Cellular and Functional Optical Coherence Tomography of the Human Retina The Cogan Lecture

An imaging modality that allows for fast, simultaneous, noninvasive probing of both three-dimensional (3D) cellular resolution retinal morphology and depth-resolved function could substantially improve the early diagnosis of various retinal diseases that are the leading causes of blindness worldwide and could contribute to a better understanding of retinal pathogenesis and enhanced therapy monitoring. In addition to user friendliness, reliability, and cost, the key technological parameters of any imaging technique that can markedly influence its clinical feasibility are axial (depth) and transverse image resolution, measurement time, detection sensitivity, penetration depth in tissue, and image contrast. Advances in photonics technology, including the development of ultrabroad bandwidth and high-speed tuneable light sources and high-speed detection techniques, have enabled a considerable improvement in the visualization capability of optical coherence tomography (OCT) in the past decade, establishing it as a state-of-the-art, noninvasive, complementary ophthalmic diagnostic methodology. Axial image resolution has been a key parameter for ophthalmic imaging because of the stratified organization of the retina. In contrast to standard or confocal microscopy, axial and transverse OCT resolution are decoupled, with the axial being mainly determined by the optical bandwidth of the used light source and the transverse by the focusing quality of the measurement beam onto the imaged tissue. In this sense, OCT bridges the gap between the high resolution and limited penetration of confocal microscopy and the low resolution and high penetration of ultrasound. These properties make OCT a unique ophthalmic diagnostic modality, in which high axial resolution, despite a long depth of focus (field)—a situation encountered in retinal imaging in vivo —can be accomplished. Furthermore, the ocular media are essentially transparent, transmitting light with minimal optical attenuation and scattering, to provide easy optical access to the retina. As a consequence, there have been numerous recent developments in OCT technology and considerable interest in this topic, especially in the field of ophthalmology. Objectively, this interest is evidenced by the great increase in publications, patents, and companies involved in the field of OCT in recent years. The market for OCT equipment is predicted to grow at a compound annual rate of more than 30% in the next 4 years, reaching 200 million by 2011. ¹ It is noteworthy that approximately 50% of all OCT publications so far have been in ophthalmology journals, demonstrating the major impact of this technique in this field. Another 25% have been published in optical journals, reflecting the numerous technical advances that have been accomplished. 

Since its invention in the late 1980s ^{2 3 4 5} and early 1990s, ^{6 7 8} the original idea of OCT was to enable noninvasive optical biopsy (i.e., the in situ imaging of tissue microstructure with a resolution approaching that of histology, but without the need for tissue excision and postprocessing). A major limitation in this approach might be the lack of optical contrast between cellular components. 

A first step toward OCT’s acting as a noninvasive optical biopsy method was the introduction of ultrahigh-resolution (UHR) OCT (Fig. 1A)enabling a noticeably superior visualization of all major intraretinal layers, especially the photoreceptor layer, because of the improved axial OCT resolution from 10 to 15 μm to 2 to 3 μm. ^{9 10 11 12} Since this approach was based on a classic technique known as time-domain OCT, it was limited in that it allowed only for two-dimensional (2D) UHR retinal imaging. With the introduction of frequency-domain (also referred to as Fourier- or spectral-domain) OCT, data-acquisition speed was greatly improved, and three-dimensional (3D) UHR OCT was achieved (Fig. 1B) . ^{13 14} Nevertheless 2D and 3D UHR OCT in this configuration could deliver only ultrahigh axial resolution, with a mismatch between axial and transverse resolution of approximately one order of magnitude. The next major advance came with the use of adaptive optics (AO), with deformable mirror technology used to correct higher order ocular aberrations during OCT image acquisition—so-called AO UHR OCT—recently allowing in vivo acquisition of retinal images with cellular resolution (Fig. 1C) . ^{15 16} The development of light sources emitting at alternative wavelengths (e.g., ∼1050 nm) enabled not only unprecedented 3D OCT with enhanced choroidal visualization, due to enhanced penetration below the retina, but also improved OCT performance in patients with cataract because there was less scattering losses in this wavelength region (Fig. 1D) . ^{17 18 19} Finally, extensions of UHR OCT have been developed that enable noninvasive depth-resolved functional imaging of the retina, providing spectroscopic, ^{20 21} polarization-sensitive blood flow or physiologic ²² tissue information. These extensions of OCT should not only improve image contrast, but should also enable the differentiation of retinal diseases via localized metabolic properties or functional (physiologic) state (Fig. 1E) . 

