Abstract
Purpose:
To detect and localize subtle changes in retinas of diabetic patients who clinically have no diabetic retinopathy (DR) or non-proliferative DR (NPDR) as compared to age- and sex- matched controls. Spectral Domain Optical Coherence Tomography (SD-OCT) and software to examine all retinal layers, including deeper layers, were used to quantify foveal avascular zone size and inner and outer retinal layer thicknesses, as well as detect axial location of prominent lesions.
Methods:
Diabetic subjects, 19 total with 16 having no DR and 3 having non-proliferative retinopathy, were matched with 19 controls with respect to age and sex. Macular-centered SD-OCT grids of 20×15 deg were taken with the Spectralis. En face or transverse images were generated from the SD-OCT data by automatically segmenting all retinal layers. The transverse images were investigated for foveal avascular zone (FAZ) size, retinal vessel caliber, and structural changes. The size of the FAZ was compared for diabetics versus controls using vendor software and manual marking in Photoshop. Inner retinal layer (IRLFAZ) and outer nuclear layer (ONLFAZ) thicknesses at the margins of the FAZ were measured using vendor software.
Results:
The FAZ area was larger for diabetics (mean ± SD = 0.388 ± 0.074 mm2) than controls (0.243 ± 0.113 mm2), t(18) = 5.27, p < 0.0001, using vendor software. The mean IRLFAZ was thicker for the diabetics (86.8 ± 14.5 µm) than controls (65.2 ± 16.3 µm), t(18) = 4.59, p = 0.00023, despite lack of exudation by clinical exam. There was no significant association between FAZ area and mean IRLFAZ for the diabetics, r = 0.099, p = 0.69. Vessels not clinically detected were visible in the NFL transverse image of most diabetics, especially for a mild NPDR patient. A prominent lesion found in the en face infra-red image of a mild NPDR subject was localized in the photoreceptor layer by SD-OCT, as well as additional outer retinal changes in other subjects.
Conclusions:
Our results demonstrate changes in inner and outer diabetic retinas not readily detectable by clinical exam. IRLFAZ had not thinned at the margins of the large FAZ’s, indicating neural mass did not yet decrease despite potential ischemia.
Keywords: foveal avascular zone, diabetic retinopathy, diabetic macular edema, transverse, imaging, optical coherence tomography
INTRODUCTION
Diabetic retinopathy (DR) and diabetic macular edema (DME) are the primary causes of vision loss in working age adults, with cases of diabetes-related vision loss expected to increase significantly because the global prevalence of diabetes is increasing at epidemic proportions.1–3 Using the US as an example of a multicultural and industrialized country, the Center for Disease Control and Prevention reports that nearly 24 million Americans have diabetes, and about a quarter (5.7 million) of them do not know that they have the disease.4 Pathological retinal vascular and neural changes are the main characteristics of these sight threatening retinal complications of diabetes.5
The initial clinical classification of DR is based on the superficial retina and larger retinal vascular changes that lead to the leakage of fluids and lipids, and including growth of new vessels on the disc or elsewhere in the retina.6,7 However, most grading schemes of DR have not yet evolved to use modern improved imaging modalities as most of these grading schemes rely on fundus photography, which originally used flood illumination and lacked scanning techniques as well as a confocal aperture to improve contrast.3
Inner retinal changes are well documented in DR and DME.8–11 Leakage of fluids and lipids, and the formation of cystoid structures, are usually more visible for the inner retina.8,9 In infrared (IR) images or color fundus images, the leakage of such substances into the inner retina may cause reflectivity changes that may obscure the visualization of underlying retinal structures. Advanced imaging techniques including optical coherence tomography (OCT) and adaptive optics scanning laser ophthalmoscopy (AOSLO) provide improved resolution and contrast for detecting DR and DME.12-14 OCT detects cysts in many layers of diabetic retinas including the inner nuclear layer and provides thickness of the retinal layers distorted by the cystoid structures. 11 In patients clinically classified as not having DME, OCT detects DME using retinal layer analysis.13 A diabetic retina could have dying neurons that lead to retinal thinning at the same time as fluid leakage is leading to retinal thickening.5,10 Thus, OCT feature-based classification schemes are still emerging. Unanticipated capillary remodeling including the growth of capillary loops and tangles are found in diabetics having only mild or moderate non-proliferative diabetic retinopathy (NPDR) using AOSLO.12 The retinal parafoveal capillary network is disrupted before the onset of DR for patients with type 2 diabetes using AOSLO.14 Thus, there are extensive retinal changes much worse than reported by the clinical classification of DR.
Disruption of parafoveal capillary network and hence a large FAZ size has been found in diabetic patients prior to the clinical detection of DR.14–24 In diabetic patients without retinopathy, significant changes in the perifoveal intercapillary areas and larger FAZ size than in age-matched controls are seen with fluorescein angiography (FA).15–21 Severity of disease is associated with larger FAZ, shown using FA.15–21 OCT angiography (OCTA) studies also show that diabetic patients without clinical DR have a significantly enlarged FAZ compared to healthy controls.22–24 Focal capillary dropout around the FAZ and hence an enlargement of the FAZ size is also found in diabetics without retinopathy using AOSLO.14 Boundaries between inner retinal layers are also disrupted in diabetic patients with current or resolved edema.11
Tissue oxygenation of the inner retinal layers is provided by the inner retinal capillaries, and the relatively constant inner retinal layer thickness at the margins of the FAZ has been attributed to the limits of the metabolic support by these capillaries.25 Hence there may be a biological penalty to the neurons at the FAZ margins as the FAZ sizes increase in diabetic patients without retinopathy.
