Abstract
purpose. To assess the application of optical frequency domain imaging (OFDI) at 1050 nm for the detection of choroidal neovascularization (CNV) in age-related macular degeneration (AMD) and its response to treatment. Three patients presenting with blurred vision and exudative AMD were imaged before and after anti-VEGF treatment with ranibizumab.
methods. The patients were imaged with OFDI, a swept-source–based, high-speed optical coherence tomography (OCT) system developed at the Wellman Center for Photomedicine. A center wavelength of 1050 nm was used that has been demonstrated to provide better imaging of the deeper structures of the retina below the RPE, such as the choroidal vasculature. Three-dimensional data sets were acquired in 2 to 4 seconds.
results. En face images were compiled from cross-sectional OFDI data and correlated with color fundus photography (CF) and fluorescein angiograms (FAs). Cross-sectional images were coregistered with CF and FA to obtain depth-resolved information about CNV, CNV volume, retinal thickness, subretinal fluid volume and height of neurosensory detachment before and after treatment with ranibizumab. A band of reduced reflectivity below the RPE was identified in all three subjects that corresponded to areas of confirmed and suspected occult CNV on FA. After treatment, this band was reduced in volume in all patients.
conclusions. High-speed 3-D OFDI at 1050 nm is a promising technology for imaging the retina and choroid in neovascular AMD. The developed system at 1050 nm provides good contrast for occult (type 1) CNV and may have advantages compared with time domain and current state of the art spectral domain OCT systems (SD-OCT) at 850 nm.
Optical coherence tomography (OCT) is a noninvasive, high-resolution technique capable of acquiring cross-sectional images of tissue morphology.
1 In ophthalmology, OCT is already a well-established clinical imaging method for the investigation of physiological properties and diseases of the eye.
2 3 4 5 Until recently, almost all commercially available systems used in retinal OCT imaging were based on first-generation time-domain OCT (TD-OCT) technology,
1 2 6 which has limited speed and sensitivity. More recently, spectral, or Fourier domain–based OCT systems (SD/FD-OCT)
7 have been developed to overcome limitations prevalent in TD-OCT.
8 9 10 11 As a result, SD-OCT achieves a 2- to 3-order-of-magnitude increase in sensitivity,
12 13 14 translating to a nearly 100-fold faster imaging method that achieves video rate speeds without sacrificing image quality.
11 15 16 Operating at higher speeds permits acquisition of three-dimensional (3-D) data sets
11 17 18 and simultaneous ultrahigh speed and ultrahigh resolution.
15 16 These improvements accumulatively result in images with more detailed information of the intraretinal layers, including the ganglion cell layer, photoreceptor layer, plexiform layers and nuclear layers.
Because of the design of the spectrometer used in SD-OCT, the effective ranging depth is limited by a depth-dependent sensitivity decay of approximately 6 dB over 1 mm.
11 An alternative technique to SD-OCT is optical frequency domain imaging (OFDI).
19 In OFDI, a rapidly tuned laser source is used and the spectrally resolved interference fringes are recorded as a function of time in the detection arm of the interferometer. Published results in healthy volunteers have shown that OFDI has better immunity to sensitivity degradation due to lateral and axial eye motion and has an effective ranging depth that is 2 to 2.5 times better then SD-OCT (depth-dependent sensitivity decay of 6 dB over 2 to 2.5 mm).
19 20 21
More important, recent research at a 1050-nm spectral range has demonstrated a better retinal penetration depth,
20 22 23 particularly important for detecting retinal abnormalities at or below the retinal pigment epithelium (RPE). A wavelength of 1050 nm has less attenuation from scattering in opaque media, commonly seen in patients with cataract.
24 Although the water absorption at 1050 nm is higher than in the 850-nm region, the difference is partially compensated for by the approximately three times higher maximum permissible exposure according to the ANSI standards (1.9 mW at 1050 nm).
25
Age-related macular degeneration (AMD) is one of the leading causes of vision loss in people over age 65 in Western countries.
