Quantitative Subanalysis of Cystoid Spaces and Outer Nuclear Layer Using Optical Coherence Tomography in Age-Related Macular Degeneration

Abstract

purpose. To use optical coherence tomography (OCT) to quantify intraretinal cystoid spaces (ICSs) and the outer nuclear layer (ONL) in patients with neovascular age-related macular degeneration (AMD) and to investigate the correlation of these parameters with visual acuity.

methods. StratusOCT (Carl Zeiss Meditec, Inc., Dublin, CA) images were collected from 53 patients receiving their initial treatment with intravitreous ranibizumab. Images were analyzed with custom software (OCTOR) that allows accurate manual segmentation of OCT B-scans and provides thickness/volume measurements of ICS, ONL, neurosensory retina, pigment epithelial detachments (PEDs), subretinal fluid (SRF), and subretinal tissue (SRT). Univariate and multivariate analyses were used to correlate OCT parameters with best corrected Snellen visual acuity. Reproducibility was assessed with weighted κ statistics and intraclass correlation coefficients.

results. A multivariate linear regression model with adjusted R ² showed that ONL volume and SRT thickness significantly correlated with Snellen visual acuity (R ² = 0.15, P = 0.002 and R ² = 0.19, P = 0.001, respectively) with an overall model R ² of 0.34. Adjustment of ONL volume for ICS did not improve correlation with visual acuity, and ICS volume did not independently correlate with visual acuity. Weighted κ statistics showed excellent intergrader agreement for both ICS and ONL measurements.

conclusions. The results suggest that an increased total volume of the ONL is associated with decreased visual acuity in neovascular AMD and that the total volume of ICS does not correlate with visual acuity. Although the correlations detected in this study are modest, quantitative subanalysis of OCT images may be of greater clinical relevance in the context of more advanced OCT technology.

Age-related macular degeneration (AMD) is the leading cause of severe visual loss and blindness in the developed world among people older than 50 years. ¹ Current treatment options for the neovascular form of AMD include thermal laser photocoagulation, photodynamic therapy with verteporfin, and antiangiogenic therapies such as ranibizumab (Lucentis; Genentech, Inc., South San Francisco, CA). ^{2 3 4 5} The efficacy of these treatments is determined primarily by assessment of visual acuity (VA) and secondarily by assessment of fluorescein angiography (FA) and optical coherence tomography (OCT) parameters. ^{2 4 5 6}  

StratusOCT (Carl Zeiss Meditec, Dublin, CA) includes image-analysis software that provides a measure of central retinal thickness, and this parameter has been widely adopted for use in clinical trials. ^{6 7 8} Despite the use of the StratusOCT across the spectrum of macular disease, the complex morphology of choroidal neovascularization (CNV) exposes the limitations of its automated analysis. Errors commonly occur in retinal boundary detection, and the software is unable to provide quantitative information regarding many important features of CNV, such as subretinal fluid (SRF), subretinal tissue (SRT), and pigment epithelial detachment (PED). ⁹ To address these problems, we have developed a software tool (OCTOR) that allows accurate manual segmentation of OCT images and facilitates the quantification of any morphologic area of interest in the neurosensory retina. ^{10 11 12 13 14 15}  

Using the publicly available OCTOR software, we have demonstrated that the presence of increased SRT thickness/volume on OCT, and, to a lesser extent, increased neurosensory retinal thickness/volume, is associated with decreased VA in patients with neovascular AMD. ¹³ The correlations, however, were modest, and only a small proportion of the variability in vision was explained by the retinal volume. To explore potential explanations for the lack of a stronger correlation, we sought to evaluate other morphologic characteristics of the retina in CNV lesions. We hypothesized that the integrity of the outer nuclear layer (ONL) and the presence of cystoid spaces are two features that may influence visual function. Photoreceptor loss, as reflected by a loss of ONL volume could be associated with visual loss. ¹⁶ Accumulation of cystoid spaces, ¹⁷ however, could result in an increased ONL volume, despite the loss of retinal cellular material. 

In this study, we used the OCTOR software to quantify the ONL and cystoid spaces in eyes with active neovascular AMD before therapy and correlated these measurements with VA. 

Methods

Data Collection

A consecutive series of patients receiving their initial treatments with ranibizumab for neovascular AMD were identified from the procedure logs at the Doheny Eye Institute. Data were collected and analyzed in accordance with the policies and procedures of the institutional review board of the University of Southern California and the tenets set forth in the Declaration of Helsinki. 