Outer retinal changes are also well documented in DR and DME.26–29 Compromise of the intact retinal pigment epithelium (RPE) layer that serves as the outer retinal blood barrier may lead to fluids from the choriocapillaris moving into the outer retina causing DME and in the process damaging the outer retinal neurons, i.e. photoreceptors.26 Such changes may not be visible on clinical fundus photography. In cases where they are visible, their axial location cannot be pinpointed. When viewing a digital color image of the macula, the red and green color channels can be separated to detect the presence of cystoid structures and hyperreflective foci, which for the outer retina are more visible in the red than in the green channel and are consistent with OCT findings in diabetic patients with DME.27 Thus, emphasis on damage to the outer blood retinal barrier, including the RPE, in DME may be imperative.3,28-34 Further, the significant impact on sight is related to the structures in deeper layers, including the deeper cysts being related to worse visual acuity,9 and photoreceptor integrity indicating which eyes had worse visual acuity.35,36 More light is required to bleach photopigment in diabetics compared to that of controls, consistent with failure of cones to capture light even though the patients were not diagnosed with macular edema or dragged maculas.37 This could potentially be due to failure of the blood outer retinal barrier in patients with undetected macular edema. Similarly, small groups of cones fail to guide light, near the fovea but not confined to locations over retinal vascular lesions, have been found to be consistent with failure to guide light in a microenvironment of edematous fluid that may arise from the failure of the blood outer retinal barrier.38
OCT provides several views for 3D assessment of the retina, which aids in visualizing damage to the deeper or outer retinal layers.34,39–41 Sequences of B-scans can be put into a stack or volume, and from the 3D OCT images, en face or transverse images of the individual retinal layers can be generated by using information segmented from specific retinal layers. Thus, transverse images at different depths, including a selection of the individual retinal layers, can be generated and hence assessment of deeper or outer retinal changes can be made.
In this paper, we discuss subtle changes in retinas of diabetic patients who clinically have been diagnosed as having no DR or NPDR which are visible in 3D using OCT. These findings may further demonstrate extensive retinal changes in some patients, much worse than reported by the clinical classification of DR.12,14,37 This new information could help clarify the prognosis for individual patients, which is important because of the large individual differences in development of DME and DR.
METHODS
Subjects
In this case control study, 19 diabetic subjects (16 with no DR, 3 with non-proliferative retinopathy) were matched with 19 controls with respect to age and sex. The mean age of diabetics was 56.4 ± 9.30 yr (range = 39–72 yr) (Table 1) and the controls was 55.6 ± 10.4 yr (range = 34–72 yr), with no significant difference in age, t(36) = 0.26, p = 0.79. There was a total of 24 females and 14 males, with 3 type 1 diabetic subjects with non-proliferative retinopathy (2 with mild NPDR and 1 with moderate NPDR) and 16 type 2 diabetic subjects. Type 1 and 2 diabetics refer to patients who have insulin dependent and non-insulin dependent diabetes mellitus, respectively (IDDM and NIDDM respectively).42 All subjects received a comprehensive ophthalmic exam by an experienced clinician at the Indiana University (IU) School of Optometry prior to being imaged. The fundus examination included dilation with 2.5% Phenylephrine and 1% Tropicamide, and used a combination of an indirect +90 D lens with the slit lamp and a binocular indirect ophthalmoscope, plus red free filters for subtle hemorrhages. Subjects’ diabetes was controlled with a mean HbA1c of 7.09 ± 0.95%, range = 5.7–8.5% (Table 1). Subjects were taking diabetes medications such as insulin, Avandia, Metformin, Amaryl, etc. and had controlled blood pressures. The mean self-reported diabetes duration was 10.5 ± 7.72 yr, range = 2–27 yr. The type 1 diabetic patients were similar in age to the mean and several Type 2 patients, and had the longest duration of diabetes, mean = 25.3 yr (Table 1). The HbA1c of the type 1 subjects were similar to that of 3 of the other subjects, 2 of whom were similar in age and had durations of diabetes of 10 years or more (Table 1). There were missing data for HbA1c. Other ocular diseases such as glaucoma were excluded, and all subjects had good visual acuity (ranging from 20/15 to 20/20). An example is shown in detail of a 62 yr old mild NPDR subject who had a visual complaint of blur even though visual acuity was 20/20. The cause of this blur could not be determined even though experienced clinicians at the clinic initially thought it may be due to a media opacity. Then multiple variations of en face IR imaging of the retina were performed, all of which showed this lesion. Even then, with knowing the retinal location, this lesion could not be seen on clinical exam. However, the exact axial location of the lesion could not be pinpointed with the en face images. Hence, this subject was included to probe further what repeated clinical examination could not find, even though this subject had a different type of diabetes and mild NPDR in contrast to the other diabetic subjects. The study adhered to the tenets of the Declaration of Helsinki, and informed consent from all subjects were obtained prior to experimental data collection. The study was approved by the Indiana University Institutional Review Board (IRB).
Table 1.
Demographics of diabetic subjects showing sex, type of diabetes, type of diabetic retinopathy (DR), age, diabetes duration, and HbA1c. NA: Not Available M/F: Male/Female. Type 1 and 2 diabetics refer to patients who have insulin dependent and non-insulin dependent diabetes mellitus respectively (IDDM and NIDDM respectively).42
Subjects | Sex (M/F) |
Diabetes Type |
DR Type | Age (years) | Duration (years) | HbA1c (%) |
---|---|---|---|---|---|---|
D001 | F | 2 | No DR | 39 | 8 | 5.7 |
D002 | F | 2 | No DR | 44 | 2 | 6.2 |
D003 | M | 2 | No DR | 44 | 3 | NA |
D004 | F | 2 | No DR | 48 | 15 | 7.5 |
D005 | F | 2 | No DR | 49 | 8 | 8 |
D006 | M | 1 | Mild NPDR | 51 | 25 | 8.1 |
D007 | F | 2 | No DR | 52 | 8 | NA |
D008 | M | 2 | No DR | 53 | 12 | 7.5 |
D009 | F | 1 | Moderate NPDR | 53 | 27 | 8.5 |
D010 | M | 2 | No DR | 58 | 3 | 5.8 |
D011 | F | 2 | No DR | 60 | 4 | 6.7 |
D012 | M | 2 | No DR | 60 | 5 | 6.9 |
D013 | F | 1 | Mild NPDR | 62 | 24 | 8.2 |
D014 | M | 2 | No DR | 62 | 15 | 6 |
D015 | F | 2 | No DR | 63 | 8 | 7 |
D016 | M | 2 | No DR | 65 | 10 | 8 |
D017 | F | 2 | No DR | 67 | 13 | 7.8 |
D018 | F | 2 | No DR | 70 | 6 | 5.7 |
D019 | F | 2 | No DR | 72 | 3 | 6.9 |
Procedure
Image Acquisition:
A total of 38 eyes (20 right eyes and 18 left eyes) were imaged with SD-OCT (Spectralis 1, and software upgrades through version 6.3.2.0, Heidelberg Engineering, Heidelberg, Germany). Macular centered SD-OCT grids of 20×15 deg were taken with an average of 133 b-scans across the macula, 30 µm apart, with an average of 16 frames per b-scan.
