New OCT Devices: Implications for AMD Management
SriniVas R. Sadda, M.D.
Associate Professor of Ophthalmology
Director, Medical Retina Unit, Ophthalmic Imaging Unit, Doheny Image Reading Center
Doheny Retina Institute
Doheny Eye Institute
Keck School of Medicine
University of Southern California
The past year has witnessed the introduction of new instruments from a variety of vendors which are said to herald a new era in optical coherence tomography (OCT) and a “golden era” in retinal imaging. These devices, which carry a variety of names, all fall under the rubric of “Fourier Domain OCT” (FDOCT) technologies. Other synonyms which the ophthalmologist may encounter include “Spectral Domain OCT”, “Frequency Domain OCT”, “3D-OCT”, and “High Definition OCT.” These names all refer to devices which represent a higher-resolution, but more importantly, higher-speed, successor to conventional time domain OCT.
Since its first description in 1991 by Huang and coworkers1, time domain OCT has revolutionized the practice of ophthalmology and, in particular, the diagnosis and management of patients with retinal disease. The latest iteration of time domain OCT, the Stratus OCT (Carl Zeiss Meditec, Dublin, CA), has become an integral component of most retina practices. Among its many capabilities, OCT provided ophthalmologists with two important new functions: cross-sectional depiction of retinal disease and quantification of retinal morphology. Instead of qualitative descriptors of disease severity such as “mild”, “moderate”, and “severe,” clinicians were able to characterize the status of their patient’s disease with much greater precision. OCT-derived measurements rapidly became incorporated into clinical trials and clinical practice for monitoring response to therapy, particularly for patients with macular edema secondary to retinal vascular diseases such as diabetes.2-4 Clinicians, however, looked to broaden the application of OCT to other diseases, including neovascular age-related macular degeneration (AMD).5-6 For example, a MedLine search reveals that in the past 18 months, over 80 peer-reviewed articles have been published describing quantitative data from OCT in assessing patients with AMD. Furthermore, in the PrONTO study,7 OCT data were utilized to guide retreatment decisions in patients receiving anti-angiogenic therapy for neovascular AMD. With the widespread utilization of time-domain OCT, the limitations of the technology began to become more apparent. Several investigators reported a high frequency of artifacts in the retinal thickness maps produced by the Stratus OCT machine.8 Some of these artifacts relate to failure of the Stratus OCT software to properly identify the inner and outer retinal boundaries in some cases, thereby impeding the ability to calculate retinal thickness accurately. In a study of 200 cases of retinal disease, we observed boundary detection errors in 92% of cases, with the highest frequency of severe errors in eyes with AMD.9 This observation likely reflects the fact that the original boundary detection algorithms for OCT were designed for diabetic macular edema and not for cases with more complex tissue interfaces such as AMD. Although segmentation errors account for some of the retinal thickness map artifacts, many appear to be related to the speed of time domain OCT.
