Novel Noninvasive Detection of the Fovea Avascular Zone Using Confocal Red-Free Imaging in Diabetic Retinopathy and Retinal Vein Occlusion

Abstract

Purpose.: To report confocal red-free imaging as a novel, noninvasive imaging modality for the detection of the foveal avascular zone (FAZ) and to compare its effectiveness to that of fluorescein angiography (FA) in diabetic retinopathy (DR) and retinal vein occlusion (RVO).

Methods.: This was a retrospective, observational, cross-sectional study. The authors enrolled 50 eyes with DR and 20 eyes with RVO. All subjects underwent confocal red-free imaging and FA using a scanning laser ophthalmoscope. For all subjects, the FAZ was identified on both confocal red-free and corresponding FA images and was delineated by two independent readers. The authors evaluated the concordance between the two imaging methods by comparing the sizes of the delineated areas and determining the overlapping ratio after image processing.

Results.: The image analysis showed a high correlation (r > 0.9) in the mean size of the FAZ between the red-free and corresponding FA images with DR or RVO. Reliable agreement between the two methods was confirmed by the comparison of size (P = 0.925 on independent t-test) and overlapping correspondence (overlapping ratio, 0.77) of the delineated area.

Conclusions.: The findings suggest that confocal red-free imaging is a simple, reliable, safe, and noninvasive method for effectively imaging the FAZ. This procedure has the potential to be used for the noninvasive detection and quantification of FAZ in screening, initial evaluation, and follow-up observation of progressive ischemic retinopathies such as DR and RVO.

Diabetic retinopathy (DR) and retinal vascular occlusion (RVO) can lead to complications, including macular edema, macular ischemia, optic neuropathy, and vitreous hemorrhage due to retinal circulation disturbance. ^1,2 Among these complications, ischemic maculopathy is especially closely related to a change in the foveal avascular zone (FAZ) induced by the loss of perifoveal capillaries arising from hemodynamic stress. ^3,4 The FAZ is a clinically important zone on the retina without capillaries in the macula, which has a mean diameter of around 400 to 500 μm and an area of 0.2–0.4 mm² in normal subjects, although there are individual differences. ^3,5,6 Several studies have reported that enlargement of the FAZ is correlated with disease progression and visual disturbance in DR, ^{3,6 –8} and an Early Treatment of Diabetic Retinopathy Study (ETDRS) report suggested a qualitative classification of FAZ morphology based on the enlargement of the FAZ detected by fluorescein angiography (FA). ⁹ Abnormalities in the FAZ, including irregular FAZ margins, capillary dropout, and intercapillary space widening, can be seen on an FA examination in ischemic maculopathy. ^{3,7,10 –12} Macular ischemia is one of the main causes of vision loss in the macular RVO type that can change the morphology of the FAZ and, therefore, an understanding of FAZ status is also vitally important when predicting visual prognosis. ¹³  

At present, the gold standard method for detecting the FAZ is FA. The fovea shows a characteristic dark appearance on a normal angiogram because of columnar pigment epithelium, a greater concentration of xanthophyll and melanin than that in the remainder of the fundus, and the absence of retinal vessels in the fovea ⁵ ; however, the FA-based method has some disadvantages, including invasiveness due to the injection of contrast media into subjects' vessels, adverse effects of fluorescein dye ranging from mild skin eruption to rare but life-threatening shock, and the procedure is time consuming because it requires 7 to 8 minutes until the end of the late phase, thereby prohibiting its repeated use. ¹⁴ Therefore, a simple, noninvasive method has been highly sought after to visualize the FAZ in the field of ophthalmology. Most previous reports have made an effort to detect the FAZ on FA images ^{9,15 –19} and there have recently been studies that have noninvasively visualized the FAZ and parafoveal capillaries with techniques using the entopic phenomenon or adaptive optics imaging, although those methods remain experimental. ^{20 –24}  

Until now, conventional red-free imaging captured with the green filter of a fundus camera has been used to highlight retinal nerve fiber layers for diagnosis of glaucoma or to improve retinal vessel contrast. However, this technique is limited in visualizing other retinal pathologies because it produces low-contrast and image resolution. Red-free imaging using a green light on a fundus camera produces a lower contrast image than that with a blue light because shorter wavelengths yield higher resolution, although imaging with blue light is also affected by media opacities. ²⁵ With the recent introduction of a commercially available confocal scanning laser ophthalmoscope (SLO), the quality and contrast of images is improving, which increases the likelihood of finding a new approach to diagnose retinal pathologies. ²⁶ Therefore, in this study, confocal red-free imaging using blue reflectance was performed to visualize the FAZ. The purpose of this study was to report on confocal red-free imaging with a blue wavelength of light as a novel, noninvasive imaging modality for FAZ detection and to compare the effectiveness of this confocal red-free imaging with FA in DR and RVO. 