For inclusion in the study, patients were required to have StratusOCT imaging performed before receiving an initial treatment with intravitreous ranibizumab. Although dense spectral domain OCT (SDOCT) data are now becoming available for patients with neovascular AMD, StratusOCT data were used for this study because of the availability of manual grading tools and the impracticality of manually segmenting boundaries on the large number of B-scans in dense SDOCT volume scans. All images were obtained by using the Radial Lines protocol of six high-resolution B-scans on a single StratusOCT machine. Data for each case were exported to an external hard drive by using the export feature available in the StratusOCT version 4.0 analysis software. 

Each patient’s best corrected VA was recorded at the time of initial diagnosis by using Snellen VA charts. The number and type of any previous treatments for CNV secondary to AMD in the study eye were recorded. Other data collected included patient age and sex, as well as the color photographs and fluorescein angiographic images for each patient. 

Computer-Assisted Grading Software

The software used for OCT analysis, OCTOR, was written by the software engineers at the Doheny Image Reading Center to facilitate manual grading of OCTs. OCTOR (available in the public domain at http://www.diesel.la/ Doheny Eye Institute, Los Angeles, CA) has been described and validated in previous reports. ^{11 15} OCTOR imports data from the StratusOCT machine as a raw data file and allows the user to delineate boundaries of interest in each radial line B-scan (Fig. 1) . After annotation of the desired layers in each B-scan, OCTOR calculates the distance in pixels between the manually drawn boundary lines for each of the various defined spaces and converts these to micrometers to yield a thickness measurement at each location. The thickness of all unsampled locations is interpolated based on a polar approximation, to yield a thickness map analogous to the StratusOCT analysis. Thicknesses are converted into volumes (cubic millimeters) by multiplying the average thickness measurements by the sampled area. The interpolation algorithm, intergrader reliability, and intragrader reproducibility have been reported. ^{11 15}  

Analogous to the StratusOCT software, OCTOR provides a report showing the calculated thickness and volume values for the nine Early Treatment Diabetic Retinopathy Study (ETDRS) macular subfields. The means and standard deviations of the foveal center point (FCP) thickness are also calculated. In contrast to the StratusOCT output, OCTOR provides separate maps for the various macular spaces (e.g., subretinal tissue, subretinal fluid, retina, pigment epithelial detachment, intraretinal cystoid space [ICS], and ONL). 

Grading Protocol

Boundaries drawn in each of the six OCT B-scans for this study included the internal limiting membrane, outer border of the photoreceptors, borders of subretinal fluid and subretinal tissue (if present), inner surface and estimated normal position of the RPE (in cases of RPE elevation), inner border of the ONL, and boundaries of ICSs. All boundaries were drawn in accordance with the standard OCT grading protocol of the Doheny Image Reading Center as previously described, ¹¹ except for the additional boundaries of the ICSs and the ONL, which are described in the following text. After completion of the grading, OCTOR was used to calculate output parameters for the various spaces including the entire neurosensory retina, as well as the ONL alone. 

The ONL was identified as the hyporeflective layer located just internal to the hyperreflective band believed to correspond to the inner segment—outer segment (IS-OS) junction (Fig. 1E) . In some cases, as a result of the destructive effects of the choroidal neovascular process, the IS-OS junction was not reliably identifiable. In those cases, other more anterior layers of the retina (e.g., inner nuclear layer, outer plexiform layer) were used as a reference to localize the ONL. Similar to the grading protocol for other boundaries described in previous reports, the borders of the ONL were initially drawn in the periphery of the B-scans where the disease’s effects are usually least severe and the retinal layers are most easily distinguishable. ¹¹ Additional segments of the ONL boundaries were then drawn in areas where these boundaries could be recognized with the greatest confidence. Discontinuities between segments of ONL boundaries (i.e., areas of uncertainty, usually due to disease) were interpolated by using adjacent areas in which the boundaries were clearly identifiable. 

An ICS was defined as any intraretinal hyporeflective space that was greater than 5 × 5 pixels (Figs. 1B 1C) . Smaller areas were not easily drawn or reliably recognized on successive grading of the same OCT scans and therefore were not included. Tissue septae were often observed within larger cystoid spaces. To provide the most precise assessment of the cystoid space volume, we excluded these septae when drawing the boundaries of the cystoid space. To study the influence of the cystoid spaces, we calculated adjusted ONL and neurosensory retinal volumes by subtracting the ICS volume from the ONL and total retinal volumes, respectively. All boundary determinations were assessed by two graders, and the differences were adjudicated as described in previous reports. 

