Wide-Field Optical Coherence Tomography of the Choroid In Vivo

Abstract

purpose. To demonstrate high-speed, high axial resolution optical coherence tomography (OCT) at 1060 nm with penetration to the sclera. The clinical feasibility of dense, high-speed sampling for higher levels of detail at the macula and optic nerve head is explored with respect to motion artifacts.

methods. A three-dimensional (3D) OCT system making use of a high-speed camera operating at 47,000 depth scans/s was developed. The 1010- to 1080-nm wavelength band leads to 6.7 μm effective axial resolution and enables the acquisition of retinal and choroidal 130 Megavoxel volumes of human subjects within 7 seconds. Motion artifacts were reduced by numeric postprocessing techniques.

results. Drift motion artifacts could be suppressed within fields up to 38° × 38° (approximately 1 cm²) using acquisition speeds of up to 74 frames/s at 512 × 512 pixel/frame. This isotropic OCT sampling of the human retina in vivo allowed reconstruction of the retinal microvasculature solely on vessel reflectivity, without the use of contrast agents, and revealed three interconnected capillary meshworks. Simultaneously in the choroid, the structure of the choriocapillaris, Sattler’s layer, and Haller’s layer were differentiated, and the choroidal-scleral interface was clearly delineated in densely sampled narrow- and wide-angle scans (>38°). At the optic nerve head, the 3D fine structure of the lamina cribrosa and the circle of Zinn-Haller were visualized.

conclusions. OCT almost centered within the 1060-nm water transmission window significantly profits from lower scattering and allows investigation of the retina and choroid at an unprecedented combination of penetration and high speed at high resolution and may provide superior clinical feasibility to commercial 800-nm devices.

Optical coherence tomography (OCT) ¹ extracts three-dimensional (3D), volumetric reflectivity information through transparent media in vivo, at a resolution comparable to that of histologic examination, ^{2 3} enabling early diagnosis of retinal abnormalities. ⁴ New light sources, ⁵ developments in camera technology and electronics, and the advent of frequency domain acquisition, originating from optical frequency domain reflectometry, have recently built the foundation for an increase in light efficiency exceeding 20 dB and a successive increase in scanning speed. ^{6 7 8 9 10 11 12 13 14 15} OCT has been particularly successful in ophthalmology in visualizing the intraretinal microstructure of the fovea and the optic nerve head (ONH). High acquisition speed and high axial resolution ^{16 17 18} are decisive factors for in vivo, especially clinical, OCT applications because motion artifacts impose an upper limit on the overall exposure time of a volume scan, whereas noise level, detector efficiency, and maximum optical power as constrained by safety regulations ^{19 20} limit sensitivity. It is, therefore, essential to investigate new technologies that improve image quality and to find the optimum speed that balances improved sensitivity at or higher than 90 dB for reflectivity measurements in biological material ²¹ compared with motion artifacts. In contrast to the common 800-nm region, scattering is reduced at the 1060-nm water absorption window, ²² which was demonstrated with time domain OCT-based systems ex vivo in porcine retinas ²³ and in vivo in healthy human subjects. ²² Improved visualization of the choroid and better penetration through cataracts could be shown with 3D spectrometer-based OCT at 1060 nm. ²⁴ These studies triggered ongoing interest at this wavelength band for 3D visualization of the vascular structure of the retina and the choroid and for investigating the involvement of individual parts of the blood supply with the development and progress of retinal diseases such as age-related macular degeneration (AMD). ^{25 26 27 28 29 30}  

Ophthalmic spectrometer-based OCT relies on the availability of high-speed camera technology and suitable broadband light sources (>70 nm bandwidth). Although silicon-based cameras use the 400- to 920-nm wavelength region, high-speed linear arrays (≥20,000 spectra/s) with sufficient (>1024) pixels for 1 μm and beyond were not readily available. To circumvent this problem, alternative high-speed tunable sources have been developed ^{30 31} that are difficult to operate, expensive, require complicated detection electronics, and offer only limited optical bandwidth usually centered at the shorter portion of the 1060-nm wavelength band, where melanin absorption is up to 2.5 times higher, restricting high axial resolution OCT imaging and penetration. 