To detect and localize subtle changes within the retina, slabs of b-scans that differed in depth were created by first segmenting the b-scans between specific boundaries of retinal layers and combining all data between 2 user-selected boundaries, using vendor software. The resulting image was displayed as an en face view of that retinal layer, termed a transverse image (Figure 1). We generated transverse images of the nerve fiber layer (NFL), ganglion cell layer (GCL), inner plexiform layer (IPL), inner nuclear layer (INL), outer plexiform layer (OPL), outer nuclear layer (ONL), photoreceptor layer (PRL), and the RPE. The PRL transverse images were generated from the combination of the inner and outer segments of the photoreceptors. En face IR images were exported from the Spectralis for further analysis. Figures 1a and b show a b-scan with NFL segmentation (red boundary demarcation) and NFL slab (slab thickness of 20 µm), respectively, of a type 1 diabetic with mild NPDR. Figures 1 c and d show a b-scan with NFL segmentation (red boundary demarcation) and NFL slab (slab thickness of 20 µm), respectively, of an age-matched control to the diabetic subject. Figures 1 e and f show NFL transverse images of the same diabetic subject and control respectively.
Figure 1.
SDOCT images of a diabetic patient and age-matched control showing the computed transverse/en face images, specification of slabs, and cross-sectional images. (a) segmentation of the NFL (red boundary demarcation) and (b) 20 µm thick NFL slab of a type 1 diabetic with mild NPDR. (c) segmentation of the NFL (red boundary demarcation) and (d) 20 µm thick NFL slab of an age-matched control. (e) NFL transverse image displayed en face of the diabetic subject and (f) NFL transverse image displayed en face of an age-matched control subject.
Image Processing:
To reduce unwanted noise (Figure 2a) in the transverse images, we used a custom routine (Matlab, Mathworks, Natick, MA) first to compute the mean, maximum, and minimum intensity for each row of pixels in the b-scan. Next, to smooth out the large changes in intensity row to row, a mask was created by first subtracting the maximum in each row from the row mean, followed by inverting the values, then generating a 2D mask with these values, and finally adding the mask back to the original image (Figure 2b). This process reduced the extraneous intensity changes among the rows, as seen by comparing Figure 2a to Figure 2b, allowing smaller or more subtle features to be more distinct.
Figure 2.
Noise reduction in the transverse images. (a) NFL transverse image corresponding to a 20×15-degree field centered on the macula of a diabetic subject with mild NPDR before the intensity mask was applied. (b) the NFL transverse image after the intensity mask was applied. The arrows in (a) show noise that is absent in (b).
Image Analysis:
FAZ size and Retinal Thickness Measurements:
Transverse images of all the fundus layers of both diabetics and age-matched controls were investigated to determine which layer contained the clearest FAZ, and to compute FAZ size. The FAZ was most visible in the NFL transverse image (Figure 3a). Figures 3a-h show the relative visibility of the FAZ in transverse images of the NFL, GCL, IPL, INL, OPL, ONL, PRL, and RPE, respectively, of a control subject, with arrows showing the FAZ. Figures 4a and b also show the marked FAZ in the NFL transverse image and the magnified image of the FAZ region of the same subject shown in reverse contrast respectively.
Figure 3.
Marking the FAZ area in the transverse images. 20×15-deg field transverse images of (a) NFL, (b) GCL, (c) IPL, (d) INL, (e) OPL, (f) ONL, (g) PRL, and (h) RPE centered on the macula of a control subject. Arrows show the FAZ in each retinal layer. The FAZ (a) is most visible in the NFL transverse image.
Figure 4.
(a) FAZ marked in red in the NFL transverse image of a control subject. (b) Magnified image of the FAZ region of the same subject, shown in reverse contrast. Scale bar = 200 µm.
The size of the FAZ was compared for diabetics versus controls using two methods: vendor software (Spectralis) that also provided focal retinal thickness measurements at the FAZ margins (Figure 5) and manual marking of the FAZ (Adobe Photoshop CS6, San Jose, CA). Using the vendor software, we measured the inner retinal layer (IRLFAZ) thickness from the boundary of the inner limiting membrane to the OPL/ONL boundary, and the outer nuclear layer (ONLFAZ) thickness from the boundary of the OPL/ONL to the external limiting membrane at the margins of the FAZ (Figure 5). Nasal and temporal FAZ measurements were obtained, then averaged, for IRLFAZ and ONLFAZ for each subject. We measured the horizontal FAZ diameter (FAZH) and the vertical FAZ diameter (FAZV) (Figure 5 a). The second method for FAZ diameter was computed from the FAZ size/area values read from the vendor software, defined as the diameter of the circle whose area was equivalent to the area of the FAZ. Hence the FAZ effective diameter = (4*FAZ area/π)1/2.
Figure 5.
FAZ and retinal thickness measurements of a diabetic subject with mild NPDR. (a) NFL transverse image showing FAZ demarcation (green demarcation). The vertical FAZ diameter (FAZV) corresponds to the vertical yellow line, and the horizontal FAZ diameter (FAZH) corresponds to the horizontal yellow line as shown by the blue arrow. (b) Corresponding segmented B-scan showing the inner retinal layer (IRLFAZ) thickness at the margins of the FAZ. Measurements were taken from the inner limiting membrane (top red line) to the OPL/ONL boundary (middle pink line), as shown by the dotted blue arrows. The outer nuclear layer (ONLFAZ) thickness at the FAZ margins were taken from the OPL/ONL boundary to the external limiting membrane, as shown by the solid red arrows.
For manual FAZ marking in Photoshop, the NFL transverse images with the FAZ marked were converted into reverse contrast images that were exported as flattened tiffs into Matlab. The number of pixels covered by the FAZ area was computed, then converted into microns by multiplying them by the micron to pixel ratio in the X and Y directions from the vendor’s fiducial marks on the transverse images.
Reflectivity and Structural Changes in diabetic retinas:
The underlying retinal vessels that were clinically not visible as a result of obscuration by the presence of overlying reflective structures were investigated in the transverse images, and shown in detail for a mild NPDR subject. These overlying reflective structures included normal retinal tissues and reflectivity changes due to the leakage of fluid, lipids, proteins, etc. from retinal vessels. Outer retinal changes not clinically detected for the type 1 subjects with non-proliferative retinopathy were further localized axially, using the b-scans and the transverse images, for example, the location of a pathology found in the IR image of a 62 yr old mild NPDR subject. Two graders assessed the visibility of retinal vessels in the en face image and all slabs of the transverse images. This was done by marking small branches of the vessels around the FAZ in the en face image and transverse images of the individual retinal layers. Retinal vessels outside the FAZ are generally expected to be most visible in NFL slab, and in the en face images when not masked by reflectivity changes. However, for diabetic subjects at some retinal locations, there was poor visibility of the small branches of retinal vessels in the NFL slab, but detectible in at least one other image. The location of these retinal vessels was marked by two graders working in a masked manner. The graders were masked to the depth or type of transverse image they were viewing, and the results of the other grader.