Statistical Methods

The mean and SD of the FCP thickness, as well as the total volume (subfields 1–9), were calculated for each space in each case. Volume was measured in cubic millimeters and thickness was measured in micrometers. Snellen VA was converted to logMAR VA to facilitate statistical analysis. Univariate and multivariate regression was used to test for associations between logMAR VA, age, sex, and OCT parameters. Commercial programming language (SAS; SAS Institute, Cary, NC) was used for all the analyses. 

The reliability of ONL and ICS grading was determined with weighted κ statistics and intraclass correlation coefficients (ICCs) in a subset of cases (the first 28 consecutive cases included in the study, i.e., more than half of the study cohort), which were graded independently by a second grader. As described in previous reports, the κ statistic is a measure of intergrader concordance on categorical scales that adjusts for chance agreement. ¹¹ The κ statistics were interpreted using the ranges described earlier: 0 to 0.20, slight agreement; 0.21 to 0.40, fair agreement; 0.41 to 0.60, moderate agreement; 0.61 to 0.80, substantial agreement; and >0.80, almost perfect agreement. ¹¹ ICC is a separate measure of correlation between graders that takes into account the differences in individual ratings. Previous studies have suggested that these two statistics present two different types of information regarding agreement and both were used herein to increase the confidence in our assessments. Bland-Altman plots were generated to demonstrate the level of agreement between graders. ¹¹  

Results

Baseline Characteristics

Fifty-three consecutive patients presenting for initial ranibizumab therapy for neovascular AMD met the inclusion criteria for the study. Demographic characteristics of the population included 19 (36%) men and 34 (64%) women. The average age of the patients was 79 ± 7 years (mean ± SD) with a range of 55 to 92 years. Mean VA at the time of initial treatment was 20/100. The neovascular lesions were categorized by fluorescein angiography as retinal angiomatous proliferation (3 eyes, 6%), as predominantly (8 eyes, 15%) or minimally (7 eyes, 13%) classic lesions, or as occult with no lesion (32 eyes, 60%). Baseline characteristics of the study population are summarized in Table 1 . 

Reproducibility Analysis

Twenty-eight consecutive sets of radial line scans (168 scans total), from 28 patients, were graded and analyzed for reproducibility of ONL and ICS grading. Mean total ICS volumes ranged from 0.14 mm³ for grader 1 to 0.10 mm³ for grader 2. Qualitative comparison of discordant cases revealed that the largest differences between graders were evident in the grading of smaller cysts that were eccentrically located, were irregularly shaped, and had poorly demarcated borders. Despite this qualitative difference, weighted κ for the total ICS volume (entire EDTRS grid) was 0.83. Agreement on grading of cystic spaces located in the central subfield was excellent (weighted κ = 1.00). 

Mean total ONL volume ranged from 2.98 mm³ for grader 1 to 3.17 mm³ for grader 2. Qualitative assessment of discordant cases revealed that the largest discrepancies occurred in areas of severe disease, particularly those with extensive SRT and SRF. These areas often had profound loss of both the IS-OS junction and the hyporeflective-hyperreflective boundary between the ONL and outer plexiform layer (OPL). Overall κ statistics for ONL volume in all ETDRS areas was 0.58 with much better agreement in the central subfield (κ = 0.80). Table 2summarizes the reproducibility analysis between the two graders. Bland-Altman plots were generated to show the level of agreement between graders (Figs. 2 3) . 

Univariate Analysis

A univariate linear regression model with adjusted R ² showed no significant correlation between ONL volume and Snellen VA (R ² = 0.06, P = 0.09). Adjustment of ONL volume by subtraction of cystoid space volume did not improve the correlation with VA. Overall retinal volume was significantly correlated with VA (R ² = 0.13, P = 0.008). Adjustment of overall retinal volume by subtraction of ICS volume also did not improve the correlation with VA. Total ICS volume did not independently correlate with VA (R ² = 0.02, P = 0.26). As we have shown previously, SRT volume correlated significantly with Snellen VA (R ² = 0.11, P = 0.01), whereas PED and SRF volume did not (R ² = 0.004, P = 0.64 and R ² = 0.002, P = 0.73, respectively). ¹³ Table 3summarizes the data from the univariate analysis. 