All OCT technology relies on the use of a near infrared light source, most commonly produced by a superluminescent diode (SLD). The emitted light reflects from the numerous tissue interfaces in the eye, with the intensity of the reflection dependent on various optical characteristics including the difference in refractive index of the materials/tissue at either side of an interface (e.g. the vitreous – retina interface). In order to identify the depth or “axial location” of a particular reflection, time domain OCT relies on a moving mirror within the OCT instrument which also reflects the light emitted by the SLD. The detector and image processing components of the OCT device effectively compare the time it takes for the light to reflect from the mirror to the time it takes to reflect from the various interfaces in the eye in order to reconstruct the depth information and construct an “A-scan.” Transverse movement of the light source allows multiple adjacent A-scans to be acquired to generate a “B-scan.” Because of the mirror movement required, time domain OCT takes time; for example, the Stratus OCT instrument may only be able to acquire 400 A-scans in one second, or take 1.25 seconds to acquire one high-resolution B-scan. Because a patient’s eye, particularly in patients with poor fixation due to disease, moves during scan acquisition, A-scans which are depicted on B-scans and thickness maps to be in adjacent locations may not be in close proximity at all. This problem undermines one’s confidence in OCT thickness measurements (especially metrics relying on the averaging of only a few points such as the foveal center point thickness) and in comparing these measurements over time. As an example, we recently performed a Stratus OCT on a patient who was being considered for a clinical trial requiring a foveal center subfield thickness over 300 microns. The first thickness map from the patient measured 249 microns, whereas a second map obtained only a minute later measured 302 microns. Inter-measurement reliability in retinal disease is only one limitation imposed by time domain OCT. Another related problem is poor registration. Registration refers to the ability to know that a given point on an OCT B-scan corresponds to a particular location in a patient’s fundus. Registration is important for comparing B-scans obtained at different visits (e.g. assessing the amount of subretinal fluid), but also in correlating findings on clinical examination with structures noted on OCT. A final critical limitation related to the slow-speed of time domain OCT is poor sampling density. In constructing a “retinal thickness map”, the 6 radial line B-scans actually sample less than 5% of the mapped area – the remainder is interpolated. More dense sampling is not feasible with time domain OCT, as even longer acquisition times will further exacerbate errors related to fixation. As a result, important pathologic findings, such as a pocket of subretinal fluid or a pigment epithelial detachment (PED), may be completely missed between the B-scans.
FDOCT addresses many of these critical limitations of time domain OCT and promises dramatic improvement in our management of patients with retinal disease such as AMD. Although not the chief benefit of FDOCT, most of the new devices on the market also offer improvements in axial resolution. The stated axial resolution of Stratus OCT is around 8-10 microns, whereas the resolution of the currently available FDOCT instruments ranges from 4-7 microns. The improved axial resolution allows visualization of finer structures, which is of value in assessing the outer retina and the integrity of the photoreceptors in patients with AMD. For example, we have identified many patients with unexplained vision loss or poor visual outcomes following therapy who were found to have subtle disruption of the photoreceptors. These findings were missed on time domain OCT. The improved resolution and overall better sensitivity may also improve the success of segmentation algorithms for detecting boundaries of retinal structures accurately – though this remains to be demonstrated.
The primary advantage of FDOCT over time domain OCT is its speed. Instead of a moving mirror to assess depth, FDOCT devices use a spectrometer and Fourier mathematics to extract the depth information. In essence, an “A-scan” is acquired “all at once,” thus dramatically increasing the acquisition speed. The current FDOCT devices on the market tout scanning rates of 18,000 – 40,000 A-scans/second – a 45-100 fold advantage over Stratus OCT. The high speed of FDOCT has several implications for imaging in patients with diseases such as AMD. High speed A-scan acquisition allows dense sampling of the macula. This eliminates concerns that pockets of subretinal fluid, PEDs, or areas of retinal edema will be missed. Identifying these abnormalities may be critical in diagnosing possible choroidal neovascularization (CNV) in patients whose angiographic findings may be equivocal. Perhaps even more important, detecting these areas of exudation may be essential to clinicians who are utilizing OCT findings to guide retreatment decisions for patients undergoing ranibizumab or bevacizumab therapy for CNV. We have observed numerous cases in which subretinal fluid was absent on Stratus OCT but detected by FDOCT imaging. These observations raise concerns that reliance on Stratus OCT findings could result in “under-treatment” of some patients with neovascular AMD.
Another major benefit of the high speed of FDOCT devices for AMD management is improved registration. The rapid and dense acquisition of OCT data facilitates alignment of the B-scans with other fundus images (infrared or color) acquired at the time of the OCT scan acquisition. One FDOCT instrument, the Spectralis (Heidelberg Engineering, Vista, CA) device, takes this one step further by using a second laser to track the position of the retina in real-time to insure B-scan acquisition at the precisely desired location. The enhanced registration capabilities of FDOCT allow clinicians to compare data between visits with much greater confidence. This capability is of great value in determining, for example, whether a patient’s retinal edema or subretinal edema is responding to therapy. In the near future, advances in OCT segmentation software will allow clinicians to have access to measurements of specific structures of interest, such as volumes of subretinal fluid, retinal pigment epithelial detachments, and retinal cysts. This will allow better monitoring of therapeutic response.