Statistical Analysis
All statistical analysis was done using IBM SPSS Statistics for Windows, Version 23.0 (IBM Corporation, Armonk, NY, USA). All values are presented as mean ± SD. A p value < 0.05 was considered statistically significant. A paired sample t-test was used to compare the mean FAZ area of the diabetics and age-matched controls. Linear regression was used to examine associations between FAZ/IRLFAZ and HbA1c and duration of diabetes, separately due to missing data. A Bland-Altman plot was used to evaluate the 95% limits of agreement between the FAZ area measured by the vendor software and the manual marking. A linear regression of the differences between the two methods versus the average tested whether there was a proportional bias in the Bland-Altman plot.43 FAZ area was plotted to investigate age and compare diabetics and controls. The relations between the FAZ area versus FAZ effective diameter, IRLFAZ versus FAZ area, and IRLFAZ versus FAZ effective diameter were examined using Pearson’s product moment correlation.
RESULTS
Quantitative Results
The mean ± SD FAZ area was significantly larger for the diabetics (0.388 ± 0.074 mm2) vs. controls (0.243 ± 0.113 mm2), using the vendor software, and similarly for diabetics (0.388 ± 0.092 mm2) vs. controls (0.253 ± 0.120 mm2) using the manual method, t(18) = 5.27, p < 0.0001 and t(18) = 4.53, p = 0.00026, respectively, (Figure 6). This was true even when the type 1 subjects with non-proliferative retinopathy and their age-matched controls were excluded, diabetics (0.383 ± 0.079 mm2), controls (0.233 ± 0.108 mm2), t(15) = 4.75, p = 0.0003 using vendor software. The mean FAZ of the type 1 subjects with non-proliferative retinopathy (0.417 mm2) fell within the 95% confidence interval (CI) of the mean FAZ (0.341 mm2, 0.425 mm2) of the type 2 diabetics without retinopathy. There was no significant effect of either HbA1c level or duration of diabetes on FAZ area of the diabetics, F1,15 = 0.002, p = 0.96, R2 = 0.00016 and F1,17 = 1.33, p = 0.26, R2 = 0.073 respectively. There was also no effect of either HbA1c level or duration of diabetes on the FAZ area of only the type 2 diabetics without retinopathy, F1,12 = 0.15, p = 0.71, R2 = 0.012, and F1,14 = 0.84, p = 0.38, R2 = 0.056 respectively.
Figure 6.
Mean plot with standard deviation error bars showing the mean FAZ area of the diabetics versus age-matched controls from the vendor software (a) and manual method (b). The mean FAZ area of the diabetics was significantly greater than that of the controls.
There was no statistical significant difference between FAZ area measured by the custom vendor software (0.316 ± 0.119 mm2) vs. the manual method (0.320 ± 0.125 mm2), t(37) = −1.14, p = 0.26. The 95% limits of agreement ranged from −0.045 to 0.054 with a mean of the differences (Avgdiff) of 0.0047 (Figure 7). There was no proportional bias in the Bland-Altman plot, as determined from a linear regression of the differences between the two methods versus the average, F1,36 = 2.22, p = 0.15. Outliers were small deviations and in the middle of the range, i.e. no trend with size of the FAZ. The FAZs of the 3 outliers with vendor software were 0.37, 0.36, and 0.27 mm2 (0.333 ± 0.055 mm2). The corresponding FAZs with manual method were 0.28, 0.29, and 0.21 mm2 (0.26 ± 0.044 mm2) respectively. Outliers were the result of the noise reduction program which decreased the contrast of smaller vessels around the fovea in the 3 subjects and led to an underestimation of the FAZ in these subjects with the manual method. Thus, subsequent results are reported from the vendor software only.
Figure 7.
A Bland-Altman plot of the difference (diff) between the FAZ area (mm2) measured by the custom vendor software and manual marking as a function of the mean of the FAZ area from the two techniques. The 95% limits of agreement were found by mean of the differences (Avgdiff) ± 1.96*standard deviation of the differences (s).
FAZ area did not strongly depend on age (Figure 8a), F1,17 = 0.34, p = 0.57 with an R2 = 0.020 for the diabetics and F1,17 = 1.89, p = 0.19 with an R2 = 0.10 for the controls. Excluding the type 1 subjects with non-proliferative retinopathy and their age-matched controls, FAZ area did not strongly depend on age for the type 2 diabetics without retinopathy, F1,14 = 0.30, p = 0.59, R2 = 0.021, or their age-matched controls, F1,14 = 1.10, p = 0.31, R2 = 0.073. Setting up a 95% confidence interval (CI) around the mean FAZ area for the controls (0.189 µm, 0.298 µm), 16 (84.2%) diabetic subjects fell outside this CI (Figure 8a).
Figure 8.
FAZ area plotted as a function of age and IRLFAZ plotted as a function of FAZ. (a) Plot showing overlap between the FAZ area of the diabetics and controls and lack of significant association between FAZ area and age for diabetics and controls. (b) Plot showing overlap between the IRLFAZ of the diabetics and controls and lack of significant association between IRLFAZ and FAZ for diabetics and controls.
The mean IRLFAZ was significantly thicker for the diabetics (86.8 ± 14.5 µm) than controls (65.2 ± 16.3 µm), t(18) = 4.59, p = 0.00023. This was true even when the type 1 subjects with non-proliferative retinopathy and their age-matched controls were excluded, t(15) = 3.45, p = 0.004. The mean IRLFAZ of the type 1 subjects with non-proliferative retinopathy (78.7 µm) fell outside the 95% CI of the mean IRLFAZ (80.3 µm, 96.3 µm) of the type 2 diabetics without retinopathy. There was no significant effect of either HbA1c level or duration of diabetes on IRLFAZ of the diabetics, F1,15 = 2.92, p = 0.11, R2 = 0.163, and F1,17 = 2.23, p = 0.15, R2 = 0.116 respectively. Excluding the type 1 diabetics with non-proliferative retinopathy, there was no significant effect of either HbA1c level or duration of diabetes on IRLFAZ of only the type 2 subjects without retinopathy, F1,12 = 0.86, p = 0.37, R2 = 0.067, and F1,14 = 1.16, p = 0.30, R2 = 0.076 respectively. Setting up a 95% CI around the mean IRLFAZ for the controls (57.4 µm, 73.0 µm), 15 (78.9%) diabetic subjects fell outside this CI (Figure 8b). A type 1 diabetic subject and three type 2 subjects did not fall outside the 95% CI for the controls (D003, D005, D013, and D014 in Table 1). The mean ONLFAZ however did not differ between the diabetics (97.5 ± 11.0 µm) and the controls (101 ± 16.7 µm), t(18) = −0.68, p = 0.51. This was also true even when the type 1 diabetics with non-proliferative retinopathy and their age-matched controls were excluded, t(15) = −0.941, p = 0.362. The mean ONLFAZ of the type 1 subjects with non-proliferative retinopathy (97.3 µm) fell within the 95% CI of the mean ONLFAZ (91.7 µm, 103 µm) of the type 2 diabetics without retinopathy.