Multivariate Analysis

Two multivariate linear regression models were generated controlling for age and sex. In Model 1, independent variables for stepwise selection were only those variables for which the univariate model had P < 0.10, excluding total retinal volume. In this model, SRT thickness in the central macular subfield and total ONL volume were both significantly correlated with VA (R ² = 0.19, P = 0.001 and R ² = 0.15, P = 0.002, respectively). The model R ² for this analysis was 0.34. In Model 2, independent variables for stepwise selection were similar to those in Model 1 but also included total retinal volume. In this model, SRT thickness and total retinal volume correlated significantly with VA (R ² = 0.19, P = 0.001 and R ² = 0.15, P = 0.001, respectively). The results of the analyses are summarized in Table 4 . 

Discussion

In this report, we describe the association between ONL and cystoid space volume on best-corrected Snellen VA in patients with active neovascular AMD. An increase in total neurosensory thickness/volume was found to correlate modestly with a decrease in Snellen VA. Other investigators have reported similar modest correlations. ¹⁸ We also confirm our previous findings that SRT thickness correlates modestly with a decrease in VA. ¹³ We have hypothesized that consideration of additional morphologic characteristics of CNV lesions on OCT could improve the observed correlations. The ONL, which contains the photoreceptor nuclei, constitutes only a fraction of the overall retinal thickness. Although photoreceptor loss is a well-established mechanism of irreversible vision loss in patients with neovascular AMD, the inner retina in these patients is generally well preserved. ¹⁹ Consequently, small changes in ONL volume from photoreceptor loss may be masked if the whole neurosensory retina is considered for correlation with visual function. On the other hand, accumulation of fluid in cystoid spaces in the retina may thicken the retina without any increase in the cellular content of the retina. Similarly, cystoid spaces may mask the presence of retinal neuronal cell loss if the whole retinal or ONL volume is considered without adjusting for the cystoid spaces. In the present study, however, quantification of ONL alone and adjustment of the ONL and retinal volume by subtraction of the ICS volume did not result in an improved correlation with VA (Tables 3 4 ; Fig. 4 ). 

There are many potential reasons why a better correlation was not observed. The grading of the ONL can often be difficult in the setting of severe disease that destroys landmarks, including the IS-OS junction and the high-contrast boundaries between retinal layers. In our grading of the ONL, we attempted to use the IS-OS junction and the outer border of the OPL to define the ONL anatomically. In most cases, one or both of these landmarks were readily identifiable in some areas, if not most, of the B-scan, allowing interpolation of ONL boundaries in sections where anatomic landmarks are not as evident. Qualitative analysis of the grading shows that there was very good agreement between two independent graders in defining the ONL boundaries. The most significant disagreements between graders occurred in more eccentric scan locations and are reflected in the lower weighted κ for subfields 1 to 9 versus the central subfield alone (0.58 vs. 0.80, respectively). ICC statistics showed a similar trend. Grading of ICS was similarly very reliable overall, but more reliable in the central subfield than in the more eccentric scan locations (weighted κ, 0.83 vs. 1.00, respectively). Based on subsequent qualitative comparisons of discordant cases, this discrepancy appeared to be due to the much smaller size and poorly demarcated boundaries of more eccentric cystoid spaces. The results of these reproducibility analyses suggest, however, that grading reproducibility is not the main factor that explains the lack of correlation between the ONL or ICS and VA. ¹¹  

There are several other potential explanations for the lack of correlation between these OCT parameters and VA. First, the consideration of the ONL and cystoid spaces are only attempts to more precisely quantify the dry retinal volume and provide anatomic evidence of retinal neuronal preservation. In active or recent-onset neovascular AMD, however, vision loss may be due to disruption of photoreceptor function rather than photoreceptor loss. It is possible, that quantification of the ONL and cystoid spaces may show better correlation with vision in patients with eyes with chronic or long-standing CNV. In addition, it is possible that these parameters may be more predictive of visual outcome or prognosis rather than VA at presentation. Second, subtraction of cystoid spaces alone still does not provide a quantification of the true dry retinal volume. For example, fluid exudation into the neurosensory retina could result in diffuse thickening without accumulation in cystoid spaces. In the present study, spaces smaller than 5 × 5 pixels were not considered, as we determined that smaller spaces could not be graded reproducibly. Third, there are probably several other OCT parameters that may have shown better correlation but were not considered. For example, eccentricity of the edema from the foveal center was not accounted for. Fourth, although distance VA did not correlate, it is possible that other parameters of visual functions such as reading speed or contrast sensitivity may have shown better results. 