2D UHR OCT: Laser–Tissue correlations

In evaluating the potential of 2D UHR OCT for enhanced visualization of intraretinal structures and to provide an improved basis for correct interpretation of in vivo ophthalmic 2D UHR OCT tomograms of high clinical relevance, studies have been conducted to compare and correlate 2D UHR OCT cross-sectional images of ex vivo pig ²³ and monkey (Macaca fascicularis) ^{24 25} retinal specimens with histology. The pig retina has many features in common with the human retina, but the primate retina provides the closest match available that can be prepared by perfusion fixation techniques. For the exact interpretation of OCT tomograms and correlation to histology, it is important to use fresh retinal specimens. Pig retinas were therefore chosen because of the easier availability within 1 to 2 hours postmortem, compared with human retina samples. Figures 2A and 2Bshow a comparison of in vitro ultrahigh-resolution OCT imaging of a pig retina 2 hours postmortem, acquired with 1.4-μm axial and 3-μm transverse resolutions (Fig. 2A)with the corresponding frozen section imaged by differential interference contrast (DIC) microscopy (Fig. 2B) . From the proximal (top) to the distal (bottom) retina, alternate dark/light bands of signal in the OCT image correlate closely with the retinal layers. Clear demarcations can be seen between the retinal nerve fiber layer (NFL), ganglion cell layer (GCL), inner plexiform layer (IPL), inner nuclear layer (INL), and outer plexiform layer (OPL). Distal to a band attributable to the outer nuclear layer (ONL), a more delicate, bright layer is likely to represent both the external limiting membrane and the myoid portion of the cone inner segments. The adjacent dark, stippled signal is aligned with the cone ellipsoids seen prominently in the DIC image. A distal dark band is possibly associated with the cone outer segment tips and, finally, a dark/light band is attributable to the pigment epithelium/choriocapillaris complex. Figures 2C and 2Dshow a comparison of in vitro ultrahigh-resolution OCT imaging (Fig. 2D)with histology (Fig. 2C)of a macular scan of a monkey retina, again demonstrating excellent correlation of the OCT and tissue boundaries and the potential of 2D UHR OCT, to visualize all major intraretinal layers. The results of these studies allow the extraction of structural retinal information of high clinical relevance with in vivo ultrahigh-resolution ophthalmic OCT tomograms and have mostly eliminated the ambiguities that arose with 2D UHR OCT tomograms in trying to discern the correlation of more than 10 intraretinal layers. 

Despite these studies comparing histology with 2D UHR OCT for the correct interpretation of the obtained tomograms, it is noteworthy that a comprehensive, reliable delineation of all intraretinal layers has not yet been accomplished. The distal part of the retina, including the retinal pigment epithelium (RPE), Bruch’s membrane, choriocapillaris, and choroid complex, remains a particular challenge. Currently, the last (i.e., strongest) continuous distal signal in UHR OCT tomograms has been interpreted as being the RPE layer. Although literature describing the light–RPE interaction in the near infrared region around 800 nm would confirm this, the relatively thick (up to 20–30 μm) appearance in UHR OCT tomograms is not consistent with the RPE as a monocellular layer. Visualization of Bruch’s membrane by OCT is also problematic. The membrane thickens with age and, at most, is 5 μm thick in patients in the 75- to 90-year range, as measured histologically in postmortem tissue. ^{26 27} Since the in vivo UHR OCT axial resolution is, at best, 2 to 3 μm at 800 nm, it is not clear that UHR OCT is capable of resolving and visualizing Bruch’s membrane. Caution is therefore imperative regarding the interpretation of the distal part of the retina, and more comprehensive studies are needed to clarify this issue. 

In several clinical feasibility studies, ^{11 28 29 30 31 32} 2D UHR OCT imaging has been performed with a laboratory prototype as well as a commercially available (Integral; Femtolasers Produktions GmbH, Vienna, Austria) femtosecond titanium:sapphire laser light source, that can generate an 800-nm wavelength with a center bandwidth of 176 nm. ³³ This light source can allow 3-μm axial resolution in the retina as used in a commercially available ophthalmic OCT system (OCT 1; Carl Zeiss Meditec, Inc., Dublin, CA). OCT imaging was performed with axial scan rates up to 250 Hz, using up to 800 μW incident power in the scanning OCT beam, well below the ANSI exposure limits. More than 300 eyes of 200 patients with a range of macular diseases such as macular hole, macular edema, age-related macular degeneration, central serous chorioretinopathy, epiretinal membranes, and detachment of the pigment epithelium and sensory retina were studied. In addition, patients with glaucoma and various hereditary retinal diseases were imaged. The purpose of these studies was to investigate the clinical feasibility of 2D UHR OCT to visualize and quantify intraretinal morphology changes, especially in the inner (IS PR) and outer (OS PR) segments of the photoreceptor layer in healthy subjects and patients with various macular diseases. 