There was not a significant association between FAZ area and the mean IRLFAZ for the diabetics, r = 0.099, p = 0.69 (Figure 8 b), and similarly between FAZ effective diameter and the mean IRLFAZ for the diabetics, r = 0.083, p = 0.74. There was also no significant association between FAZ area and mean IRLFAZ, r = 0.17, p = 0.53 for the diabetics when the type 1 subjects with non-proliferative retinopathy and their age-matched controls were excluded. The same was true for the association between FAZ effective diameter and the mean IRLFAZ, r = 0.18, p = 0.50.
Qualitative Results
FAZ area was clearly marked in both diabetic patients (Figures 9a,c,e) and controls (Figures 9b,d,f). The marked FAZ area varied in shape and size among individuals (Figure 9).
Figure 9.
FAZ areas shown in reverse contrast. (a) FAZ regions of diabetics. (b) FAZ regions of age-matched controls to (a). (c) FAZ regions of diabetics. (d) FAZ regions of age-matched controls to (c). (e) FAZ regions of diabetics. (f) FAZ regions of age-matched controls to (e). The FAZ regions vary in shape and size among individuals for both diabetic patients and controls. Scale bar = 200 µm.
There were striking reflectivity changes in some of the en face IR images that indicated pathological changes, particularly in the en face IR image of a 62 yr old type 1 diabetic subject who clinically had been diagnosed as having only mild NPDR (Figure 10a). Clinically, these reflectivity changes obscured the visualization of underlying retinal structures (Figure 10a). However, extensive retinal changes consistent with more severe retinopathy than clinically diagnosed were visualized in 3D beneath the overlying hyper reflective changes (Figure 10b). For instance, there were retinal vessels that did not appear in the NFL transverse image (Figure 10b), but were seen in at least one other image. These could have been causing shadows due to being poorly perfused, although we cannot rule out that these small vessels were localized deeper than usually found for normal eyes. Comparison of the transverse areas localized the lesion to deeper fundus layers. The b scan also demonstrates the lesion (Figure 10c). Other similar outer retinal changes involving the photoreceptor layer were observed for the other type 1 diabetics with non-proliferative retinopathy (10 d-e). The lesion found in the IR image (Figure 11a) was absent in the transverse images of the inner retina (Figures 11b-e) but was present in the transverse images of the outer retina (Figures 11 f-i). Hence the lesion was due to outer retinal damage rather than inner retinal damage. Considering the layers of the outer retina, the damage was least prominent in the OPL (Figure 11 f) and most prominent in the PRL (Figure 11h) and the RPE image (Figure 11i) transverse images. Therefore, the lesion was a result of photoreceptor and RPE damage since it was most prominent in the PRL (Figure 11h) and the RPE transverse images (Figure 11i), consistent with damage to the outer blood retinal barrier. Further, in the transverse images, retinal vessels were visible that were not seen in the en face image. This may be due in part to the reflectivity changes in the en face image that mask the vessels, but could also be due to decreased vessel contrast associated with diabetic changes to the retinal vessels per se.
Figure 10.
20×15-deg en face, NFL transverse, and b scan images centered on the macula of a 62 yr old type 1 diabetic with mild NPDR and b scans of other type 1 diabetics with non-proliferative retinopathy. (a) Blue circle shows areas of reflectivity changes in the IR image. There is a striking reflectivity change indicated by the blue arrow, and focal regions of lower reflectivity. These reflectivity changes obscure the visualization of underlying retinal structures. (b) Arrows show vessels that appeared to be non-perfused in the corresponding NFL transverse image, but were detected in one or more of the other layers or the en face image. Small, hyperreflective dots are consistent with early vascular changes. (c) B scan with blue arrow showing the axial location of the lesion in (a) to be a damage to the PRL and RPE. (d-e) B scans of 51 and 53 yr old type 1 diabetics with mild and moderate non-proliferative retinopathy, respectively, showing outer retinal changes (blue arrows) similar to that seen in panel c.
Figure 11.
Comparison of a 20×15-deg en face image and successively deeper transverse images centered on the macula, showing that the origin of the greatest reflectivity change (red box) is in the deeper layers for a 62 yr old type 1 diabetic with mild NPDR. (a) IR image, (b) NFL, (c) GCL, (d) IPL, (e) INL, (f) OPL, (g) ONL, (h) PRL, and (i) RPE transverse images of a type 1 diabetic with mild NPDR, from the same data set and retinal region. The lesion in (a) is absent in the inner retinal layers in panels b-e, with no box shown, but present in the outer retinal layers (red box) f-i. The boxes in panels f-i correspond to that in panel a.