One additional OCT parameter that may correlate with VA, which was not considered in this study, is the integrity of the IS-OS junction. This parameter may be an early indicator of the health and function of the photoreceptors before the development of frank cellular loss. ^{20 21} Although we were able to use the approximate location of the IS-OS junction to delimit the ONL, we found it difficult to reliably quantify the volume of the IS-OS junction itself, since it was only a few pixels in width. The higher resolution and speed afforded by new SD-OCT technology may eventually facilitate identification and reliable quantification of the IS-OS boundary (as well as the photoreceptor inner and outer segments themselves) in the setting of disease, but better automated segmentation algorithms may be required. 

Finally, there are several potential reasons for lack of correlation, related to the limitations of the study design. For example, only best corrected (frequently pinhole) Snellen VA was used and protocol refractions were not performed. It is of course, well known that Snellen VA measurements are more variable than ETDRS measurements, particularly in the VA range greater than 20/100. ²² Among our 53 patients, 25 had VA better than 20/100, and 28 had vision that was equal to or worse than 20/100. In addition, the retrospective nature of the study introduces the possibility of unknown biases that may have confounded the analysis. Furthermore, the sample size is relatively small, and the study may have been underpowered to detect a relationship. Unfortunately, manual tracing of all the cysts in each B-scan is an extremely time-consuming process and limits the feasibility of conducting a much larger study. Future development of automated algorithms to segment retinal cysts may allow this limitation to be addressed. The OCT technology itself is another limitation of this study. Since the six radial line scans from the StratusOCT were used for this analysis, calculation of cyst volumes required interpolation between scan lines. Fortunately, most of the cysts were in the central macula, thus reducing the extent of interpolation. Nonetheless, interpolation can introduce significant artifacts when considering small structures such as cysts compared with larger structures such as the whole retina. Dense volume scanning with SDOCT technology may help address this problem, but is impractical for manual segmentation because of the large number of B-scans that must be assessed. 

Despite these limitations, our study appears to confirm previous findings of a weak to modest correlation between SRT and neurosensory retina volume and VA. Of note, the correlation between total retinal volume or ONL volume and VA in our statistical models was very similar, which may suggest that ONL volume is the main component of retinal volume that is driving the observed correlation with VA. Although the correlations are weak, they are also in line with results in previous studies of OCT and VA in other diseases. Several studies of eyes with diabetic macular edema have shown that the correlation between FCP and foveal central subfield retinal thickness and VA ranges from 0.08 to 0.54 in diabetic patients. ^{23 24 25 26 27 28 29 30 31 32} Our correlation between VA and retinal thickness in neovascular AMD falls within the range of correlations reported for diabetic macular edema. ³³  

In summary, consideration of ONL and ICSs did not improve the correlation of StratusOCT morphologic parameters with Snellen distance VA in this small series of patients with active neovascular AMD. Further study of these and other parameters may be warranted when automated subanalysis of SDOCT volume scans becomes feasible. 

 Supported in part by National Institutes of Health Grant EY03040 and National Eye Institute Grant R01 EY014375.

 Submitted for publication August 8, 2008; revised December 8, 2008; accepted April 14, 2009.

 Disclosure: A.H. Kashani, None; P.A. Keane, None; L. Dustin, None; A.C. Walsh, Topcon Medical Systems (C, P), Heidelberg Engineering, Inc. (P); S.R. Sadda, Topcon Medical Systems (C, P), Heidelberg Engineering, Inc. (P), Optovue, Inc. (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.

 

OCT B-scan (A) demonstrating ICSs and subretinal tissue. The clinically relevant boundaries (B) were graded with OCTOR (a computer-assisted manual grading tool) software, which then computed the volumes of the spaces defined by these boundaries. Gray shaded areas: demonstration of (C) graded ICSs, (D) total volume of the neurosensory retina, and (E) total volume of the ONL.

Table 1.