Because of the recent availability of a novel InGaAs diode array with an unprecedented readout rate of 47,000 lines/s, in vivo 3D OCT at 1060 nm for retinal and choroidal imaging became possible. We investigated the possibilities for highly sampled wide-field volumetric OCT imaging of the retina and the choroid, reaching the choroidal-scleral interface (CSI) and the suppression of image artifacts caused by eye motion. 

Methods

Spectrometer-Based 1060-nm OCT System

The key component of the OCT detection system was a recently developed high-speed InGaAs camera (SU-LDH 1024; SUI Goodrich, Princeton, NJ) with a linear 1024-pixel array. The large vertical dimension of the pixels (25 × 500 μm²) gives it a distinct advantage for stability because of its imperviousness to astigmatism and to misalignment caused by thermal distortions or vibrations. The entire array is approximately 25 mm wide and is integrated with high-speed electronics that connect to a personal computer through a base Camera Link standard bus at a sampling rate of 50 MHz. The array is optimized for high-speed spectrometry and can deliver a sustained line rate of approximately 47 kHz sampling at 14 bit per pixel. A similar 1024 × 25 × 100 μm² array operating at a sampling frequency of 38 kHz was used in a parallel study by Suichi et al. ²⁹ The OCT spectrometer, specifically designed for the camera, uses all reflective, off-the-shelf components in an optically simple and mechanically stable Czerny-Turner geometry with two spherical mirrors and a gold-coated 1200 lines/mm grating. ²⁴ Optical modeling of the spectrometer, with ZEMAX (Bellevue, WA) software predicts a chromatic spot size (including geometric aberrations) between 14 and 22 μm in the spectral wavelength region from 1015 to 1095 nm. The efficiency of the instrument was experimentally verified with an amplified spontaneous emission light source (NP Photonics, Tucson, AZ) delivering 72-nm bandwidth at full-width-half-maximum (FWHM) and 19-mW fiber-coupled optical power. The fiber-based Michelson interferometer, with an 80/20 splitting ratio, included a physical dispersion-balanced reference arm. The patient interface was based on a commercial OCT-2 system (Carl Zeiss Meditec, Dublin, CA) with modified optics to support approximately 100-nm bandwidth FWHM at 1060 nm with minimal chromatic aberrations. A 66-D lens at the exit of the fundus camera allows for nearly unvignetted viewing within a scanning angle of approximately 40° × 40° and approximately 10 mm working distance (measured to the corneal surface). The transverse resolution was limited by the approximately 1.2-mm diameter beam and the optics of the eye to approximately 20 μm. With transversal oversampling—that is, the distance between successive depth scans is smaller than half the transversal resolution that decreases the influence of stochastic noise—and successive local filtering, speckles and noise can be suppressed. ³² Through this technique it was possible to generate overviews across large retinal sections, with lateral angles greater than 20°, that included the optic nerve head and the macula to investigate the overall morphology. 

The axial OCT depth-scanning range was determined to be 3 mm in air, the signal-to-noise ratio was 92 dB at the zero delay, and depth-dependent signal roll-off caused by nonlinear discrete sampling of the line array typical for the frequency domain was 19 dB, extrapolated at the end of depth. The effective axial OCT resolution, as determined from the coherence function with a mirror, was 9 μm in air, resulting in 6.7 μm in tissue. It does not degrade with depth across the full scanning range for positive or for negative frequencies. This indicates the good quality of nonlinearity compensation and λ→k mapping, which is necessitated by spectrometer-based acquisition. 

Compared with the predecessor OCT system operating at this wavelength, ²⁴ the depth-scanning range could be enhanced from 1.7 to 3 mm in air. To profit from the camera’s data acquisition speed of 47,000 lines/s, it was set to free running mode during frame acquisition, triggered by the DSP system, which also controlled the timing of the galvanometer scanners. Between frames, a time window of 2.56 ms was introduced to avoid the mechanical instabilities of the scanners. The frame rate was 41 frames/s for a stack of 1024 × 256 lines (depth scans) and a stack of 74 frames/s for an isotropic 512 × 512 lines. 