Figure 3shows the visualization and quantification of the IS and OS PR layer thickness in three patients, two with different stages of macular hole and one with Stargardt’s disease. Figure 3Adepicts the right eye of a 59-year-old woman with a visual acuity of 20/25. A 2D UHR OCT examination showed early dehiscence of the inner retinal layers at the margin of the fovea, which form flat intraretinal spaces at the inner border of the ONL. Temporally, the space seems to be only intraretinal, but on the nasal aspect the spaces appear to communicate with the vitreous cavity. The outer retinal layers, especially the external limiting membrane (ELM), the IS PR, the highly reflective interface between IS and OS PR, the OS PR, and the RPE are clearly visualized and appear unchanged. Higher magnification (Fig. 3B)allows better identification of small intraretinal spaces as well as the quantification of the PR layer, which is approximately 98 μm thick in the central and approximately 70 to 77 μm thick in the parafoveal region, indicating normal thickness despite the retinal disease. For the conversion of optical to geometric (i.e., absolute) retinal thickness a phase group refractive index of 1.35 for retinal tissue has been assumed. Figure 3Cdepicts the fundus image with OCT scan indication (white arrow). Figure 3Dshows the left eye of a 63-year-old woman with loss of vision from 20/20 to 20/50. 2D UHR OCT revealed total disruption of the outer retinal layers, especially the ELM, IS PR and the highly reflective interface between the IS and OS PR. The outer retinal layers seem to be pulled up toward the center of the lesion, with approximately 100- to 117-μm PR layer thickness (Fig. 3E) . These appearances are the result of a retinal detachment in this area that has resulted in the stretched appearance of the PR OS. In the parafoveal region, the PR layer thickness is approximately 76 to 77 μm, which lies within the normal range. Figure 3Fdepicts the fundus image with OCT scan indication (white arrow). Figure 3Gshows a 2D UHR OCT scan through the fovea of a 30-year-old female patient with Stargardt’s disease, central microperimetry measurement (Fig. 3H) , and the corresponding location of the 2D UHR OCT scan relative to the fluorescein angiogram (Fig. 3I) . Due to a substantial atrophy of all intraretinal layers, but especially of the outer nuclear layer, IS and OS PR), reduction of the total retinal thickness decreased to 65 μm, as revealed by 2D UHR OCT. In the parafoveal region, small remnants of the IS and OS PR with a thickness of approximately 30 to 33 μm, can be visualized by 2D UHR OCT. Microperimetry measurements (performed by Erdem Ergun, Medical University, Vienna) revealed the existence of an absolute scotoma (Fig. 3H) . The size of this absolute scotoma corresponded with the area of transverse photoreceptor loss as visualized by 2D UHR OCT (Fig. 3G , red arrows) demonstrating the correlation of morphology (detected by 2D UHR OCT) and retinal function (detected by microperimetry). These clinical studies demonstrated that 2D UHR OCT enables unprecedented visualization of all major interaretinal layers, which until then had been possible only with histopathology. 

3D UHR OCT

2D UHR OCT, because it is based on time-domain OCT, has a fundamental time–bandwidth tradeoff, in that imaging can be accomplished with sufficient sensitivity either at ultrahigh speed with standard resolution or at standard speed with ultrahigh-resolution, but not at both especially because of the low power levels allowed for in vivo retinal imaging (∼600–800 μW at 800 nm). As a consequence of slow scanning/measurement time, motion artifacts have to be corrected by using postprocessing algorithms, and only a limited number of A-scans per B-scans can be obtained. State-of-the-art time-domain–based OCT systems have been developed to provide high scanning speeds of up to 8 kHz, (i.e., 8000 A-scans per second), ³⁴ with the drawback of lower system sensitivity at the faster scanning speeds and/or broader optical bandwidth, to accomplish ultrahigh resolution. One possible approach to performing 3D retinal imaging is an extension of time-domain OCT called en face OCT. ^{35 36 37 38 39 40 41} Another alternative to time-domain OCT for high-speed imaging is a technique by which the entire spatially resolved tissue reflectance (A-scan) is obtained simultaneously, thereby removing the need for depth scanning. ^{42 43 44 45} This technique has been referred to in the literature as Fourier-domain OCT (FD OCT), spectral-domain OCT, frequency-domain OCT, or swept-source OCT. This technical improvement is accomplished either by dispersing the interferometric information in space using a spectrometer setup (Fourier-domain OCT or spectral-domain OCT) or by encoding information in time by tuning a monochromatic light source in combination with a single photodetector (frequency-domain OCT or swept-source OCT). The acquisition speed for both approaches, with a spectrometer used as a detector or a tuneable light source, is then mainly limited by the readout rate of the charge-coupled device (CCD) camera or by the tuning speed of the light source, respectively. Since the scanning range and electronic detection band width are decoupled, both approaches offer substantial improvements in sensitivity and a dramatic increase in the line rate (A-scan rate) without reducing the imaging performance compared with time domain OCT. Although the basic principle of frequency-/Fourier-/spectral-domain OCT has been known since 1995, ⁴² CCD technology and proper recognition delayed the experimental and clinical evaluation of this technique for almost a decade. Three groups in parallel have demonstrated the great advantage of this technique in data acquisition speed and sensitivity. ^{46 47 48} FD OCT with a spectrometer used for detection has recently been shown to offer high-speed, ^{49 50 51} ultrahigh-resolution, ^{13 52 53} functional imaging. ^{54 55}  

The high measurement speed of FD OCT enables in vivo 3D UHR OCT. In the time required for a single 2D UHR OCT B-scan (using time domain OCT), the equivalent of 50 to 100 of these 2D UHR OCT B-scans can now be acquired by FD OCT, covering the entire retinal volume (Fig. 4A) . By analogy with scanning laser ophthalmoscopy measurements, 3D UHR OCT can raster scan the selected retinal area and acquire full-depth morphologic information at any given point without the need for scanning in depth. A volumetric data set is acquired for the imaged area that that can then be viewed and analyzed in ways similar to those used with CT or MRI scans (Figs. 4A 4B) . Hence, the imaged volume can arbitrarily be cut according to the necessary diagnostic needs (e.g., as in a UHR SLO in en face [C-mode] tomograms [Figs. 4C 4D ]). In several pilot studies, 3D UHR OCT has been evaluated in the clinical imaging, and frequency-domain OCT has provided video rate imaging of retinal disease. ^{14 56 57} Figure 4demonstrates 3D UHR OCT in the foveal region of a patient with RPE atrophy (note that the axial dimension is two times larger than the other two dimensions, for better visualization). The 3D representation (developed in collaboration with Carl Glittenberg and Susanne Binder, Ludwig Boltzmann Institute der Krankenanstalt Rudolfstiftung, Vienna, Austria) of the macular region is presented at different angled views (Figs. 4A 4B)depicting the pathological change in the topography of the foveal depression as well as enabling unprecedented views in which the retina can be observed from any direction, including from below (Fig. 4B) . Figures 4Cto 4Hshow a virtual biopsy/surgery performed with 3D UHR OCT in combination with 3D data rendering, which allows the user to excise and remove any given layer or part of the retinal volume to visualize intraretinal morphology. An important advantage of this method is that it is reversible, in that the virtual retina can be reconstructed. Scanning a retinal volume with nearly isotropic (reasonably small equidistant) sampling intervals also avoids the need to decide on the orientation of any line scans that would be required with the time domain OCT. Figure 5shows 3D UHR OCT of both eyes of a patient with a macular hole. The patient’s right eye (Fig. 5A)shows clear tomographic (Fig. 5B)and topographic (Fig. 5C)impairments due to the macular hole. The left eye (Fig. 5D)has been diagnosed as normal with standard diagnostic techniques. By scanning the whole central foveal volume, 3D UHR OCT reduces the risk that retinal features indicating subtle tomographic (Fig. 5E)and topographic (Fig. 5F)photoreceptor impairment will be missed. 

