Choroidal Thickness Influences Near-Infrared Reflectance Intensity in Eyes With Geographic Atrophy Due To Age-Related Macular Degeneration

Abstract

Purpose: To evaluate the effects of retinal and choroidal thickness on near-infrared reflectance (NIR) scanning laser ophthalmoscopy in eyes with geographic atrophy (GA) secondary to non-neovascular age-related macular degeneration (AMD).

Methods: This was a cross-sectional review of the clinical records and multimodal imaging data of eyes diagnosed with GA secondary to non-neovascular AMD. Imaging modalities included color fundus photography, fundus autofluorescence, NIR, and structural spectral-domain optical coherence tomography (SD-OCT). On SD-OCT images, the foveal retina thickness and the subfoveal choroidal thickness were measured by two independent readers. Near-infrared reflectance intensity within areas of GA was subjectively graded as hyperreflective, isoreflective, or hyporeflective and objectively estimated by using ImageJ to calculate the mean gray scale value within each GA area. A linear regression analysis was performed to model the relationship between mean NIR gray scale value and retinal and choroidal thickness.

Results: One hundred four eyes of 104 patients with a mean age of 81.3 years (SD: ±8.3) were included. The area of GA was hyperreflective on NIR in 88 eyes (85%), isoreflective in 13 eyes (12%), and hyporeflective in 3 eyes (3%). The mean foveal retinal thickness was 101.5 μm (SD: ±54) showing no significant relationship with mean NIR (P = 0.464); and the mean subfoveal choroidal thickness was 172.6 μm (SD: ±114.7) showing a statistically significant relationship with mean NIR intensity in the linear regression analysis (r = 0.590; r² = 0.348; P < 0.00001).

Conclusions: Variations in choroidal thickness appear to influence NIR intensity in areas of GA and have the potential to affect image interpretation. The recognition of this relationship may provide useful information regarding choroidal thickness.

Geographic atrophy (GA) is an advanced and progressive manifestation of age-related macular degeneration (AMD) that occurs in 20% to 35% of cases of late-stage AMD.^1–4 It is characterized by the presence of areas of degeneration and loss of retinal pigment epithelium (RPE), photoreceptors, and choriocapillaris with a diameter of at least 175 µm which may progressively enlarge over time and correspond to areas of absolute scotoma.^5,6 

High-quality stereoscopic color fundus photographs have been the gold standard for the assessment of GA for many years.^1–6 However, advances in imaging technology, particularly the use of fundus autofluorescence (FAF), have provided diagnostic and prognostic information valuable for the analysis of GA.⁷ Currently, various multimodal imaging approaches inclusive of color fundus photography, FAF, and cross-sectional and en face spectral-domain optical coherence tomography (SD-OCT) are being evaluated for the assessment of GA size and progression rates. These methods are helpful in the identification of potential candidates for therapeutic clinical trials and the evaluation of the effects of these novel therapies.^8,9 

Near-infrared imaging of the fundus is an alternative modality for retinal imaging that has shown to be of special value in the visualization of subretinal structures because it easily penetrates the neuroretinal tissue.^10,11 Current confocal scanning laser ophthalmoscopes use an 820-nm wavelength to provide high definition near-infrared reflectance (NIR) images.¹² The main advantage of this technique compared to conventional color fundus imaging is enhanced ability to penetrate media opacities, as no flash is required.¹² The confocal aperture helps eliminate backscattering light from outside the plane of focus and thereby enhances image contrast and spatial resolution. The utility of NIR in the diagnosis of non-neovascular AMD has been previously reported.⁹ Increased NIR intensity seen in areas of GA shows good correspondence with areas of hypoautofluorescence in FAF⁹ (Fig. 1). 

Figure 1

 

Right eye of an 84-year-old male with GA secondary to non-neovascular age-related macular degeneration. (A) Fundus autofluorescence shows multifocal hypoautofluorescent areas representing GA. The fovea is involved. (B) Near-infrared reflectance shows well-delineated areas of hyperreflectivity corresponding to the areas of GA.

In this study, we analyzed the effects of retinal and choroidal thickness on NIR intensity in areas of GA secondary to non-neovascular AMD. We showed that variations in choroidal thickness influencing NIR intensity may have the potential to effect image interpretation. 

Methods

This cross-sectional observational study was approved by the Western Institutional Review Board (Olympia, WA, USA). It adhered to the tenets of the Declaration of Helsinki and complied with the Health Insurance Portability and Accountability Act of 1996. 