OCT Data and Image Processing

The spectrometer was characterized by means of a free space interferometer to obtain proper mapping from wavelength to frequency to correct for nonlinearities. Spectral correction was applied with the standard deviation of the spectral signal and fitting to a quasi-Gaussian spectrum, therefore suppressing smaller side lobes otherwise visible at high-contrast reflectors. Global numeric dispersion correction used wavelength-dependent phase shifting of the complex signal with numeric optimization of second- and third-order dispersion with respect to image entropy. ³³ Because OCT at approximately 1-μm wavelength with ∼70-nm bandwidth is less susceptible to dispersive shifts attributed to the vicinity of the zero dispersion point of water and the limited bandwidth of the device, no local (depth-dependent) dispersion compensation was needed, despite the enlarged penetration depth. Postprocessing for motion artifact suppression used image registration based on a modified variable resolution-registering algorithm within successive frames to compensate with translation ³⁴ and optional warping ³⁵ for a complete 3D data set. Furthermore, speckle reduction was optionally integrated by fast dyadic wavelet transform filtering ³⁶ of the high-frequency components. For image registration no external image was necessary, though it was possible to use a 2D en face snapshot fundus camera image or even multiple volumetric scans to fully correct for transversal shifts and to extract the absolute geometric position. 

In Vivo Retinal and Choroidal Imaging of Vasculature in Healthy Subjects

Imaging was performed on a group of 10 healthy subjects who had different retinal pigmentation and were of different ages (range, 19–41 years) and was compared with a spectrometer-based OCT system using a 2048-pixel Si camera operating at 20,000 lines/s with a superluminescent diode centered at 830 nm and with approximately 50 nm bandwidth, resulting in comparable axial resolution of approximately 7 μm in air. This study was approved by the ethics committee at the Cardiff University School of Optometry and Vision Sciences and followed the tenets of the Declaration of Helsinki. Informed consent was obtained from the subjects after the nature and possible consequences of the study were explained to them. The feasibility of the novel high-speed, 1060-nm system was investigated, especially for visualizing the retinochoroidal interface and the CSI at the fovea and the ONH. As an indicator of signal quality, visualization of the CSI was evaluated in the en face stack. Specifically, the ability to discriminate the front and back of this interface at full circumference around the macula was chosen. Total exposure time was kept to less than 7 seconds in the isotropically spaced set of 512 × 512 depth scans of variable angular excursions. To investigate the vascularization of the retina and of the choroid, this scanning protocol covered either the local fine vascular structure or the wider network of larger vessels by including 17°-wide foveal or 38°-wide field scans, respectively. Furthermore, the ONH was investigated with special emphasis on visualizing the 3D structure of the lamina cribrosa. In particular, the shape of its pores is known to play an important role in the early development of glaucoma. ³⁷  

Results

High-Speed, 1060-nm OCT Imaging of the Normal Human Fovea

In 7 of 10 subjects, 3D images could be acquired that visualized the retina, RPE, choriocapillaris, multiple choroidal layers, main choroidal vasculature, and CSI. Figure 1adepicts an OCT fundus image obtained by intensity integration along each single depth-scan. The dashed line indicates the location of the cross-section shown in Figure 1b , where a darker followed by a lighter stripe at the CSI, probably corresponding to the lamina fusca sclerae, are visible. Figures 1c 1d 1e 1f 1g 1h 1i 1j 1k 1l 1m 1nare extracted from a volume of 512 × 512 × 512 voxels over 1.5 × 1.5 × 0.5 mm³ generated by axial averaging to reduce speckle and enhance contrast. Corresponding depth locations and local thicknesses of the en face images are color coded in Figure 1b(borders of Figs. 1c 1d 1e 1f 1g 1h 1i 1j 1k 1l 1m 1n ). The fiber bundles in the nerve fiber layer (NFL) are clearly visualized in Figure 1cdespite the focusing to the choroid. The superficial meshwork of larger capillaries below the NFL is associated with the ganglion cell layer, and the radial parafoveal capillaries (Figs. 1d 1h , inset) can be separated. More distally (Fig. 1e) , a fine net of small inner capillaries (Fig. 1i , inset), situated at the posterior part of the ganglion cell layer (GCL) close to the inner plexiform layer, is visualized. In addition, another layer of capillaries at the junction between the inner nuclear and outer plexiform layers (Fig. 1f)is found that can be used to visualize the avascular zone (Fig. 1j) , even in the slightly defocused case. 