DISCUSSION
Diabetic patients currently may not be treated for the earliest clinical signs of DR, yet transmission of eye exam results of these patients to a primary care physician or diabetologist for the control of systemic glucose level provides a more complete assessment of the patient’s status.3 An updated clinical classification of DR is emerging, since it is well-established that there can be retinal changes more severe than seen on clinical exam or fundus photography.12,14,37 For example, disorganization of the inner retinal layers can occur in patients who may show resolved macular edema on clinical exams.11 Vessel remodeling occurs in patients with a clinical classification of only mild or moderate non-proliferative retinopathy.12 Thus, the early referral of diabetic patients is crucial when treatment or systemic glucose control can still prevent vision loss,3 by improving the present clinical classifications of DR.12,14,37
The association between FAZ enlargement and DR severity has well been documented using FA.15–21 The idea is that diabetic patients with more severe DR have a larger FAZ size with increase in FAZ size observed at all stages of DR even in diabetic subjects without clinical retinopathy as compared to age- matched controls.15–21 Enlargement of the FAZ size in diabetic subjects with no DR compared to controls has also been found using other imaging modalities such as AOSLO,14 and OCTA.22–24 By visualizing the FAZ in 3D using OCT, our results also showed enlarged FAZ size in diabetic patients, despite that 16 had no DR and 3 had non-proliferative retinopathy as compared to their age-and gender- matched controls (Figure 7). The enlarged FAZ in the diabetics was also true even when the diabetic patients with non-proliferative retinopathy and their age-matched controls were excluded (p = 0.0003). Thus, our results from the 3D en face/transverse OCT agree with previous studies that posit retinal vascular remodeling at the FAZ in diabetics before the clinical detection of DR.14–24
The mean FAZ area of our normal subjects (0.243 ± 0.113 mm2) using the 3D en face/transverse OCT was comparable to that measured with OCTA in two different studies; (0.25 ± 0.06 mm2)23 and (0.266 ± 0.097 mm2)44 as well as with FA; (0.231 ± 0.06 mm2),18 and (0.221 ± 0.09 mm2).16 There was also agreement for the mean FAZ area of our diabetics without retinopathy (0.383 ± 0.079 mm2) and the FAZ area of diabetics without DR measured with OCTA; (0.37 ± 0.07 mm2).23 These findings corroborate the idea that the clinical classification of DR is insufficient and that there are more extensive retinal changes in diabetic patients than detected on clinical exam or fundus photography.12,14,37 As OCTA can be challenging in patients with poor fixation,22 the present method may provide similar information.
Tissue oxygenation of the inner retinal layers is provided by the inner retinal capillaries. The relatively constant inner retinal layer thickness at the margins of the FAZ has been attributed to the limits of the metabolic support by these capillaries.25 This suggests that when there is an increase in the size of the FAZ, there is reduced support from the inner retinal capillaries to these inner retinal neurons and hence the possibility of a biological penalty to these layers. The retinal neurovascular unit consists of retinal capillaries, Muller cells, astrocytes and neurons that are closely interconnected to ensure the supply of oxygen and nutrients to the retinal neurons.5 This is evident in the upregulation of retinal blood flow in response to the increase lactate level and partial oxygen pressure in the retinal neurons.45 Hence in the case of reduced oxygen supply to the neurons as can occur in a large FAZ, the buildup of lactate in the retinal neurons may cause the neurons to swell or become edematous. This may explain the thicker IRLFAZ seen in the diabetics as compared to their age-matched controls positing that the inner retinal neurons were edematous rather than shrinking. In this study, the IRLFAZ was thicker in the diabetics than controls even though there was lack of exudation on clinical exams not generally detected by changes in the OCT b-scans. There was a lack of significant association between the IRLFAZ and FAZ area and hence lack of measurable thinning of the inner retinal layers to reduce the neural mass at the large FAZ. Further research is needed to quantify neural cell counts in addition to counts of the potentially more robust cones, which are likely to receive metabolic support via the choriocapillaris and RPE.
We also found a lack of significant association between FAZ size and age in both diabetics and controls (Figure 8), similar to findings using OCTA for healthy subjects.44 This suggests that for our sample, we cannot support or disprove that diabetes mellitus is just like aging with respect to the changes in FAZ integrity. These associations may be different for a larger sample size in a different study.
We also found retinal changes consistent with more severe retinopathy than clinically detected, e.g. for a 62 yr old type 1 diabetic subject with only mild NPDR (Figure 10). Overlying reflective changes in diabetic retinas as a result of leakage of fluid, lipids, proteins, etc. from retinal vessels can obfuscate the visualization of underlying retinal structures. By imaging the retina in 3D using OCT, en face or transverse images of different retinal layers at different depths can be generated. This makes it possible to visualize the underlying retinal structures in the presence of overlying reflective changes, even though these changes may decrease the contrast of underlying retinal structures on clinical exam or fundus photography. The patient was diagnosed as having only mild NPDR even though there were extensive underlying retinal changes. We found a surprising lack of visible retinal vessels in the NFL transverse image (Figure 10 b), particularly for a 62 yr old type 1 diabetic subject with mild NPDR. These underlying retinal changes were however not detected on clinical exam nor the en face image of the retina as a result of the overlying reflective retinal structures.
DR is typically classified as an inner retinal disease based on changes to the superficial retinal vasculature and neurons.6,7 However, there is growing evidence that outer retinal changes are equally important and have a significant impact on visual acuity.35–37 Early outer retinal changes have been found in diabetic subjects without DME or dragged maculas, as shown in the greater amount of light required to bleach cone photopigment in these subjects potentially due to failure of the outer retinal blood barrier.37 In diabetic eyes, dark cones near the fovea but not retinal vascular lesions have been found to be consistent with the failure to guide light in a microenvironment of edematous fluid.38 The integrity of the photoreceptor layer and its effect on visual acuity has also been found,35,36 positing the importance of outer retinal changes in diabetic subjects. The axial location of a lesion found in the IR image of a 62 yr old type 1 diabetic subject with mild NPDR was found to be the photoreceptor and RPE layers (Figure 11). Although this lesion could be seen in the en face image, one could not conclude that the lesion was damage to the photoreceptor and RPE layers. Hence the presence of photoreceptor and RPE damage in this diabetic subject was clinically missed altogether. Similar outer retinal changes involving the photoreceptor layer were observed for the other type 1 diabetics with non-proliferative retinopathy (Figures 10 d-e). This shows that the outer retinal change observed in the 62 yr old type 1 diabetic was not limited to that subject only, but was observed in other type 1 subjects as well. Also, similar outer retinal changes have been observed for type 2 diabetics with dark cones,38 which is a sign found in both type 1 and type 2 diabetes that is consistent with early failure of the function of outer blood retinal barrier. In addition to OCT, hyperreflective foci and cystoid structures in the deeper retinal layers have been better visualized in red channel than green channel images in diabetic subjects with DME.27 Since diabetes does not only cause microvascular changes to the inner retina but also leads to outer retinal changes even in the early stages of no DR or DME,37 it is imperative that detection of early changes to the outer retinal layers in diabetic subjects are emphasized.