This was a cross-sectional review of the clinical records and multimodal imaging data of eyes diagnosed with GA secondary to non-neovascular AMD seen by a single physician (KBF) at the Vitreous Retina Macula Consultants of New York. Inclusion criteria comprised patient age > 50 years with GA secondary to non-neovascular AMD. Only one eye was included for each patient. If both eyes were eligible, then only the right eye was included in the analysis. Exclusion criteria included the diagnosis of other ocular conditions associated with macular atrophy such as neovascular AMD, presence of choroidal neovascularization secondary to other retinal disease, macular dystrophies, intraocular inflammatory disorders, angioid streaks, high myopia (>6 diopters), or a history of prior macular laser. 

Data were collected through a review of medical records and prior multimodal imaging. Imaging data included color fundus photography (Topcon TRC 501X fundus camera; Topcon Medical Systems, Tokyo, Japan), fluorescein angiography (Topcon TRC 501X fundus camera; or Heidelberg Spectralis HRA-OCT [Heidelberg Engineering, Dossenheim, Germany]), indocyanine green angiography (Topcon TRC 501X fundus camera or Heidelberg Spectralis HRA-OCT), fundus autofluorescence (Topcon TRC 501X fundus camera or Heidelberg Spectralis HRA-OCT), NIR (Heidelberg Spectralis HRA-OCT), and structural SD-OCT (Heidelberg Spectralis HRA-OCT). 

Demographic data collected included age, sex, and affected eye. Clinical data collected included the presence of uni- or multifocal GA lesions and the presence or absence of foveal involvement as determined by multimodal assessment including color fundus photographs, FAF, NIR, and SD-OCT. The SD-OCT scans centered at the fovea were analyzed. Retinal thickness at the foveal center was manually measured with the caliper function within the Heidelberg Eye Explorer software (version 6.3.4.0; Heidelberg Engineering) as the distance between internal limiting membrane and the innermost point of the RPE–Bruch's membrane complex expressed in micrometers. At the same point, the subfoveal choroidal thickness was measured as the distance between the outermost point of the RPE–Bruch's membrane complex and sclerochoroidal junction (micrometers). The measurements were provided by two independent readers, and the mean of both values was used for the statistical analysis. In cases where the measurements differed by more than 10 μm, a third reader was used to provide a third measure, and the mean of the three values was used. 

The NIR images were extracted from the SD-OCT scan by using the Heidelberg Eye Explorer software and then analyzed by using ImageJ (http://imagej.nih.gov/ij/; provided in the public domain by the National Institutes of Health, Bethesda, MD, USA). To obtain an objective estimate of NIR intensity, we adopted methods used in prior studies quantifying fundus autofluorescence by using ImageJ to calculate the mean gray scale value within each GA area.¹³ The mean gray value function calculates the average gray value within a manually selected area (area of GA) by dividing the sum of gray values for each pixel within this region by the total number of pixels. The values are expressed in pixels. In eyes showing ill-defined GA margins on NIR, FAF and color photographs were used to help define the region to be analyzed. In eyes showing multifocal GA, the largest area was used for the analysis. In addition to the objective grading of GA NIR intensity, a subjective grading of NIR intensity within the largest area of GA was performed for each eye. These GA areas were classified as hyperreflective, isoreflective, or hyporeflective from their appearance compared to surrounding areas of healthier retina visible within each image. To control for variability in the gray intensity values that might be related to variations in illumination intensity during scan acquisition, pupil size, or media opacity, a correction factor was applied from the mean gray value obtained over a large retinal vein present within an area unaffected by GA. It was assumed that, unlike many other retinal structures visible within the NIR images, the NIR intensity of the large veins would be minimally affected by AMD. Both the mean gray value and the corrected mean gray value were used in the statistical analysis. 

All data were collected in an Excel document (Microsoft Excel 2016 for Mac, version 15.14; Microsoft Corp., Redmond, WA, USA). The statistical analysis was performed with IBM SPSS statistics software for Mac (version 20.0.0; SPSS, Inc., Chicago, IL, USA) by using linear regression to model the relationship between the quantified reflectivity of NIR images and the retinal and choroidal thickness. Statistical significance was defined as P < 0.05. 