3D UHR OCT based on FD OCT enables acquisition of substantially more sampling points (measurements) across regions of interest that are sufficiently large for the derivation of clinically relevant retinal thickness maps based on segmentation of intraretinal layers. ⁵⁸ Of importance, 3D UHR OCT is capable of providing topographic information that is equivalent to that provided by scanning laser tomography—for example, with the HRT (Heidelberg Retina Tomograph; Heidelberg Engineering, Heidelberg, Germany) or quantification of circumpapillary nerve fiber layer thickness similar to that estimated by scanning laser polarimetry (GDX VCC; Carl Zeiss Meditec, Inc.). 

Cellular Resolution OCT

In retinal OCT, the typical transverse resolution using an incident beam of approximately 1 mm diameter (at 800 nm) to scan the retina, is on the order of ∼20 μm. which is approximately one order of magnitude worse than the axial OCT resolution and results in a depth of focus of approximately 750 μm. The spot size can be reduced (to improve the transverse resolution and reduce transverse speckle size) by increasing numerical aperture, but this has the disadvantage that it also reduces depth of focus. For ophthalmic retinal OCT imaging, where the cornea and the lens act as the imaging objective, the numerical aperture can be increased by dilating the pupil and increasing the measurement beam diameter. In practice, however, for large pupil diameters, ocular aberrations limit the minimum focused spot size on the retina, even for monochromatic illumination. When ultrabroad bandwidth light is used, chromatic aberration will pose an additional limit to the smallest possible spot size. ^{59 60 61} An alternative and promising approach is to use AO to minimize ocular aberrations and to reduce the spot size. AO, which was originally developed to improve the resolution of astronomical imaging, has been an important step forward in improving resolution in ocular imaging. Methods for the correction of ocular aberrations were suggested in the early 1960s, but the first substantial improvements in the resolution of retinal imaging were not seen until the 1980s, pending the development of appropriate AO technologies. ^{62 63} Deformable mirrors ⁶⁴ and liquid-crystal spatial light modulators ⁶⁵ are readily available and offer a cost-effective solution not only for static, but also for closed-loop dynamic correction of ocular aberrations. ⁶⁶ Recently, they have been successfully interfaced to ophthalmic imaging techniques such as conventional flood illuminated ^{63 67} or scanning imaging systems, ⁶⁸ to provide higher contrast and transverse resolution. In this mode, they have enabled the in vivo en face visualization of photoreceptors as well as ganglion and RPE cells in animal models. ⁶⁹  

The interface of AO to 3D UHR OCT has the potential to enable further improvement in the visualization possibilities of ophthalmic OCT. In addition, equivalent axial and transverse resolution in the retina combined with ultrahigh-speed OCT would enable isotropic resolution and sampling, raising the possibility that cellular-resolution retinal imaging is possible with the proviso that the optical contrast of cellular features, (e.g., photoreceptors or ganglion cells) is sufficient at the imaging wavelength region (typically around 800 nm). First attempts to combine an en face coherence-gated camera with adaptive optics resulted in en face OCT imaging with standard axial resolution (14 μm) and high transverse resolution. ⁷⁰ The first AO UHR OCT was developed by using a compact (300 × 300 mm) closed-loop AO system, based on a real-time (30 Hz) Hartmann-Shack wave front sensor and a deformable mirror with 37 actuators membrane. These were interfaced to a time-domain 2D UHR OCT system, based on a commercial OCT instrument, with a compact titanium:sapphire laser with 130-nm bandwidth ¹⁵ providing the illuminating beam. These modifications delivered a two- to threefold improvement in transverse resolution compared with previously used UHR OCT systems and a significant signal-to-noise ratio improvement of up to 9 dB. Volumetric AO UHR OCT of the living retina, by a 3D UHR OCT delivering 25,000 axial scans per second and wavefront correction with a programable phase modulator, enabled the visualization of cellular retinal features that might correspond to terminal bars of photoreceptors at the level of the external limiting membrane. ¹⁶ AO based on one to two deformable mirrors has also been applied to an OCT with moderate axial resolution, to provide an unprecedented level of segmentation of the intraretinal layers. ^{71 72 73}  