The results of this study should be interpreted in light of the following limitations. The disagreement among some of the FAZ data could be attributed to noise in the transverse images produced by the vendor software. Our noise reduction algorithm reduced the contrast of small vessels around the FAZ leading to the three outliers with the manual method. However, both methods measured similar FAZ sizes as can be seen with the Bland-Altman plot (Figure 7). There was no trend in the plot that compared the two methods because the outliers were in the middle of the range, i.e. no trend with size of the FAZ. There was also no proportional bias when a linear regression of the differences between the two methods versus the average was done. The range of HbA1c in our subjects may not have been sufficiently large to uncover an effect on FAZ, and our study was not longitudinal. Also, the duration of diabetes and HbA1c are self-reported information from the subjects which may not always be accurate. There were also missing data on HbA1c for some of the diabetic subjects (Table 1). There was a small number of type 1 participants with retinopathy compared with the much larger number of type 2 participants without retinopathy.
In conclusion, the presence of photoreceptor and RPE damage, thicker inner retinal neurons at the FAZ margins, larger FAZ area, and underlying vessels seen in NFL transverse image but not seen in en face or seen in other transverse images but not seen in NFL transverse in these diabetic subjects show that there are extensive retinal changes, greater than detected on standard color fundus photography. In our subjects, the IRLFAZ had not thinned to decrease the neural mass at the margins of the large FAZ indicating that the inner retinal neurons at the FAZ margins were edematous rather than only shrinking. These results are consistent with the idea that the clinical classification of DR is insufficient and that there are extensive retinal changes in diabetic patients, with individual differences early on.12,14,37 Based on our results, these changes occur both in the inner and outer retina. The need for a reclassification of the different stages of DR with improved imaging modalities that have better resolution and contrast than color fundus cameras may be important in ocular and systemic management of diabetic patients.
Acknowledgments
Acknowledgements: This work was supported by NIH (EY007624 and EY026105) to AEE and (EY024315) to Dr. Stephen Allan Burns (SAB), Indiana University School of Optometry, 800 E. Atwater Avenue, Bloomington, IN 47405.
Footnotes
Disclosure: The authors report no conflicts of interest and have no proprietary interest in any of the materials mentioned in this article.
REFERENCES
- 1.Blumenkranz MS. Optimal current and future treatments for diabetic macular edema. Eye 2010; 24: 428–434. [DOI] [PubMed] [Google Scholar]
- 2.Danaei G, Finucane MM, Lin JK et al. National, regional, and global trends in systolic blood pressure since 1980: systematic analysis of health examination surveys and epidemiological studies with 786 country-years and 5.4 million participants. Lancet 2011; 377: 568–577. [DOI] [PubMed] [Google Scholar]
- 3.Elsner AE & King BJ. Screening for macular disorders: the optometrist’s perspective. Clinical Optometry 2015; 7: 15–38. [Google Scholar]
- 4.Centers for Disease Control and Prevention. Improving the Nation’s Vision Health: A Coordinated Public Health Approach. Atlanta: CDC; 2006. Available from: http://www.cdc.gov/visionhealth/pdf/improving_nations_vision_health_508_final.pdf. Accessed September 11, 2016. [Google Scholar]
- 5.Antonetti DA, Klein R & Gardner TW. Diabetic retinopathy. N Engl J Med 2012; 29: 1227–1239. [DOI] [PubMed] [Google Scholar]
- 6.Grading diabetic retinopathy from stereoscopic color fundus photographs – an extension of the modified Airlie House classification. ETDRS report number 10. Early Treatment Diabetic Retinopathy Study Research Group. Ophthalmology 1991; 98: 786–806. [PubMed] [Google Scholar]
- 7.Wilkinson CP, Ferris FL 3rd, Klein RE et al. Global Diabetic Retinopathy Project Group. Proposed international clinical diabetic retinopathy and diabetic macular edema disease severity scales. Ophthalmology 2003; 110: 1677–1682. [DOI] [PubMed] [Google Scholar]
- 8.Arend O, Remky A, Elsner AE, Bertram B, Reim M & Wolf S. Quantification of cystoid changes in diabetic maculopathy. Invest Ophthalmol Vis Sci 1995; 36: 608–613. [PubMed] [Google Scholar]
- 9.Beausencourt E, Remky A, Elsner AE, et al. Infrared scanning laser tomography of macular cysts. Ophthalmology 2000; 107: 375–385. [DOI] [PubMed] [Google Scholar]
- 10.Antonetti DA, Barber AJ, Bronson SK et al. Diabetic retinopathy: seeing beyond glucose-induced microvascular disease. Diabetes 2006; 55: 2401–2411. [DOI] [PubMed] [Google Scholar]
- 11.Sun JK, Radwan SH, Soliman AZ et al. Neural retinal disorganization as a robust marker in current and resolved diabetic macular edema. Diabetes 2015; 64: 2560–2570. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Burns SA, Elsner AE, Chui TY et al. In vivo adaptive optics microvascular imaging in diabetic patients without clinically severe diabetic retinopathy. Biomed Opt Express 2014; 5: 961–974. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Browning DJ, McOwen MD, Bowen RM Jr & O’Marah TL. Comparison of the clinical diagnosis of diabetic macular edema with diagnosis by optical coherence tomography. Ophthalmology 2004; 111: 712–715. [DOI] [PubMed] [Google Scholar]
- 14.Tam J, Dhamdhere KP, Tiruveedhula P et al. Disruption of the Retinal Parafoveal Capillary Network in Type 2 Diabetes before the Onset of Diabetic Retinopathy. Invest. Ophthalmol. Vis. Sci 2011; 52: 9257–9266. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Arend O, Remky A, Evans D, Stüber R & Harris A. Contrast sensitivity loss is coupled with capillary dropout in patients with diabetes. Invest. Ophthalmol. Vis 1997; 38: 1819–1824. [PubMed] [Google Scholar]
- 16.Arend O, Remky A, Harris A, Bertram B, Reim M & Wolf S. Macular microcirculation in cystoid maculopathy of diabetic patients. Br J Ophthalmol 1995; 79: 628–632. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Arend O, Wolf S, Harris A & Reim M. The relationship of macular microcirculation to visual acuity in diabetic patients. Arch Ophthalmol 1995; 113: 610–614. [DOI] [PubMed] [Google Scholar]