Precise registration is also valuable in correlating fundus findings with OCT features. An instrument such as the 3D OCT 1000 (Topcon Medical Systems, Paramus, NJ) provides a near-simultaneous color fundus image that facilitates this comparison. For example a user can simply “click” on a fundus finding they believe may be an area of retinal angiomatous proliferation and then see if it in fact corresponds to an area of intraretinal hyper-reflectivity on the B-scan. Similarly, the various FDOCT devices allow the user to utilize the registration capabilities to align the OCT data with other imaging or diagnostic modalities. For example, the OCT/SLO (Opko Health, Miami, FL) instrument also allows the user to correlate OCT findings with functional data from microperimetry. The Spectralis allows correlation with fundus autofluoresence and angiographic images. All of these tools and capabilities are useful to the clinician in better understanding their patients’ disease processes.
Another benefit of FDOCT and the dense acquisition of image data is the ability to generate three dimensional renderings of the disease morphology. This is particularly useful in surgical planning for vitreo-retinal interface disease, but can also aid the clinician in understanding AMD. For example, areas of retinal angiomatous proliferation can be readily identified. In addition, 3-D maps can be a useful tool in patient education.
Although the benefits of FDOCT are significant and FDOCT has already become an essential tool in our practice, these devices are not without some negatives. In addition to the additional capital investment required to obtain the device, the tremendous amount of data acquired by the instruments presents several challenges. First, clinicians must be able to view the data in a rapid and efficient fashion if the instruments are to be successfully incorporated into their practice environment. This requires the instruments to have an ergonomic interface and to be network capable. Second, the large size of the dataset has implications for data transfer as well as storage. Fortunately, with advances in computing and networking, these issues are solvable.
In summary, FDOCT instruments represent a revolutionary advance in imaging technology for retinal disease. The benefits of FDOCT have important implications in our management of patients with AMD.
- Huang, D., et al., Optical coherence tomography. Science, 1991. 254(5035): p. 1178-81.
- Brown, J.C., et al., Detection of diabetic foveal edema: contact lens biomicroscopy compared with optical coherence tomography. Arch Ophthalmol, 2004. 122(3): p. 330-5.
- Panozzo, G., et al., Role of OCT in the diagnosis and follow up of diabetic macular edema. Semin Ophthalmol, 2003. 18(2): p. 74-81.
- Panozzo, G., et al., Diabetic macular edema: an OCT-based classification. Semin Ophthalmol, 2004. 19(1-2): p. 13-20.
- Arias, L., et al., Optical coherence tomography analysis of a randomized study combining photodynamic therapy with intravitreal triamcinolone. Graefes Arch Clin Exp Ophthalmol, 2007.
- Coscas, F., et al., Optical coherence tomography identification of occult choroidal neovascularization in age-related macular degeneration. Am J Ophthalmol, 2007. 144(4): p. 592-9.
- Fung, A.E., et al., An optical coherence tomography-guided, variable dosing regimen with intravitreal ranibizumab (Lucentis) for neovascular age-related macular degeneration. Am J Ophthalmol, 2007. 143(4): p. 566-83.
- Ray, R., S.S. Stinnett, and G.J. Jaffe, Evaluation of image artifact produced by optical coherence tomography of retinal pathology. Am J Ophthalmol, 2005. 139(1): p. 18-29.
- Sadda, S.R., et al., Errors in retinal thickness measurements obtained by optical coherence tomography. Ophthalmology, 2006. 113(2): p. 285-93.