As broad-spectrum bandwidth light sources became a necessary requirement in UHR OCT, the importance of correcting both monochromatic and chromatic aberrations (pancorrection) became apparent if the potential increase in transverse resolution in AO UHR OCT images were to be realized. For this reason an achromatizing lens has been designed that could compensate the average values of chromatic aberration in the normal human eye, in the 700 to 900 nm range. ⁶⁰ The compensating lens was then implemented in an AO system for full aberration correction, by using a novel magnetic force-based deformable mirror. The correcting device exhibited an unprecedented ability to compensate for large monochromatic aberrations. Use of a 140-nm optical bandwidth in combination with compensation of the eye’s chromatic aberration not only enables ultrahigh axial OCT resolution in the retina but also reducing the axial speckle size in the tomograms. As a result of these developments, high-resolution views of the living retina can now be obtained that have high (nearly isotropic) resolution and less speckle noise. 

Figure 6depicts a comparison of histology in the parafoveal region of a monkey retina (Fig. 6A)versus in vivo AO UHR OCT of a normal human retina at ∼2° parafoveal location (Fig. 6B ; comparison obtained in collaboration with Peter Ahnelt, Department of Physiology, Medical University of Vienna, Austria). It should be mentioned that the OS of the monkey retina showed some artifacts (tilted OS) due to histologic preparation. Only the distal part of the retina from the myoid to the RPE is visualized by AO UHR OCT, with ∼75-μm transverse extensions. Because of ultrahigh axial and transverse resolution, morphologic details of single photoreceptors (cones) can be visualized for the first time. At the top of the figure, weak, elongated signals followed by strong, elliptical features, clearly differentiated by AO UHR OCT, probably correspond to ellipsoids and myoids, together constituting the cone inner segments. Small diameter cylindrical elements emerging from the ellipsoids constitute the basal outer segments which appear to be transmuting into elements giving stronger brighter signals (green arrows), possibly including cappings by pigment epithelium processes and interphotoreceptor matrix. 

Black (indicating weak OCT signal) regions between the cones (Fig. 6 , asterisk) may relate to the interphotoreceptor space but may also include rods interspersed between cones that cannot be resolved, yellow arrows may correspond to pigment epithelium components including melanin granules, known mainly to absorb light. Further studies are needed to clarify and interpret the appearance of the single cones in these preliminary AO UHR OCT tomograms. 

3D OCT with Enhanced Penetration into the Choroid

So far, commercial and scientific clinical ophthalmic OCT has mainly been performed in the 800-nm region, due to the availability of light source technology in this wavelength range. ⁷⁴ Although 800-nm OCT systems can resolve all major intraretinal layers, they enable only limited penetration beyond the retina, resulting in limited visualization of the choriocapillaris and choroid. Furthermore, in clinical OCT imaging, cataract represents a significant challenge when imaging the retina. In the 600- to 1200-nm region, scattering decreases monotonically with increasing wavelength while the scattering behavior of light in biological tissues shows a significant decrease with longer wavelengths (proportional to ∼λ⁻⁴). For that reason OCT imaging at 1050 nm can deliver deeper tissue penetration structures beneath the RPE, as well as delineation of the choroidal structure. Figure 7Ashows 2D retinal OCT imaging in vivo in the foveal region of a healthy human subject performed with a time-domain–based system at 1050 nm (data from Boris Hermann, Cardiff University, Wales, UK) indicating superior visualization of the choroid due to improved penetration at longer wavelengths. ^{17 75} In addition, depending on the melanin levels in the RPE of the subject, images can also be obtained of the choroidal–scleral interface (Fig. 7A , arrows). In a recent study, the clinical feasibility of 3D high-resolution (∼7-μm axial resolution) OCT at 1050 nm was investigated in patients with retinal diseases and cataract. ¹⁹ 3D in vivo retinal imaging at 1050 nm was performed by using an FD-OCT system at 20 frames/s, demonstrating enhanced penetration into the choroid in which the reduced scattering at this wavelength significantly improved the imaging performance in patients with cataract, thereby widening the clinical applicability of ophthalmic OCT. 3D OCT at 1050 nm in a patient with cataract was performed (Fig. 7B)and compared with those accomplished with 3D UHR OCT at 800 nm (Fig. 7C) . The image obtained at 800 nm (Fig. 7C)had sufficient signal strength to allow 3D imaging. However, the image obtained at 1050 nm (Fig. 7B)provided considerably greater detail of retinal structure with preserved integrity of images covering the deeper retinal layers. An important application of imaging at longer wavelengths is that since they do not stimulate the retina, they can be used to probe the effects of retinal stimulation by other, shorter, more visible wavelengths. In this context, it has been demonstrated that OCT can be used to detect depth-resolved physiological correlates of neuronal activity ^{22 76} within the retina. 