 

Baseline Characteristics of Study Population

Mean VA
LogMAR	0.77 ± 0.44
Snellen	20/100
Angiographic classification
Occult with no classic CNV	32 (60)
Predominantly classic CNV	8 (15)
Minimally classic CNV	7 (13)
RAP	3 (6)
Not available	3 (6)
Total eyes	53
Pretreatment
None	26 (49)
Avastin	9 (17)
Macugen	4 (8)
Photodynamic therapy	5 (9)
Multiple treatments	8 (15)
Other	1 (2)
Total	53

 Data are expressed as the number (% of total group), unless otherwise stated. RAP, retinal angiomatous proliferation; multiple treatments, patients who received more than one treatment modality.

Table 2.

 

Intergrader Comparison

OCT Parameter	Grader 1 Mean	Grader 2 Mean	Mean Absolute Difference (Median, Maximum)	ICC	Weighted κ
ICS
Volume (9)	0.04	0.03	0.01 (0, 0.04)	0.96	1.00
Volume (1–9)	0.14	0.10	0.03 (0, 0.17)	0.97	0.83
FCP	72.76	59.56	13.76 (0, 79)	0.98	0.96
ONL
Volume (9)	0.14	0.16	0.02 (0.02, 0.04)	0.94	0.80
Volume (1–9)	2.98	3.17	0.22 (0.19, 0.90)	0.81	0.58
FCP	211.40	228.08	25.16 (26, 76)	0.95	0.77

Figure 2.

 

Bland-Altman plots demonstrating intergrader reproducibility of ICS grading in the foveal central subfield (A) and entire ETDRS grid (B). Dashed lines: 95% limits of agreement.

Figure 3.

 

Bland-Altman plots demonstrating intergrader reproducibility of ONL grading in foveal central subfield (A) and entire ETDRS grid (B). Dashed lines: 95% limits of agreement.

Table 3.

 

Univariate Correlation between OCT Parameters and Snellen Visual Acuity

OCTOR Measured Parameter	Mean ± SD	Correlation Coefficient (r)	Model R 2	P
ICS
FCP	53.85 ± 103.84	0.14	0.02	0.29
Total volume	0.25 ± 0.84	0.14	0.02	0.26
ONL
FCP	198.66 ± 98.28	0.08	0.007	0.54
Total volume	3.12 ± 0.64	0.24	0.06	0.09
ONL (adjusted for ICS)
FCP	144.81 ± 65.66	0.1	0.01	0.45
Total volume	2.87 ± 1.01	0.00	0.00	0.89
Neurosensory retina
FCP	276.79 ± 132.60	0.37	0.14	0.007
Total volume	7.67 ± 1.06	0.37	0.13	0.008
Neurosensory retina (adjusted for ICS)
FCP	222.94 ± 114.21	0.3	0.09	0.03
Total volume	7.42 ± 1.32	0.2	0.04	0.17
PED
FCP	102.40 ± 130.96	0.2	0.04	0.16
Total volume	3.19 ± 17.10	0.06	0.004	0.64
Subretinal tissue
FCP	48.98 ± 54.33	0.36	0.13	0.009
Total volume	0.25 ± 0.48	0.33	0.11	0.01
Subretinal fluid
FCP	21.09 ± 39.66	0.00	0.00	0.95
Total volume	0.24 ± 0.38	0.04	0.002	0.73

 Data are expressed as the mean ± SD (FCP, micrometers; total volume, cubic millimeters).

Table 4.

 

Multivariate Correlation between OCT Parameters and Snellen Visual Acuity

OCTOR Measured Parameter	Model 1	Model 2
ICS	NC	NC
ONL	Correlated	NC
Volume ETDRS 1–9	R 2 = 0.15, P = 0.002	NC
ONL (adjusted for ICS)	NC	NC
Neurosensory retina		Correlated
Total volume	Excluded	R 2 = 0.15, P = 0.001
(adjusted for ICS)	Excluded	NC
PED	NC	NC
Subretinal Tissue	Correlated	Correlated
ETDRS 9	R 2 = 0.19, P = 0.001	R 2 = 0.19, P = 0.001
Subretinal Fluid	NC	NC
Overall R 2	0.34	0.44

 NC, no correlation.

Figure 4.

 

(A) ONL versus logMAR Snellen VA. (○) Uncorrected ONL measurements; (×) ONL adjusted for ICS (ONL – ICS). (B) Total retinal volume (TRV) versus LogMAR Snellen VA. (○) Uncorrected TRV measurements; (×) TRV adjusted for ICS (TRV – ICS).