- 18.Arend O, Wolf S, Jung F et al. Retinal microcirculation in patients with diabetes mellitus: dynamic and morphological analysis of perifoveal capillary network. Br J Ophthalmol 1991; 75: 514–518. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Arend O, Wolf S, Remky A et al. Perifoveal microcirculation with non-insulin-dependent diabetes mellitus. Graefe’s Arch Clin Exp Ophthalmol 1994; 232: 225–231. [DOI] [PubMed] [Google Scholar]
- 20.Bresnick GH, Condit R, Syrjala S, Palta M, Groo A & Korth K. Abnormalities of the foveal avascular zone in diabetic retinopathy. Arch Ophthalmol 1984; 102: 1286–1293. [DOI] [PubMed] [Google Scholar]
- 21.Leite E, Mota MC, Faria DA et al. Quantification of the foveolar zone in normal and diabetic patients. J Fr Ophtalmol 1988; 12: 665–668. [PubMed] [Google Scholar]
- 22.de Carlo TE, Romano A, Waheed NK & Duker JS. A review of Optical Coherence Tomography Angiography (OCTA). Int J Retina Vitreous 2015; 1: 5. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Takase N, Nozaki M, Kato A, Ozeki H, Yoshida M & Ogura Y. Enlargement of foveal avascular zone in diabetic eyes evaluated by en face optical coherence tomography angiography. Retina 2015; 35: 2377–2383. [DOI] [PubMed] [Google Scholar]
- 24.Di G, Weihong Y, Xiao Z et al. A morphological study of the foveal avascular zone in patients with diabetes mellitus using optical coherence tomography angiography. Graefe’s Arch Clin Exp Ophthalmol 2016; 254: 873–879. [DOI] [PubMed] [Google Scholar]
- 25.Chui TYP, VanNasdale DA, Elsner AE & Burns SA. The association between the foveal avascular zone and retinal thickness. Invest Ophthalmol Vis Sci 2014; 55: 6870–6877. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Bhagat N, Grigorian RA, Tutela A & Zarbin MA. Diabetic macular edema: pathogenesis and treatment. Surv Ophthalmol 2009; 54:1–32. [DOI] [PubMed] [Google Scholar]
- 27.Alhamami MA, Elsner AE, Malinovsky VE et al. Comparison of Cysts in Red and Green Images for Diabetic Macular Edema. Optom Vis Sci 2017; 94: 137–149. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Vinores SA, Campochiaro PA, Lee A, McGehee R, Gadegbeku C & Green WR. Localization of blood-retinal barrier breakdown in human pathologic specimens by immunohistochemical staining for albumin. Lab Invest 1990; 62:742–750. [PubMed] [Google Scholar]
- 29.Vinores SA, McGehee R, Lee A, Gadegbeku C & Campochiaro PA. Ultrastructural localization of blood-retinal barrier breakdown in diabetic and galactosemic rats. J Histochem Cytochem 1990; 38: 1341–1352. [DOI] [PubMed] [Google Scholar]
- 30.Xu HZ & Le YZ. Significance of outer blood-retina barrier breakdown in diabetes and ischemia. Invest Ophthalmol Vis Sci 2011; 52: 2160–2164. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31.Dahrouj M, Alsarraf O, Liu Y, Crosson CE & Ablonczy Z. C-type natriuretic peptide protects the retinal pigment epithelium against advanced glycation end product-induced barrier dysfunction. J Pharmacol Exp Ther 2013; 344: 96–102. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32.Dahrouj M, Alsarraf O, McMillin JC, Liu Y, Crosson CE & Ablonczy Z. Vascular endothelial growth factor modulates the function of the retinal pigment epithelium in vivo. Invest Ophthalmol Vis Sci 2014; 55: 2269–2275. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33.Garcia-Ramírez M, Hernández C, Palomer X, Vázquez-Carrera M & Simó R. Fenofibrate prevents the disruption of the outer blood retinal barrier through downregulation of NF-κB activity. Acta Diabetol 2015; 53: 109–118. [DOI] [PubMed] [Google Scholar]
- 34.Abozaid MA, Scoles D, Goldberg M, Carroll J & Han DP. En Face Optical Coherence Tomography of Outer Retinal Discontinuity and Fan-Shaped Serous Macular Detachment in Diabetic Macular Edema. JAMA Ophthalmol 2015; 133: 961–963. [DOI] [PubMed] [Google Scholar]
- 35.Otani T, Yamaguchi Y & Kishi S. Correlation between visual acuity and foveal microstructural changes in diabetic macular edema. Retina 2010; 30: 774–80. [DOI] [PubMed] [Google Scholar]
- 36.Shen Y, Liu K & Xu X. Correlation between visual function and photoreceptor integrity in diabetic macular edema: spectral-domain optical coherence tomography. Curr Eye Res 2016; 41: 391–399. [DOI] [PubMed] [Google Scholar]
- 37.Elsner AE, Burns SA, Lobes LA Jr & Doft BH. Cone photopigment bleaching abnormalities in diabetes. Invest Ophthalmol Vis Sci 1987; 28: 718–724. [PubMed] [Google Scholar]
- 38.Sawides L, Sapoznik KA, de Castro A et al. Alterations to the Foveal Cone Mosaic of Diabetic Patients. Invest Ophthalmol Vis Sci 2017; 58: 3395–3403. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 39.Wolff B, Basdekidou C, Vasseur V, Sahel JA, Gaudric A, Mauget-Faÿsse M. “En face” optical coherence tomography imaging in type 2 idiopathic macular telangiectasia. Retina 2014; 34: 2072–2078. [DOI] [PubMed] [Google Scholar]
- 40.Schaal KB, Legarreta AD, Gregori G et al. Widefield en face optical coherence tomography imaging of subretinal drusenoid deposits. Ophthalmic Surg Lasers Imaging Retina. 2015; 46: 550–559. [DOI] [PubMed] [Google Scholar]
- 41.Cheng S, Leng T. Noninvasive detection of microaneurysms in diabetic retinopathy by swept-source optical coherence tomography. Clin Ophthalmol 2016; 10: 1791–1795. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 42.Kulkarni M, Leszczynska A, Wei G et al. Genome-wide analysis suggests a differential microRNA signature associated with normal and diabetic human corneal limbus. Sci Rep 2017; 7:3448. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 43.Giavarina D Understanding Bland Altman Analysis. Biochem Med. 2015; 25: 141–151. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 44.Samara WA, Say EA, Khoo CT et al. Correlation of foveal avascular zone size with foveal morphology in normal eyes using optical coherence tomography angiography. Retina 2015; 35: 2188–2195. [DOI] [PubMed] [Google Scholar]
- 45.Pournaras CJ, Rungger-Brandle E, Riva CE, Hardarson SH & Stefansson E. Regulation of retinal blood flow in health and disease. Prog Retin Eye Res 2008; 27: 284–330. [DOI] [PubMed] [Google Scholar]