Results

One hundred four of 104 patients (45 males, 59 females) diagnosed with GA secondary to non-neovascular AMD were included in the analysis. There were 92 right eyes (88%) and 12 left eyes (12%). Mean patient age was 81.3 years (SD: ±8.3; range, 55–99 years). A single lesion was present in 55 eyes (53%), whereas 49 eyes (47%) showed multifocal areas of GA. Foveal involvement was present in 93 eyes (89%), and 11 eyes (11%) showed foveal sparing. The area of GA appeared hyperreflective on NIR in 88 eyes (85%), isoreflective in 13 eyes (12%), and hyporeflective in 3 eyes (3%) (Figs. 2, 3). Focus adjustments during acquisition of NIR images did not appear to alter the difference in NIR intensity between the areas of GA and surrounding fundus regions (Supplementary Fig. S1). 

Figure 2

 

Three left eyes from different patients showing the variation in NIR intensity occurring within areas of GA. (A–C) Fundus autofluorescence showing areas of unifocal GA in the left eye of three different patients. (D–F) Near-infrared reflectance images of the same eyes shown in (A–C) show variable reflectivity within the area of GA: (D) hyperreflective, (E) isoreflective, and (F) hyporeflective. (G) Spectral-domain OCT of the same eye shown in (A) and (D). The horizontal scan centered at the fovea shows hypertransmission in the area of atrophy associated with a thin choroid (subfoveal choroidal thickness: 88 μm). (H) Spectral-domain OCT scan of the same eye shown in (B) and (E). The horizontal scan centered at the fovea shows hypertransmission in the area of atrophy associated with normal choroidal thickness (subfoveal choroidal thickness: 181 μm). (I) Spectral-domain OCT scan of the same eye shown in (C) and (F). The horizontal scan centered at the fovea shows hypertransmission in the area of atrophy associated with a thick choroid (subfoveal choroidal thickness: 460 μm).

Figure 3

 

Variations in near-infrared reflectance intensity within areas of multifocal geographic atrophy. (A–C) Near-infrared reflectance images showing variable reflectivity: (A) hyperreflective, (B) isoreflective, and (C) hyporeflective. (D) Spectral-domain OCT of the corresponding patient in (A). The scan centered in the fovea shows the classical choroidal hypertransmission sign in the areas of atrophy associated with a thin choroid (subfoveal choroidal thickness: 98 μm). (E) Spectral-domain OCT scan of the corresponding patient in (B). The scan centered in the fovea demonstrates multifocal areas of choroidal hypertransmission associated with a preserved choroidal thickness (subfoveal choroidal thickness: 240 μm). (F) Spectral-domain OCT scan of the corresponding patient in (C). The scan centered in the fovea evidences a diffuse choroidal thickening without visualization of the scleral–choroidal boundary and limited visualization of the choroidal hypertransmission sign (subfoveal choroidal thickness: 364 μm).

The mean foveal retinal thickness was 101.5 μm (SD: ±54; range, 12–239 μm) and the mean subfoveal choroidal thickness was 172.6 μm (SD: ±114.7; range, 17–533 μm). The mean gray value of the area of GA was 148.4 pixels (SD: ±26.5; range, 79–211 pixels). The mean gray value of the vein wall was 59.8 pixels (SD: ±18.3; range, 26–107 pixels). After applying the correction factor, the mean corrected gray value was 147.9 pixels (SD: ±27.1; range, 68–209 pixels). 

The relationship between the mean gray value and the retinal thickness did not show any significant relationship (P = 0.215), whereas the mean gray value and the subfoveal choroidal thickness showed a statistically significant relationship (r = 0.295; r² = 0.087; P = 0.002). The relationship between the mean corrected gray value and both the subfoveal choroidal thickness and the foveal retinal thickness on the linear regression analysis is shown in Figure 4. The linear regression showed a highly statistically significant relationship between the mean corrected gray value and the subfoveal choroidal thickness (r = 0.590; r² = 0.348; P < 0.00001), whereas there was no statistically significant relationship between the mean corrected gray value and the foveal retinal thickness (r = 0.021; r² = 0.005; P = 0.464). 

Figure 4

 

Linear regression curves showing the relationship between both choroidal and retinal thickness with NIR intensity within areas of geographic atrophy. There is a good correlation of the NIR reflectivity and the thickness of the underlying choroidal tissue. This relationship explains up to 35% (r² = 0.348) of the changes in NIR intensity; however, no statistical relationship is shown between the retinal thickness and the NIR reflectivity (r² = 0.005).