Between the RPE and the choroidal-scleral junction, approximately 20- to 70-μm thick (depending on the layer thickness) depth-integrated en face images unveil the choroidal substructure (Figs. 1k 1l 1m 1n) . Beneath the RPE a highly vascularized layer, the choriocapillaris (CC; Fig. 1k ) is found. Its image has a finely spotted appearance, and polygonal structures that might be associated with clusters of CC lobules of 200- to 250-μm diameter and approximately 20-μm thickness (zoom-in Fig. 1g ) with their typical central arteriole (red arrow) and an array of draining venuoles (blue arrows) were also found in literature obtained by electron microscopy of corrosion casts. ³⁸ The finer texture of the typical microcapillaries of the CC forming a mesh of approximately 10-μm inner diameter is beyond the instrument’s transversal resolution. The following two layers in the choroid, Sattler’s layer (SL, Fig. 1l ) and Haller’s layer (HL, Fig. 1m ), are distinguished by the amount of melanocytes that contribute to scattering within the choroidal stroma and by the density and thickness of the individual vessels. The latter originates from the short posterior ciliary arteries, visible as dark spots in en face scans (Figure 1n) , which perforate the lamina suprachoroidea. Another absorbing layer on top of the first scleral layer, the lamina fusca sclerae, may be associated with the stronger signal. 

High-Speed, 1060-nm OCT of the ONH and Lamina Cribrosa

The ONH was also of particular interest for 3D OCT imaging at 1060 nm. It is supported by an extended mesh of collagen bundles, the lamina cribrosa, which continues the scleral shape in the form of a sieve and gives mechanical support to the nerve fibers before they form the optic nerve. Because of its deeper signal penetration and denser sampling, the high-speed, 1060-nm system is likely to detect reflections from the deeper, almost orthogonally orientated collagen bundles. The posterior ciliary arteries perforating the sclera at the lamina suprachoroidea were visualized at a nearly circular region centered at the papilla (Figs. 2a 2b , red arrows). They originate directly from the ophthalmic artery on the outer side of the sclera that is out of reach for transvitreal optical assessment, whereas the arteries close to the papilla form a circular anastomosis (circle of Zinn-Haller; Figs. 2a 2c , yellow arrows) halfway inside the sclera. Figure 3shows depth-averaged, wide-field, 1060-nm OCT en face images of the lamina cribrosa for two different subjects. At the level of the NFL (Fig. 3a) , superficial blood vessels are visualized. Figure 3bshows an integrated depth section at the level of the lamina cribrosa with obscuring shadows cast by the more superficial vessels and, with low contrast, the structure of the lamina cribrosa. Figure 3cdepicts a full depth-integrated image, and the magnified view (Fig. 3d)unveils the cribrosal structure including its pores (Fig. 3d , yellow arrow). The better contrast in Figures 3c and 3dcompared with Figures 3a and 3bis probably caused by a thinner and thereby more transparent NFL. Visualization of the 3D microstructure of the lamina cribrosa is of particular interest in elderly subjects with neurodegenerative diseases such as glaucoma. 

1060-nm versus 800-nm OCT

Figure 4demonstrates the comparison of high-speed, 1060-nm OCT versus 800-nm OCT in the same subject. Both systems had comparable axial resolution (approximately 7 μm in tissue) but different data acquisition speeds (20 kHz for 800 nm vs. 47 kHz for 1060 nm). Enhanced penetration enabling the visualization of the choroidal-scleral interface is more pronounced in the 38° (11-mm) wide, 1024-line, 1060-nm tomogram (Fig. 4a)compared with the tomogram acquired at 800 nm (Fig. 4b) , with half the sample number and width (19°, or 5.5 mm, at 512-depth scans) despite more than two times longer integration time. The vessel luminae (dark in OCT) have a higher contrast to the vascular walls and the sclera—primarily because of the lower amount of multiple scattering—and, therefore, less noise in the 1060-nm image (Fig. 4a)and less blur. In 70% of all investigated eyes with different pigmentation of the iris, penetration into the choroid was more pronounced with the 1060-nm OCT system. Visualization of the choroidal-scleral interface and the choroidal vasculature across the complete foveal 3D volume was accomplished in 50% of investigated eyes with the 1060-nm device and in 20% of investigated eyes with the slower 800-nm system. This indicates that penetration into the choroid is, on one hand, wavelength dependent but that it also depends on the individual fundus pigmentation of the investigated subjects, which differs between subjects, but also seems to differ between the RPE and the LSF. 