Methods

All participants were enrolled after a review of the medical records of patients with DR and RVO at Hanyang University Medical Center. Data were collected from March to May 2011. The study protocol was approved by the Institutional Review Board of Hanyang University Medical Center and the research followed the tenets of the Declaration of Helsinki. 

Subjects and Imaging Studies

We screened a total of 95 eyes of 80 patients with DR (65 eyes) or RVO (30 eyes) for this study. All subjects involved in this study underwent an FA examination followed by confocal red-free imaging by SLO (F-10; Nidek, Gamagori, Japan) with a blue laser source on the same day, as well as comprehensive ophthalmological exams including fundus photography. Spectral-domain optical coherence tomography (SD-OCT; 3D OCT-2000, Topcon, Tokyo, Japan) was also performed for the detection of macular involvement in DR and RVO. Lens status was simply classified into four groups (clear or pseudophakic lens, mild, moderate, or severe cataract) by one specialist (BRL) with reference to the lens opacities classification system III. ²⁷ The value of the “image quality factor ” (provided by Topcon SD-OCT), which was similar to signal strength using time-domain OCT, was measured to identify the influence of media opacity indirectly. Before performing FA, we performed confocal red-free imaging with blue reflectance of the macular area. A macular image was subtended at an angle of 40° using both confocal red-free imaging and FA. FA was performed in the standard manner (Fig. 1) and all images were obtained by a single, well-trained technician who followed the same imaging protocol for all subjects. The resolution of all retained images was 1600 × 1200 pixels. 

Figure 1.

 

Example images of a normal FAZ. (A) Raw confocal red-free image. (B) Raw FA image in midvenous phase. Bottom images show the FAZ on confocal red-free imaging (C) and FA (D) delineated by one independent reader.

Image Selection and Comparison

Image Selection.

Only the image of the macula taken during the arteriovenous phase of the FA examination was selected. All images were interpreted by a single retinal specialist (BRL) who determined whether they could be included in this study. There was no artificial manipulation (such as contrast enhancement) for better detection of the FAZ and only raw images were used. The following criteria were used to select images for this study: (1) The enrolled eye should have only a single retinal disease (either DR or RVO) based on the FA examination; (2) the FA image should clearly show the FAZ. Cases in which the quality of the red-free image was too poor to detect the FAZ due to media opacity, including severe cataract or vitreous hemorrhage, were excluded; and (3) eyes with high myopic and hyperopic refractive errors of greater than −6.0 or +6.0 diopters were excluded from this study. 

Image Processing.

Two-staged image processing was performed for quantitative analysis of FAZ. First, two masked readers (YUS, JWS) manually outlined the area of the FAZ in each image obtained by both confocal red-free imaging and FA. Each reader delineated the FAZ on the confocal red-free image twice for the measurement of test–retest reliability. The criteria in outlining the FAZ with confocal red-free imaging were as follows: (1) The reader should recognize the central darkest area of the macula; (2) the FAZ boundary was determined by plotting the border between the central dark area and the surrounding area showing an abrupt decrease in shade; and (3) an area showing a mild decrease in shade was ignored as the border, which depended on the reader's judgment. Then, with the help of commercially available software (MatLab 2010a; TheMathWorks Inc., Natick, MA), the red-free image was overlaid on the FA image to generate one registered image (Fig. 2). The affine registration method was applied for the registration procedures between the FA and red-free images. For the procedures, five correspondence points were selected in each image with respect to obvious blood vessel branches. 

Figure 2.

 

Image processing for quantitative comparisons between red-free and FA images. The top left and bottom left images were captured using red-free imaging and FA, respectively. The first step in image processing was delineation of the FAZ shown in the confocal red-free and FA images by two independent readers. After that, two images were registered (MatLab software). Using these registered images, size of the FAZ determined by each imaging technique and the size of their union and intersection were obtained for calculating the overlapping ratio.

Image Analysis for Comparison.