26 27 The gold standard methods for the diagnosis of exudative AMD include color fundus photography (CF) and fluorescein angiography (FA). These methods provide detailed information about the en face location and global dimension of leaking macular abnormalities, but stereo FA and CF images donot offer optimal depth information.
28 TD-OCT has been compared with FA for the detection of choroidal neovascularization (CNV) in AMD.
29 Although the sensitivity of TD-OCT is high, the specificity is not sufficient for it to supplant FA. TD-OCT systems, however, can quantify retinal thickening and subretinal fluid. This has been useful in following the results of treatment of neovascular AMD.
30 31 Classic CNV has been visualized with both TD-OCT and SD-OCT.
32 33 Small studies have demonstrated regression of classic (presumably type II, or subretinal) CNV with anti-VEGF treatment.
34 A recent study by Coscas et al.
35 demonstrated that limited or complete RPE detachments could be found in 98% of patients by TD-OCT, often corresponding with areas of occult leakage on FA or ICG. In the images presented in that paper, it is difficult to see Bruch’s membrane. Drexler et al.,
36 in a study of a new ultrahigh-resolution OCT system, presented one case of occult neovascularization. In their images, Bruch’s membrane could be readily seen.
In this study we present the first application of OFDI at a longer wavelength (1050 nm) for the study of AMD. We hypothesized that the better penetration depth and deeper effective ranging depth of this instrument is particularly important for detecting retinal abnormalities expressed in patients with AMD, such as pigment epithelium detachment (PED) and visualizing occult (type 1) CNV below the RPE.
An ultrahigh-speed OFDI system was developed that constituted a nearly twofold speed improvement to 30,000 A-lines/second over the system described in detail in a previous publication regarding the retinal OFDI system.
20 Patients underwent CF, FA, TD-OCT (Stratus; Carl Zeiss Meditec, Dublin, CA) and OFDI imaging. After anti-VEGF treatment, OFDI and color fundus photography were repeated. 3-D-OFDI images were coregistered with CF and FA images. Sub-RPE fluid/CNV thickness and volume, retinal thickness, subretinal fluid volume, and the presence of the photoreceptor layer were evaluated before and after anti-VEGF treatment.
Patient 1 was a 70-year-old man with a history of mild nonproliferative diabetic retinopathy, cataract extraction with intraocular lens implantation in the right eye, and previously non-neovascular AMD who presented with a chief complaint of increasingly blurred vision in the right eye. Visual acuity with spectacle correction in the right eye measured 20/50 by ETDRS (Early Treatment Diabetic Retinopathy Study) chart. Comprehensive examination including fundus biomicroscopy, FA, and OCT imaging (Stratus; Carl Zeiss Meditec) of the macula revealed CNV in the affected eye. OFDI was then obtained, followed by nine total treatments with intravitreous ranibizumab (Genentech, Inc., San Francisco, CA) over the ensuing 10 months. Repeat imaging by OFDI was performed immediately before the ninth injection, at which time the visual acuity in the affected eye measured 20/32-1 by ETDRS chart.
Pretreatment.
Posttreatment.
Patient 2 was a 78-year-old man with an ocular history of moderate nuclear sclerotic cataracts and non-neovascular AMD who presented with new-onset vision loss and distortion in the right eye. Visual acuity in this eye measured 20/50-2 without correction by ETDRS chart. Results with examination with fundus biomicroscopy, FA, and OCT were consistent with CNV in the affected eye. OFDI was then performed, followed by monthly intravitreous injections with ranibizumab for a total of four treatments. After treatment, visual acuity of the right eye stabilized at 20/50, and there was a marked reduction in CNV-associated exudative changes, as seen on fundus biomicroscopy and confirmed by OCT. The patient persisted at this level, and treatment was withheld during a 6-month period of close observation, at the conclusion of which OFDI imaging was repeated.
Pretreatment.
Posttreatment.
Pretreatment.
Posttreatment.