Wide-Field Densely Sampled Isotropic 1060-nm OCT of Healthy Subjects

The significant increase in scanning speed of this novel, high-speed, 1060-nm OCT system enables isotropic and densely sampled wide-field volumetric imaging. A typical 1060-nm, wide-field scan is shown in Figure 5 . In the corresponding en face sequence, the detailed structure of the vascularization of the three retinal networks and the choroid can be investigated. The improved visualization of the CSI at the lamina suprachoroidea (LS) permits evaluation of the total thickness of the choroid in the cross-sections. As the fly-through of the wide field image already indicates, the LS is easily found where choroidal vessels are reduced to dark cross-sections and where the average intensity of the choroidal stroma slightly drops and rises again. This seems to correlate with the short posterior ciliary arteries piercing the sclera before another dark stripe appears in the tomogram, which was interpreted as depth-dependent oscillation in intensity because of polarization changes induced by the birefringence of the sclera ³⁹ when imaged at 800 nm with polarization-sensitive OCT. The stronger morphologic contrast of the 1060-nm tomograms suggests that this layer in fact may be associated with the first scleral layer, the lamina fusca sclerae, which contains a high number of melanocytes that have the ability to influence reflectivity and absorption within this layer. The thickness distribution itself overlaps with the major direction of the choroidal vessels, which start as thinner arteries (Fig. 5d , red arrow) from the papilla and are replaced with thicker veins (Fig. 5d , blue arrow) on their way to the ocular equator, with a bias toward larger vessels and greater thickness in the macular region, though not as strongly related as the retinal vasculature. Here the higher density of small perforating vessels at the macula, rather than merely the transversal connections from the papilla to the equator, can be investigated for the first time. Figure 6demonstrates the enhanced penetration at the longer wavelength side of the 1060-nm spectrum in a dark-pigmented eye as well as its ability to visualize the complete choroid and its structures, including the CSI. The subject’s pigmentation was graded as dark by the visibility of choroidal vessels, which appeared brighter than the choroidal stroma, forming an inverted tigroid fundus image (Fig. 6a)in the periphery ⁴⁰ (note also the more reflective NFL). 

Discussion

The overall time for in vivo imaging is constrained by blinking rate and tear-film stability, whereas involuntary eye motion during fixation (drift and microsaccades) constrain exposure time. Microsaccades as found in Figure 1are reported ⁴¹ to have amplitudes ranging from approximately 0.25° to 2.3°. Depending on their time scale, they appear every 0.5 to 3 seconds with accelerations of up to 14,500°/s² and speeds of 30° to 360°/s. ⁴² Axial and rotational motions, caused by head movements and activity of the oblique eye muscles, result in further distortions in the scanned volumetric image. ⁴³ If these motions are faster than the raster scan, scanning positions become ambiguous, leading to image distortions that cannot be compensated. Hence, the time spent and, therefore, the number of depth scans in the fast-scanning direction of a typical rectangular raster scanning retinal imaging device are limited because of restriction on the advancement speed of the slow scanning axis. For typical isotropic wide scans of angle α approximately 20°, corresponding to A∼5.7 mm (using a conversion factor of 288 μm/degree), corresponds to a depth scan separation of Δx _slow = A · N ⁻¹. For healthy subjects with maximum eye motions of ω_mot = 30 to 100°s⁻¹ (v _mot approximately 8–28 mm/s), ⁴⁴ the number of depth scans (N) in the fast direction at f _s = 47 kHz is limited by  

\[N{<}\sqrt{\frac{A\ {\cdot}\ f_{\mathrm{s}}}{v_{\mathrm{mot}}}}{=}\sqrt{\frac{{\alpha}\ {\cdot}\ f_{\mathrm{s}}}{{\omega}_{\mathrm{mot}}}}.\]