The registered images were used to measure the size of the FAZ in pixels (MatLab software). The first task was to determine the size of the delineated FAZ on both the red-free and FA images, and to compare them. The actual size of the FAZ was calculated by adjusting eyeball optics based on the Gullstrand eye model provided by built-in software (F-10; axial lengths of the enrolled eyes were not measured) and was expressed in square millimeters. The second task was to calculate the overlapping ratio, or “Jaccard Index,” to evaluate the concordance between the two imaging techniques; this method was used for comparing manually delineated images in another report. ¹⁸ The overlapping ratio is defined as the ratio of the shared area (intersection) to the combined area of the FAZ (union) observed by each imaging method using a registered image. The closer the value of this ratio is to 1, the more the two images overlap. To determine the overlapping ratio, the same independent readers delineated the areas of union and intersection of the FAZ, respectively, with the registered images generated during the second stage of image processing. These areas were measured and the overlapping ratio was calculated (MatLab software). All values measured by the two readers were averaged for image analysis. Interobserver repeatability was examined by calculating the interclass correlation (ICC). Independent t-tests and correlation analyses were undertaken to compare the size of the FAZ between the two methods. A test–retest reliability coefficient was calculated using correlation analysis. For all tests, a value of P < 0.05 was considered statistically significant. All statistical analyses were calculated using commercial software (SPSS version 18.0.0; SPSS Inc., Chicago, IL). 

Results

Of all the reviewed records, we included 38 subjects (50 eyes) with DR (regardless of stage) and 20 subjects (20 eyes) with RVO in this study after the image selection process. The images of 25 eyes were excluded because of image noise due to severe cataract (16 eyes), vitreous hemorrhage or floater (5 eyes), and poor pupil dilation (4 eyes). The baseline characteristics of the enrolled eyes are shown in Table 1. The image quality factor of SD-OCT was 75.38 ± 7.52 in clear lens or pseudophakic eyes, 69.53 ± 15.58 in eyes with mild cataract, and 53.31 ± 17.28 in eyes with moderate cataract. 

Table 1.

 

Clinical Characteristics of Enrolled Eyes

Characteristic	Value
Eyes, n	70
Age, y	58.62 ± 11.13
Ratio (M:F)	38:32
Refraction (SE), diopters	0.39 ± 1.87
Lens status (eyes), n
No cataract or pseudophakia	18
Mild cataract	40
Moderate cataract	12
Diagnosis (eyes), n
DR	50
NPDR	37
PDR	13
RVO	20
BRVO	17
CRVO	3

 SE, spherical equivalent; BRVO, branch retinal vein occlusion; CRVO, central retinal vein occlusion.

The mean sizes of all variables (FAZ on FA and red-free images, their unions and intersections) delineated by the two experts are shown in Table 2. There were no significant differences between values measured via manual delineation by the two experts (r > 0.8 for all variables, intraclass coefficient). The test–retest reliability coefficients for each reader in outlining the FAZ seen on confocal red-free imaging were 0.83 and 0.81, respectively (Pearson's correlation coefficient, P < 0.001). The mean FAZ sizes of the subjects with DR and RVO were 0.38 ± 0.11, 0.39 ± 0.09 mm² on FA imaging and 0.39 ± 0.12, 0.38 ± 0.08 mm² on confocal red-free imaging, respectively. The correlation coefficient between the FAZ sizes detected by the two methods was 0.809, which was statistically significant (Pearson's correlation coefficient, P < 0.001). 

Table 2.

 

Mean Size and Overlapping Ratio of Fovea Avascular Zone

Factor	Size of Delineated Area			Overlapping Correspondence
	FA (mm2)	RF (mm2)	P Value*	FA ∩ RF (mm2)	FA ∪ RF (mm2)	Overlapping Ratio
DR	0.38 ± 0.11	0.39 ± 0.12	0.645	0.33 ± 0.11	0.45 ± 0.14	0.77 ± 0.09
RVO	0.39 ± 0.09	0.38 ± 0.08	0.720	0.33 ± 0.09	0.43 ± 0.08	0.78 ± 0.08
Total	0.38 ± 0.10	0.38 ± 0.10	0.783	0.33 ± 0.10	0.44 ± 0.13	0.77 ± 0.09

 FA ∩ RF = the intersection area of FAZ shown in FA and red-free imaging. FA ∪ RF = the union area of FAZ shown in FA and red-free imaging. All values were measured by two independent readers and were averaged for image analysis. The overlapping ratio is defined as the ratio of the intersection of two methods and their union.

*  The difference in the size of the delineated area assessed by independent t-test.