Functional OCT

A variety of functional extensions of OCT technologies have been developed in the past, of which Doppler OCT, for measuring the blood flow velocity, ^{77 78 79 80} and polarization sensitive OCT, for imaging depth-resolved tissue birefringence, ^{81 82 83} have been the most developed and successfully applied in retinal imaging. Electrophysiology remains the gold standard for the quantification of retinal activity. This method is invasive and time intensive and has no depth resolution and poor transverse resolution. Noncontact, optical probing of retinal responses to visual stimulation with 10-μm spatial resolution, achieved using functional UHR OCT has recently been demonstrated for the first time. ²² This method relies on the observation that physiological changes in dark-adapted retinas caused by light stimulation can result in local variations in tissue reflectivity. This functional extension of OCT can be considered as an optical analog to electrophysiology and has therefore been called optophysiology. Optophysiology can be used for noncontact, high-resolution, spatially resolved probing of the physiological responses in light-stimulated retinas. ²² To determine the sensitivity of optophysiology for the detection of changes in retinal reflectivity triggered by light stimulation, a dark-adapted, living in vitro rabbit retina was exposed to a single flash of white light, and optophysiology data were acquired synchronously with electroretinogram (ERG) recordings. For in vitro experiments, the system was interfaced to a state-of-the-art fiber laser, with an emission spectrum centered at 1250 nm and a spectral bandwidth of 150 nm. A light source with longer central wavelength was chosen for these experiments to avoid prestimulation of the dark-adapted retinas during the optical recordings. Throughout the functional experiments the isolated retinas were stimulated with single, 200-ms white light flashes. A morphologic B-scan was first taken from the measurement location (Fig. 8A) . Multiple UHR OCT depth reflectivity profiles (A-scans) were then acquired at one transverse location in the retina (Fig. 8B)synchronously with ERG recordings (Fig. 8C) . The UHR OCT A-scans were combined to form 2D raw data M-tomograms presenting the retina reflectivity profile as a function of time (Fig. 8B) . The optical data were processed by using a cross-correlation algorithm to account for any movement of the retina caused by the solution flow and for calculation of the optical background (average over the prestimulation A-scans of each M-tomogram) and generation of differential M-tomograms from the raw data M-tomograms (Fig. 8D) . Optophysiological signals could be extracted from various retinal layers, so that depth-resolved optical backscattering changes that resulted from physiological processes induced by the optical stimulus could be detected (Fig. 8E) . Figure 9Ashows an OCT retinal image of the rabbit retina, demonstrating that UHR OCT is capable of distinguishing all major retinal layers. This comparison is essential in establishing the morphologic and the physiological origins of any changes in the recorded optical signal, observed in the differential M-scan. Figure 9Eshows a representative differential M-tomogram arising from a single-flash stimulus. As expected, in the nonstimulated retina (Figs. 9B 9C 9D)the optical reflectivity of the PR layer did not change significantly with time. When the retina was exposed to the light stimulus (yellow box), changes were seen in optical backscattering at locations corresponding to the IS (Fig. 9F)and OS (Fig. 9G)PR layer, which correlated with changes in the corresponding ERG (Fig. 9H) . Optical backscattering increased significantly after the light flash and then returned slowly to baseline. When KCl was applied to the retinal sample to inhibit photoreceptor function (Figs. 9I 9J 9K 9L) , the optical changes observed in the IS and OS PR were close to the optical background level and showed no correlation to the onset of the light stimulus. Depolarization of the cell membranes can occur during conduction of an action potential which could be detected by UHR OCT, but also by detection of spatially resolved change in backscattering over time. The exact origin of the detected optophysiologic signals is unclear but may be related to the dipole reorientation (and therefore refractive index changes) at the photoreceptor membrane. Alternatively, they could arise from light-induced isomerization of rhodopsin in the OS PR or metabolic changes in the mitochondria of the IS PR. Noninvasive in vivo functional optical imaging of the intact rat retina has recently been demonstrated using high-speed UHR OCT. Imaging was performed with 2.8-μm resolution at a rate of 24,000 axial scans per second. ⁷⁶  

Spectroscopic OCT

Spectroscopic and wavelength-dependent OCT is a rather new area of investigation, and only a few in vitro studies have been performed to date. Studies have been limited by the lack of light sources with sufficiently broad bandwidth. ^{84 85 86} With spectroscopic OCT (SOCT), localized absorption spectra of endogenous or exogenous chromophores in the tissue may be measured, which can be used to enhance contrast in the OCT image or to extract functional information from the tissue. ^{20 87 88 89 90 91 92 93 94 95}  

Using state-of-the-art femtosecond titanium:sapphire lasers that emit broad-bandwidth light centered at 800 nm not only enables (sub)cellular level resolution OCT, but also may provide spectroscopic information over the entire output bandwidth. This spectral region is important, because it overlaps with the so-called therapeutic window, covering the absorption features of several biological chromophores. ⁹⁶ Therefore, this extended version of OCT may enable the extraction of spatially resolved spectroscopic information and improve OCT image contrast. In addition it could allow the derivation of functional or metabolic information from investigated tissue. In standard OCT imaging, only the envelope of the interference signal was detected (Figs. 10A 10B) . Spectral information can be obtained by measuring the full interference signal and using appropriate digital signal processing (e.g., using a Morlet wavelet transform; Figs. 10C 10D ). To display the spectroscopic data in a simple color image, the “center of mass” of the spectra was calculated in these preliminary experiments (Figs. 10E 10F 10G) . SOCT imaging requires a multidimensional map. This map can be obtained by using hue, saturation, luminance (HSL) color space (not RBG) and mapping the intensity into the saturation and the center of mass of the spectrum into the hue, keeping luminance constant. With all techniques that use reflected and backscattered light for imaging of tissue, the specific absorption and scattering properties of the intermediate tissues determine the amount of light that is detected from each location within the retina. Figure 10Hdepicts in vivo spectroscopic OCT in a patient with RPE atrophy. The area of reduced/lost melanin concentration (Fig. 10H , between the two arrows) is contrasted by a discontinuous yellow band at the RPE level (Fig. 10H , asterisk). It is important to note that OCT image contrast results from a combination of absorption and scattering. Incident light is attenuated by scattering and absorption as it propagates through the tissue, is backscattered from the internal structure that is being imaged, and is again attenuated as it propagates out of the tissue. ⁹⁷ Thus, the optical properties (absorption and scattering) from deep structures are convolved with the properties of the intervening structures, making it challenging to determine the exact optical properties of a given internal structure. However, OCT provides more information than other spectroscopic imaging techniques which integrate backscattered light from multiple depths within tissue. Spectroscopic OCT may also be used to enhance image contrast, enabling the differentiation of tissue diseases via their spectroscopic properties or functional state. Spectroscopic OCT could thus function as a type of “spectroscopic staining,” analogous with staining in histopathology, and should be able to detect spatially resolved functional, biochemical tissue information over the entire emission wavelength region of the light source with a single measurement. 