to approximately 100 samples, enabling compensation only of microsaccades. Another solution to avoid image distortions by microsaccades is to restrict the complete acquisition time to less than 0.5 to 1 second. On the other hand, the 10-fold slower drift can be completely compensated by interframe registration rather than mechanical tracking. This will allow acquisition of volume scans representing the 3D structure at high precision. As shown in a recent study in which an experimental 1060-nm tunable laser system was used, ³¹ higher speeds can be used to further improve sampling density or to evade microsaccades. In contrast, the spectrometer-based system permits the user to freely choose the light source and to emphasize longer wavelengths within the 1060-nm water window for higher penetration through melanin and blood. Furthermore, more sophisticated signal processing is possible because of the higher stability of the static spectrometer. This enables visualization of different portions of the circle of Zinn-Haller and feeder vessels that perforate the sclera and proves that penetration at 1060 nm is sufficient to monitor the full retina supporting vasculature of the choroid. Wide-field, high-speed scanning, especially in future combinations with optional Doppler flow measurements, opens the possibility to partially replace invasive and risky fluorescein angiography by a completely noninvasive technique. The ability to distinguish and visualize individual layers of blood vessels above and beneath the RPE at high resolution has great potential to improve diagnostic abilities for diseases such as age-related macular degeneration, diabetes, and the effects of retinal occlusions. Better visualization of the ONH, including the lamina cribrosa and the depth-dependent density and shape of its pores, holds promise to improve glaucoma assessment. At the ONH, thick vessels or thicker NFL limit access to deeper portions at the center of the optic disc in young, healthy subjects. Older patients, especially those with glaucoma, are known to have a shallower nerve fiber layer and a more visible lamina cribrosa. Further studies on variability will probably aid our understanding of the impact of morphologic and functional differences on the visualization and structural parameters of different tissues, including layer thickness, vessel density, and melanin content. Optimizations of the patient interface, as found in the technologically similar commercial systems, are likely to improve overall sensitivity. The spectrometer allows the integration of sources with bandwidths greater than 110 nm that also operate in the better-penetrating, long-wavelength side of the 1060-nm water transmission window to improve axial OCT resolution to approximately 4 μm and to reduce speckle size, whereas camera development will improve imaging speeds. It is expected that clinical studies of abnormalities with 1060 nm OCT will lead to new insights into disease progression and will help early diagnosis and optimized treatment for a variety of patients, especially when optical access to the retina is limited because of scatterers in the anterior eye segment (e.g., cataract or corneal haze). 

 Supported by Cardiff University; FP6-IST-NMP-2 STREPT (NANOUB-017128); Action Medical Research (AP1110); DTI OMICRON (1544C); FP7 FunOCT; and Carl Zeiss Meditec Inc.

 Submitted for publication September 12, 2008; revised November 5, 2008; accepted January 22, 2009.

 Disclosure: B. Považay, None; B. Hermann, None; B. Hofer, None; V. Kajić, None; E. Simpson, None; T. Bridgford, None; W. Drexler, Carl Zeiss Meditec (F, C)

 The publication costs of this article were defrayed in part by page charge payment. This article must therefore be marked “advertisement” in accordance with 18 U.S.C. §1734 solely to indicate this fact.

Figure 1.

 

High-speed, 1060-nm OCT scan of retinal and choroidal microstructures covering the parafovea (5 × 5 mm) (Movie S1). (a) En face OCT fundus image (integrated signal along the full depth) focused into the choroid. (b) Horizontal tomogram through the central fovea (a, dashed line) depicting the retinal, choroidal, and inner scleral layers. En face images are extracted at depths indicated by color coding in b. IPL/OPL, inner/outer plexiform layer; INL/ONL, inner/outer nuclear layer; IS/OS, inner/outer photoreceptor segments; SL, Sattler layer; HL, Haller layer; LFS, lamina fusca sclerae. (c–f) Retina. (g–j) 5× magnifications of marked regions. (k–n) Choroidal structures. (c) Fiber bundles in the NFL. (d) Vessels in the GCL with terminal metarterioles (h, detail). (e) GCL vessels, close to the IPL, forming a fine capillary network (i, detail). (f) Capillaries in the INL, forming the avascular zone (j, detail). (k) CC possibly unveils polygonal CC lobuli (g, detail) of approximately 200-μm diameter with a feeder arteriole (red arrow) and the venoules (blue arrows), associated with clusters of capillaries of 25- to 50-μm diameter, at the resolution limit of the system. (l) Metarterioles of Sattler layer. At the layer (m) of the bigger arterioles of Haller layer, the image distortion caused by a microsaccade becomes apparent (green arrow). (n) CSI taken above the lamina fusca sclerae. White regions correspond to high reflectance of scleral tissue, with larger choroidal vessels penetrating the sclera, visible as black diffuse spots because of their tilt and depth integration.