Figure 3 illustrates the agreement between the two methods in a Bland–Altman plot. Overall, the overlapping ratio of the two methods was 0.77 ± 0.09. There were no statistically significant differences according to diagnosis in our comparison of the FAZ size and the overlapping ratio between the two methods. Figure 4 shows the shapes and sizes of the FAZs visualized on confocal red-free imaging, which correspond to those seen on FA. Low overlapping ratios (<0.6) were recorded in five cases because the shape of the FAZ obtained via confocal red-free imaging, which resembled a petaloid pattern, was different from that seen on FA. All these cases had severe cystoid macular edema on SD-OCT and the mean central macular thickness was 413.4 ± 55.98 μm, which was statistically larger than that of the cases with a high overlapping ratio (298.32 ± 29.82 μm) (Fig. 5). 

Figure 3.

 

Bland–Altman plot illustrating the agreement between confocal red-free imaging and FA in terms of the size of the FAZ. RF, confocal red-free imaging.

Figure 4.

 

Case series of FAZ visualization using red-free imaging and FA. Case 1 demonstrated enlargement of the FAZ on FA, which was comparable to the FAZ observed on red-free imaging. In cases 2 and 3, the FA highlighted the eccentric shape of the FAZ and the red-free images showed a similar morphology. NPDR, nonproliferative diabetic retinopathy; PDR, proliferative diabetic retinopathy; BRVO, branch retinal vein occlusion; OR, overlapping ratio.

Figure 5.

 

Images from a 67-year-old patient with severe nonproliferative NPDR. In contrast to the cases shown in Figure 4, the red-free image (A) of the fovea avascular zone in this case had a petaloid appearance that is not comparable to the FAZ on the FA image (B). The low overlapping ratio (0.471) was calculated between the delineated FAZs of the two imaging methods. Severe cystoid macular edema was found on the SD-OCT (C).

Discussion

Ischemic maculopathy is often overlooked as a reason for visual disturbance because it can be diagnosed on the FA examination, which is the only practical imaging method available to show a foveal pericapillary network. As described earlier, it is not appropriate for subjects to undergo FA frequently to determine FAZ status due to its invasiveness and side effects. There have been many studies in which the FAZ was visualized using different techniques. Bresnick et al., ³ Mansour et al., ¹⁰ and Conrath et al., ¹⁹ used conventional FA, whereas Arend et al. ⁴ used the SLO system that can yield better definition of the FAZ than a conventional fundus camera. On the other hand, there have been continuous efforts to visualize the FAZ without classic intravenous FA. Applegate et al. ²⁰ reported the recognition of the FAZ by entopic phenomenon in subjects with DR, but this was a subjective method because it depended on the subjects' responses. Recently, with the introduction of adaptive optics in vision research, Tam et al. ²⁴ used an adaptive-optics SLO and Popovic et al. ²¹ used a dual-conjugate flood illumination adaptive optics system as a noninvasive method to visualize parafoveal capillaries and to detect the FAZ. Although adaptive optics imaging produces high-definition images without using fluorescein dye, this technique is not yet commercially available. In this study, we investigated the effectiveness of a novel approach to detect the FAZ noninvasively using confocal red-free imaging, which could be used in current clinical practice. 

We found that a FAZ visualized on FA could also be observed via confocal red-free imaging. Without any manipulation of the images, two readers were able to identify the boundaries of the FAZ using raw red-free images. Comparisons of the area within demarcation lines showed that confocal red-free imaging is comparable to FA imaging for detecting the FAZ. The interobserver and a test–retest reliability were high, which indicated that the delineation process was reproducible. The mean size of the FAZ in DR observed using both imaging techniques was similar to that of a previous report ¹⁸ using automated FAZ segmentation, but was smaller than that documented in a report by Arend et al. ⁷ In RVO, Parodi et al. ¹³ reported an enlarged FAZ in macular RVO that was larger than our measurement because, in the present study, only five cases showed ischemic maculopathy with a definitively enlarged FAZ. An overlapping ratio of >0.75 was calculated, indicating high concordance between the morphologies viewed on the different images. The red-free image corresponded to that of the FA image even in cases with enlarged, irregular-margined FAZs on FA. This agreement was reinforced by the similar size of the FAZ and the high overlapping ratio between the two imaging techniques. 