Conclusion

Present clinical practice calls for the development of techniques to diagnose disease in its early stages when treatment is most effective and significant irreversible damage can either be prevented or postponed. OCT being a noninvasive, cross-sectional imaging modality providing morphologic and functional tissue information simultaneously is therefore a potential candidate. In addition to the eye’s easy accessibility, OCT was always successful in the past, since there was not really another competing noninvasive technique that could provide the clinician with the same wealth of depth-resolved morphologic and functional retinal information. Furthermore, it is noteworthy that ultrahigh axial resolution OCT is more successful in mainly single-scattering stratified organs, such as the retina, compared with optically dense organs (e.g., brain). Recent accomplishment in light source and detection technology are an important step toward accomplishing the original idea of OCT’s acting as an optical biopsy technique, (i.e., providing nearly the same information as that accomplished with histopathology). Despite the ground-breaking technological developments of the recent years, several questions/challenges remain to be resolved:

•  
  What is the actual clinically necessary and technically feasible axial and transverse resolution needed for improved ophthalmic diagnosis?
•  
  Is the 3-dB definition for axial resolution of OCT systems practicable, if OCT tomograms are displayed with up to 40-dB dynamic range?
•  
  How can in vivo axial and transverse resolution be reliably determined in the retinal tomograms? Even the highest isotropic (equidistant) resolution that is achieved in the future may not enable the visualization of certain morphologic features in the living retina (e.g., Bruch’s membrane, ganglion cells or others) if contrast is insufficient.
•  
  Are 3D UHR OCT retinal tomograms being interpreted properly?
•  
  What is the optimum central wavelength for ophthalmic OCT?

Because of the recent significant technologic improvement, perhaps there will be some saturation in future efforts to improve the key technological OCT parameters (e.g., resolution, scanning/data acquisition speed, sensitivity, and penetration). Therefore, the future of OCT based on this superb technological performance of visualizing morphology may be to localize tissue function. This situates OCT in a unique position to perform noninvasive visualization of microstructural intraretinal morphology, but due to its depth-resolving nature, it will also enable the acquisition of unprecedented depth-resolved functional tissue information—ideally, performed with a single measurement and dedicated postprocessing. 

Hence, 3D ultrahigh isotropic resolution OCT (interfaced to AO) in combination with ultrafast scanning/data acquisition has enabled a quantum leap in OCT performance in the past few years. OCT can now be considered to be an optical analog to CT or MRI offering limited penetration, but with microscopic resolution. In addition to functional extensions of OCT, this technique may have the potential to revolutionize ophthalmic diagnosis in the very near future. 

 Supported by Cardiff University Grant FP6-IST-NMP-2 STREPT (017128); ÖNB Jubiläumsfonds 10J06; Christian Doppler Society Grants FWF P14218-PSY, FWF Y 159, and CRAF-1999-70549; Femtolasers GmbH, Vienna, Austria; Carl Zeiss Meditec, Inc., Dublin, CA; Multiwave Photonics, Porto, Portugal; NP Photonics, Tucson, Arizona; and Maxon Computer GmbH, Friedrichsdorf, Germany.

 Disclosure: W. Drexler, Carl Zeiss Meditec, Inc. (C, F); Femtolasers GmbH; Multiwave Photonics; NP Photonics; Maxon Computer (F)

 

Figure 1.

 

Improving the key technological parameters of OCT. (A) Axial resolution enabled by 2D UHR OCT; (B) improved data acquisition speed enabled by 3D UHR OCT; (C) improved transverse resolution enabling cellular resolution OCT by 3D AO UHR OCT; (D) improved penetration depth into the choroid with 3D HR OCT at 1050 nm; (E) enabling the noninvasive localization of retinal tissue function (e.g., optical probing of retinal physiology-optophysiology).

Figure 2.

 

Radial view of a pig retina (A) as revealed by in vitro 2D UHR OCT imaging and (B) DIC microscopy of the matching radial frozen section. Note that OCT signal bands correlate well with the retinal layers/sublayers, from the GCL level to the IS and OS PR level. Foveal portion of corresponding semithin histologic section (C) and 2D UHR OCT image (D) of a perfusion-fixed monkey retina. gc ax, ganglion cell axon layer; gc, ganglion cells; ipl, inner plexiform layer; inl, inner nuclear layer; Hf, fibers of Henle; onl, outer nuclear layer; cis/cos, cone inner/outer segments; pe, pigment epithelial layer; ch cap, choriocapillaris; ch, choroid; ★, darker faults in foveal floor indicative of foveal strain; d, epiretinal debris.