Figure 2.

 

Sections of 3 × 3 × 1.7-mm volume ONH of a healthy 20-year-old subject imaged by high-speed, 1060-nm OCT (Movie S2). (a) Representative cross-section. Color-coded dashed lines indicate depth positions of en face tomograms depicted in b and c. Penetration beneath the choroidal-scleral border and visualization of the circular arterial anastomosis of varying thickness, the circle of Zinn-Haller (yellow arrows and en face section, c), and the emanating support arteries (red arrows and en face section, b) of the choroid inside the brighter backscattering scleral stroma give proof of the relatively high penetration of the 1060-nm radiation. Note that the circle of Zinn-Haller often has an asymmetric cross-section consisting of a thinner (c, left arrow) and a wider (c, right arrow) portion.

Figure 3.

 

(a) Depth-averaged en face view of the ONH of a 27-year-old subject, consisting of approximately 50 depth slices at the surface. (b) En face view and at the level of the lamina cribrosa (LC) acquired across a 9-mm² area simulating confocal imaging with advanced penetration. (c) Full-depth averaged wide-field (20° [5.7-mm scanning angle]) tomogram of a 40-year-old subject. (d) Magnification (10° or 2.5 mm) contrasts pores of LC (arrow).

Figure 4.

 

(a) Wide-field OCT tomogram (1024 lines, 47 kHz) at 1060 nm with wavelet despeckling. (b) 800-nm OCT tomogram (512 lines, 17 kHz), same eye, location, and sampling density followed by identical processing depicts less contrast. Because of the higher scanning speed of the 1060-nm system, wide scan imaging (38°, corresponding to 11 mm) can be performed in the same acquisition time as with the 800 nm system (19°, corresponding to 5.5 mm), maintaining the sample spacing.

Figure 5.

 

Wide-field view of a 41-year-old healthy subject at 1060-nm signal amplitude enhanced with depth (Movie S3). The OCT fundus image (a) is reconstructed across a 20° × 20° field and is sampled with 512 × 512 depth scans. Motion artifacts are compensated by cross-correlation and warping in both transversal directions. Tomogram (b), intersecting fovea, and ONH (yellow dotted line) depict the full retinal and choroidal structure. Because of depth-dependent intensity compensation, weaker signals from the posterior part of the choroid are visualized. As indicated by the red arrow, the lamina suprachoroidea delimits the CSI. Closely below the lamina fusca sclerae, which is rich in melanocytes, is displayed as a dark line before the scleral soma begins. Integration along the 100 μm (orange indicator) of this slice leads to an image in which the superficial retinal vessels become apparent (c, orange). The same procedure at a thin section below the RPE (d, green) and above the CSI (e, blue) unveils the arterial (d, red arrow), capillary, and venous (d, blue arrow; e) network of choroidal vasculature.

Figure 6.

 

(a) Fundus image of a dark-pigmented south Indian subject, acquired with white light extending across the perifovea, including the ONH. It depicts an NFL with a bright appearance and dark retinal and brighter choroidal vessels with respect to the choroidal background. (b–l) 30° field, densely sampled (512 × 512) OCT scan. Cross-sections are placed in the horizontal (b) and the sagittal (c) planes through the fovea centralis (a, dashed and dotted lines), reaching the CSI. (d) OCT fundus image generated from the 138 Mvx volume. En face coronal sections (color coded with depth) are taken at the parafoveal level through the retinal metarterioles (d) and the deeper capillary layer (e) at the GCL, inner-outer segment junction (photoreceptor-ellipsoids) (f), choriocapillaris (j), network of metarterioles in Sattler layer (k) and Haller layer with its feeding arterioles. Because of the stronger absorption (l), the choroidea-sclera interface has a lower signal, but exiting vessels are still visible as dark spots.

 

Supplementary Materials

Movie S1 - 2.5 MB (.mov) - see Figure 1 

Movie S2 - 3.3 MB (.mov) - see Figure 2 

Movie S3 - 2.6 MB (.mov) - see Figure 5 

The authors thank James Morgan, Cristiano Torti, and Alexandre Tumlinson (Cardiff University) for important discussions and contributions and Doug Malchow (SUI Goodrich) for technical support