High-speed OFDI at 1050 nm is a promising technology for imaging the retina and choroid in AMD. The developed system provides a large effective depth range (up to 2.5 mm in air), and previous work suggests that a center wavelength at 1050 nm provides better penetration below the RPE than does the 850-nm wavelength.
20 22 This method could eventually result in a better detection and quantification of sub-RPE changes, particularly in exudative AMD. The sub-RPE changes in exudative AMD demonstrated in our cases, as well as those previously reported, could represent type I CNV, sub-RPE fluid, or both. The images showed good definition of the choroidal vasculature, except in the areas of substantial subretinal blood accumulation. Three-dimensional imaging and rendering gave valuable additional volumetric information on CNV, subretinal fluid, and retinal edema. The low variability in the repeated volume measurements suggests good reproducibility of the segmentation techniques. The resultant en face images demonstrated excellent correlation with CF and FA images. Comparisons of the pre- and postvolumetric datasets showed interesting changes after treatment and made it possible to determine the effectiveness of treatment in the future. CNV thickness maps and corresponding 3-D reconstructions have been compared for both pre- and posttreatment scans and are shown in the
Supplementary Document.
In Patient 1, the cross-sectional images were coregistered with the CF and FA images before and after treatment. The locations in the pretreatment dataset with CNV are showing small well-defined structures beneath the RPE layer, and the RPE layer itself shows a significant increase in contrast in the image
(Fig. 3E.I) . Other striking features in the
Figure 3Ecross-sectional images are the existence of cystic changes, drusen, blood clots, RPE detachment, and a large subretinal fluid pool. This fluid pool appears in a large part of the scanned area, and in
Figure 3E.IIthe changes in the photoreceptor layer, which is located above the fluid pool, are clearly visualized. We hypothesize that the reduced scattering at the point of highest elevation of the subretinal fluid is a signature of the absence of the photoreceptors or their outer segments in
Figure 3E.II , feature F, due to prolonged separation from the RPE, whereas the strong scattering in the periphery (
Fig. 3E.II , feature G) indicates the presence of the photoreceptor layer. The scattering between the RPE and Bruch’s membrane could represent CNV and/or fluid. The segmentation of this structure for patient 1 included the entire area of RPE elevation (i.e., the volume of the RPE detachment and the region between the RPE and Bruch’s below the hemorrhage, by extrapolating the extension of Bruch’s membrane). The volumetric amount of subretinal fluid and presumed CNV and/or fluid was calculated from the 3-D rendering of the segmentation. The posttreatment data set showed dramatic change in elevation suggesting reduction of subretinal fluid and blood. The apparent CNV and/or fluid volume was reduced by a mean of 46% ± 7.8% (see
Table 1 and the
Supplementary Document).
In patient 2, the cross-sectional images were coregistered with the CF and FA images before and after treatment. In the pretreatment dataset, the OFDI scans showed a membrane suspected to be CNV between the RPE and Bruch’s membrane at different locations of the cross-sectional images. The cross-sectional images showed disrupted RPE layers at the suspected CNV locations. The appearances of pigment accumulations were also clearly visible in all the images. A highly scattering structure compatible with type II CNV is observed in
Figures 4E.IIand 4E.III . In
Figure 4E.III , a possible breakthrough of the RPE and cystic changes are observed, and in
Figure 4E.IVsubtle subretinal fluid is visible. The volume for the retinal edema, the subretinal fluid, and CNV was calculated from the 3-D rendering of the segmentations. The posttreatment dataset showed, besides the reduction of subretinal elevation and subretinal cysts, a reduction of the CNV volume by a mean of 73% ± 3.9% (see
Table 1 and the
Supplementary Document). A dark band was visible in the cross-sectional images and the en face image, indicating change in scattering at the location where CNV had been detected. We believe this band was due to overlying atrophy of the RPE.