We assumed that the reason the FAZ appears darker than the surrounding retinal tissue on confocal red-free imaging with blue reflectance may be differences in the absorption coefficient between hemoglobin and retinal chromophores. ⁵ Retinal reflectance is determined by a combination of the absorption coefficient of hemoglobin and retinal tissue. ²⁸ If there is no blood flow in the retinal tissue, the appearance of the avascular retina depends on the degree of retinal pigmentation. ²⁸ By contrast, a normally perfused retina has a small whitish area because hemoglobin partially reflects the blue wavelength of light based on the spectral reflectance of the retina. The absorption coefficient of hemoglobin is 10% lower than that of melanin at 490 nm. A FAZ that is not supplied by retinal circulation appears as a dark spot due to the combination of xanthophyll and melanin because only retinal pigmentation, which absorbs most of the blue light, influences retinal reflection. ²⁸  

We expect that confocal red-free imaging can be applied in clinical situations. This technique may be helpful for DR screening, especially in large-scale health check-ups or telemedicine. DR has been classified and screened using fundus photography, as suggested by the ETDRS group. ^29,30 However, fundus photography cannot detect the FAZ and ischemic maculopathy can occur independent of this stage of DR, making this method insufficient as a screening examination. Confocal red-free imaging can provide more detailed information about disease status at screening and may also lead to the creation of new DR classifications including FAZ status in the future. The process for obtaining red-free images is simple, fast, and safe, as is fundus photography. One single captured image is sufficient to detect the FAZ in confocal red-free imaging, in contrast to FA, in which of midvenous phase images are needed to detect the FAZ. Repeated, frequent monitoring of FAZ status by noninvasive confocal red-free imaging is potentially possible because using only repeated FA for the evaluation of FAZ status is not recommended. Specifically, this confocal red-free imaging method may be the only feasible alternative method for FAZ evaluation in subjects who experience serious side effects from fluorescein. 

Although the use of confocal red-free imaging is promising, several issues should be considered when interpreting the FAZ visualized using this method. First, with confocal red-free imaging, the exact boundary of the FAZ cannot be determined because the perifoveal capillary structures are not visible using this method. The FAZ can be detected subjectively using this technique based on the difference in tone of the pixels between the FAZ and the surrounding tissue. To overcome this limitation, we are currently developing an algorithm for automated FAZ detection with confocal red-free imaging through optical density analysis, which will provide a more objective measurement of the FAZ. Additionally, we plan to apply a frame-averaging technique to improve the image quality of confocal red-free imaging, making it more precise for detecting the FAZ. Second, imaging with blue light is affected by media opacities because of its short wavelength. Even though media opacities such as severe cataract or vitreous haziness make it difficult to interpret both FA and red-free images, the image quality of FA with the help of fluorescein is superior to that of the red-free image. The image quality factor, which is the value of the presenting image quality of SD-OCT affected by media opacity, was 37.58 ± 9.89 in the eyes with severe cataract in this study, and those eyes were excluded; however, we found that the quality of confocal red-free imaging was sufficient to observe the FAZ in cases with mild to moderate cataract, which had an image quality factor of >50. Third, the degree of retinal pigmentation, the presence of retinal hemorrhage, or a retinal nerve fiber bundle defect, which appears as a dark region in confocal red-free imaging, can influence FAZ interpretation because this imaging technique is based on the difference in the spectral reflectance of retinal tissue. The FAZ morphology on confocal red-free imaging was inconsistent with that on FA in cases of severe cystoid macular edema. These cases had increased central macular thicknesses of >400 μm, which indicated that changes in central macular thickness might affect FAZ interpretation on confocal red-free imaging because intraretinal cystic fluid could alter the reflectance of the macular area. Therefore, when interpreting confocal red-free images, correlations with fundus photography or OCT may be required. Further study is warranted with a larger sample size and a greater diversity of cases. 

Another limitation is that the study population was composed only of Korean subjects. Differences in retinal pigmentation may influence results for other populations. Future studies involving subjects of other races and ethnicities are therefore necessary. A small sample size and subjective FAZ delineation are additional limitations of the present study. We plan to investigate long-term changes in the FAZ in DR and RVO observed on confocal red-free imaging combined with automated FAZ detection software. 

There have been no reports of FAZ evaluation by confocal red-free imaging. We found that this imaging modality can reliably detect the FAZ and is highly correlated with FA. It has the potential to be used for noninvasively detecting and quantifying the FAZ in screening, initial evaluation, and follow-up observation of ischemic maculopathy, such as DR and RVO. 

Footnotes

 Presented in part at the annual meeting of the Association for Research in Vision and Ophthalmology, Fort Lauderdale, Florida, May 2011.

Footnotes

 Disclosure: Y.U. Shin, None; S. Kim, None; B.R. Lee, Nidek (C); J.W. Shin, None; S.I. Kim, None