates that UHR SD-OCT is a promising modality that can reveal outer retinal alterations associated with normal aging and early AMD pathology. Ultrahigh resolution OCT promises to enable new imaging markers for AMD diagnosis, monitoring progression and response to therapy, as well as to enable investigations of AMD pathogenesis in vivo and accelerate future therapeutic trials.
Volumetric retinal imaging with ultrahigh-resolution spectral-domain optical coherence tomography and adaptive optics using two broadband light sources.
All authors have completed and submitted the ICMJE disclosures form.
The authors made the following disclosures: N.K.W.: Consultant – Apellis, Nidek, Boehringer Ingelheim; Research support – Carl Zeiss Meditec, Heidelberg Engineering, Nidek, Optovue, Topcon, Regeneron, all outside the current work; Shareholder – Ocudyne; Office holder – Gyroscope Therapeutics.
C.A.C.: Research support – Heidelberg Engineering; Consultant – Apellis, Astellas, Boehringer Ingleheim, all outside the current work.
J.G.F.: Royalties from intellectual property owned by MIT and licensed to Optovue. The other authors have no proprietary or commercial interest in any materials discussed in this article.
Supported by the National Institutes of Health (Bethesda, MD, R01EY011289 to J.G.F. and R01EY028282 to C.A.C.); Retina Research Foundation (Houston, TX); Beckman-Argyros Award in Vision Research (Irvine, CA); Greenberg Prize to End Blindness; Champalimaud Vision Award (Lisbon, Portugal); Topcon Medical Systems (Tokyo, Japan); Massachusetts Lions Eye Research Fund (Belmont, MA); Macula Vision Research Foundation (West Conshohocken, PA); Research to Prevent Blindness (New York, NY); and EyeSight Foundation of Alabama (Birmingham, AL). The sponsor or funding organization had no role in the design or conduct of this research.
HUMAN SUBJECTS: Human subjects were included in this study. The study protocol was approved by institutional review boards at the Tufts Medical Center and Massachusetts Institute of Technology. All study procedures adhere to the tenets of the Declaration of Helsinki and comply with the Health Insurance Portability and Accountability Act of 1966. Written informed consent was obtained after explaining the purpose, procedures, benefits, and risks of the study.
No animal subjects were used in this study.
Author Contributions:
Conception and design: Chen, Curcio, Fujimoto
Data collection: Chen, Qamar, Messinger, Baumal, Witkin, Liang, Waheed
00009-X&id=gr1.jpg){kind=link}
00009-X&id=gr2.jpg){kind=link}
00009-X&id=gr3.jpg){kind=link}
00009-X&id=gr4.jpg){kind=link}
00009-X&id=gr5.jpg){kind=link}