Patient 3 had exudative AMD with obvious, though occult CNV. Correlation of OFDI images with CF and FA at the suspected areas showed changes between the RPE band and Bruch’s membrane that correspond to occult CNV. The cross-sectional images of the pretreatment dataset consisted of 512 A-lines and therefore had a noisier appearance, which can also be partly due to poor focusing of the eye lens in this patient. The rendered 3-D volume of this dataset showed a large, oval, presumed area of CNV, with good correlation with the oval-shaped diffuse fluorescence in the FA images (
Figs. 5E.I 5E.II 5E.III , feature C). The large cystic edema compartment was visible only in the cross-sectional images and the rendered en face 3-D volume (
Figs. 5E.II 5E.III , feature D). This limitation was also the case for the retinal edema (
Fig. 5E.I 5E.II 5E.III , feature A). Furthermore, the scanned area did not completely cover the area of retinal edema due to the limitations of the hardware in the slit lamp. The different volumes could be calculated for the retinal edema and the presumed CNV area. Again the posttreatment dataset showed a marked change in subretinal elevation. The retinal edema had completely disappeared. The presumed CNV was still visible, covering approximately the same area, but had decreased in volume by a mean of 42% ± 3.3% (see
Table 1 and the
Supplementary Document).
The changes demonstrated between the RPE band and Bruch’s membrane could represent fluid or type I CNV, or both. We doubt that they represent hemorrhage, given the lack of characteristic color on fundus photographs. The variable extent of regression of these volumes after anti-VEGF treatment, in contrast to the virtually complete resolution of retinal edema and subretinal fluid, suggests that at least some of the volume represents type I CNV.
In conclusion, we demonstrated high-speed 3-D imaging of exudative AMD with occult CNV before and after treatment, using OFDI at 1050 nm. We believe that these images demonstrate occult, type I CNV below the RPE. This relatively new OCT technology in combination with the wavelength in the 1-μm region can have a valuable contribution to early AMD research and as a follow-up treatment-imaging tool.
Contributed equally to the work and therefore should be considered equivalent authors.
Presented in part at the annual meeting of the Association for Research in Vision and Ophthalmology, Fort Lauderdale, Florida, May 2007, and at the International Society for Optical Engineering meeting (SPIE), San Jose, California, January 2007.
Supported in part by Grant R01 RR019768 from the National Institutes of Health, Grant R01 EY014975 from the National Eye Institute, and Grant F4 9620-01-1-0014 from the Department of Defense.
Submitted for publication December 4, 2007; revised February 19, 2008; accepted July 31, 2008.
Disclosure:
D.M. de Bruin, None;
D.L. Burnes, None;
J. Loewenstein, None;
Y. Chen, None;
S. Chang, None;
T.C. Chen, None;
D.D. Esmaili, None;
J.F. de Boer, NIDEK, Inc. (F, P, R)
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.
Corresponding author: Johannes F. de Boer, VU University, De Boelelaan 1081, FEW T0.67, 1081 HV Amsterdam, The Netherlands;
[email protected].
Table 1. Pre- and Posttreatment Segmentation Volumes of CNV, Subretinal Fluid and Retina Edema for Patients 1–3
Table 1. Pre- and Posttreatment Segmentation Volumes of CNV, Subretinal Fluid and Retina Edema for Patients 1–3
| Patient Scan | Mean Segmentation Volumes (mm3) | | |
| Choroidal Neovascularization | Subretinal Fluid | Retinal Edema |
| Patient 1 | | | |
| Pretreatment | 0.238 ± 0.0107 | 0.961 ± 0.0294 | N/A |
| Posttreatment | 0.127 ± 0.0139 | 0.00314 ± 0.000171 | N/A |
| Patient 2 | | | |
| Pretreatment | 0.231 ± 0.0182 | 0.0620 ± 0.00740 | 0.136 ± 0.00969 |
| Posttreatment | 0.0617 ± 0.00493 | 0 | 0 |
| Patient 3 | | | |
| Pretreatment | 0.479 ± 0.00742 | N/A | 1.93 ± 0.133 |
| Posttreatment | 0.277 ± 0.0149 | N/A | 0 |