Figure 3.

 

2D UHR OCT enables qualitative, quantitative analysis and correlation to functional measurement of the photoreceptor layer in various retinal diseases. Horizontal 2D UHR OCT image of patients with different stages of macular holes (A–F) and Stargardt’s disease (G–I), magnification of the central foveal region with quantification of the PR layer thickness (B, E), microperimetry (H), and fundus photograph (C, F), and fluorescein angiogram (I). (C, F, I, arrow) Scan location.

Figure 4.

 

3D UHR OCT enables unprecedented volumetric representation of the macular region of a patient with RPE atrophy from views at different angles (A), including from below (B). Virtual biopsy/surgery using 3D UHR OCT (C–H) allows the user to excise and remove any given layer or part of the retinal volume to visualize intraretinal morphology (developed in collaboration with Carl Glittenberg and Susanne Binder, Ludwig Boltzmann Institut der Krankenanstalt Rudolfstiftung Vienna, Austria).

Figure 5.

 

3D UHR OCT of both eyes of a patient with a macular hole. The patient’s right eye (A) shows clear tomographic (B; intraretinal cysts [top arrows] and photoreceptor impairment [bottom arrows]) as well as topographic (C; arrow, topographic change of foveal depression) impairments due to the macular hole. The left eye (D) had been diagnosed as normal. 3D UHR OCT clearly indicates early, subtle tomographic (E; arrows, subtle morphologic photoreceptor changes) as well as topographic (F) impairment.

Figure 6.

 

Comparison of histology in the monkey retina (A) versus in vivo AO UHR OCT of a normal human retina (B, in collaboration with Peter Ahnelt, Department of Physiology, Medical University of Vienna, Austria). The OS of the monkey retina show some artifacts (tilted) due to histologic preparation. AO UHR OCT enables the visualization of single PR (cones) OS morphology. CC, choriocapillaris; (*) regions that may relate to the inner photoreceptor matrix or rods; green arrows: cone OS tips; yellow arrows: pigment epithelial processes.

Figure 7.

 

(A) 2D in vivo retinal OCT imaging in the foveal region of a healthy human subject performed by a time-domain–based system at 1050 nm (preformed by Boris Hermann, Cardiff University, Wales, UK), indicating superior visualization of the choroid and choroidal–scleral interface (arrows). (B) 3D HR OCT at 1050 nm in a cataract patient versus 3D UHR OCT at 800 nm (C). The image obtained at 1050 nm (B) provides considerably more detail of retinal structure with preservation of the integrity of the deeper retinal layers.

Figure 8.

 

Optophysiology: basic principle. (A) Morphologic B-scan from the measurement location; (B) multiple UHR OCT depth reflectivity profiles (A-scans) are acquired at one transverse location in the retina synchronously with ERG recordings (C). (D) Generation of differential M-tomograms of the raw data M-tomograms; (E) different optophysiological signals can be extracted from various retinal layers, enabling depth-resolved optical backscattering changes that are due to physiological processes induced by the optical stimulus.

Figure 9.

 

In vitro optophysiology: optical probing of depth-resolved retinal physiology. (A) OCT retinal image of the rabbit retina; (B–D) response to no light stimulus; (E–H) representative single-flash stimulus differential M-tomogram and extracted responses from the inner (F) and outer (G) photoreceptor layer; (I–L) case of KCl-inhibited photoreceptor function; yellow boxes: time duration of the light stimulus. (D, H, L) Simultaneous ERG recordings.

Figure 10.

 

(A, B) Standard OCT uses only the envelope of the interference signal; (C, D) spectral information can be obtained by measuring the full interference signal and using appropriate digital signal processing; (E–G) to display the spectroscopic data in a simple color image, the “center of mass” of the spectra is calculated; (H) in vivo spectroscopic OCT in a patient with RPE atrophy. The area of reduced or lost melanin concentration (H, between the two arrows) is contrasted by a discontinuous yellow band at the RPE level (H, *).

The authors thank Boris Herrmann, Bernd Hofer, James E. Morgan, Boris Považay, and Angelika Unterhuber, from the School of Optometry and Vision Science, Cardiff University; Adolf F. Fercher, Leopold Schachinger, and Harald Sattmann from the Centre of Biomedical Engineering and Physics, Medical University Vienna, Austria; Enrique J. Fernandez, and Pablo Artal from the Laboratorio de Optica, Universidad de Murcia, Spain; Kostadinka Bizheva, Department of Physics and Astronomy, University of Waterloo, Canada; Carl Glittenberg, Florian Zeiler, Christiane Falkner, and Susanne Binder, Ludwig Boltzmann Institut der Krankenanstalt Rudolfstiftung Vienna, Austria; Renate Pflug, Peter K. Ahnelt, Christian Schubert, Elisabeth. M. Anger, and Herbert A. Reitsamer from the Institute of Physiology, Medical University Vienna, Austria; Andreas Stingl and Tuan Le, Femtolasers Produktions GmbH, Vienna, Austria; Michael Stur, Christoph Scholda, Matthias Wirtitsch, Erdem Ergun, Oliver Findl, Ursula Schmidt-Erfurth, Stefan Sacu, Stephan Michels, and Christian Ahlers from the Department of Ophthalmology and Optometry, Medical University of Vienna, Austria.