Discussion

While traditionally the diagnosis of GA has relied primarily on color fundus photography, other modes of fundus imaging including FAF, SD-OCT, and NIR may offer certain advantages in terms of sensitivity, reproducibility, and lesion quantification. The availability of NIR, often acquired simultaneously with SD-OCT, can provide high-quality images of the fundus having a precise correlation with the tomographic features.¹⁴ As has been reported previously,⁹ we found that most areas of GA in AMD eyes show NIR hyperreflectivity. This observation was confirmed in 85% of the eyes in our series. However, 12% of our eyes showed isoreflective lesions and 3% of our cases showed hyporeflective lesions. We were able to show a significant association between variations in estimated NIR intensity within areas of GA and the underlying choroidal thickness, with differences in choroidal thickness appearing to account for up to 9% to 35% of this variation (Fig. 4). 

A benefit of NIR over many other imaging modalities is minimal light scattering through hazy media, which enhances its ability to image the retina through small pupils and fundus hyperpigmentation, hemorrhage, and exudation.^11,15,16 This noninvasive technique is more comfortable for patients as compared to the bright flash of flood-illuminated fundus photography or the intense blue light of confocal scanning laser ophthalmoscopy FAF. Imaging with NIR has been used to detect early changes in the outer retina, RPE, and choroid^11,16 and is frequently used as the fundus imaging modality acquired with structural SD-OCT scans.¹⁴ One potential limitation of NIR imaging is the presence of hyperreflective artifacts in up to 25% of eyes.¹⁷ This artifact appears to be related to reflection or light-scatter due to posterior chamber intraocular lenses, as it occurs almost exclusively in pseudophakic patients.¹⁷ This artifact should be differentiated from the well-know physiologic increase of reflectivity at the level of the fovea in NIR imaging. ^9,17 

Previous studies have reported the utility of NIR in detecting the choroidal lesions associated with neurofibromatosis,^18,19 choroidal nevi,¹² and the fundus changes occurring in pseudoxanthoma elasticum,²⁰ acute macular neuroretinopathy,²¹ and paracentral acute middle maculopathy.²² Imaging with NIR has been shown to be superior to FAF for detecting lacquer cracks in high myopia before the development of RPE alterations.²³ In eyes with neovascular AMD, NIR has been described as useful for the detection of pigment epithelial detachments and other exudative changes.^24–26 Measurements of altered NIR appear to show a strong correlation with areas of neovascularization detected on FA,²⁶ with different neovascular lesion subtypes showing characteristic NIR features.^26,27 

The value of NIR imaging for non-neovascular AMD has been reported.^9,14 Near-infrared reflectance provides high-quality images of drusen and pigmentary changes, with areas of GA typically appearing hyperreflective compared to surrounding retinal structures.⁹ Presumably, this increase in NIR intensity is related to hypertransmission of the scanning laser in areas where loss of outer retinal structures and RPE enhances the detection of light reflected from the underlying sclera. In this study, we found hyperreflective GA lesions in 83% of cases. However, the lesions showed a wide range of NIR intensity. Interestingly, 15% of our eyes showed isoreflective GA lesions, and 2% of eyes showed hyporeflective lesions (Figs. 2, 3). It has been reported that the sensitivity of the detection of foveal involvement by GA may be increased with the inclusion of NIR compared to FAF alone.⁹ On FAF, foveal xanthophyll pigments produce central hypoautofluorescence that may be difficult to distinguish from that related to GA.⁹ Similarly, our results showed that increased choroidal thickness may confound NIR interpretation, as eyes with thick choroids may show isoreflective or even hyporeflective GA areas. We hypothesize that despite hypertransmission of the near-infrared scanning laser in areas of GA, a choroid of sufficient thickness may mask the reflectance from the underlying sclera (Fig. 5). In support of this theory is the effect of subretinal fluid occurring over areas of RPE loss on the NIR intensity. When subretinal fluid is present, NIR intensity is reduced but, following its resolution, reflectivity increases (Supplementary Fig. S2). 

Figure 5

 

Variations in NIR intensity related to changes in choroidal thickness in an 84-year-old female with central GA. (A) Fundus autofluorescence shows a homogenous region of hypoautofluorescence corresponding to the area of GA. (B) Near-infrared reflectance shows an isoreflective appearance within the GA region in the central and temporal area (blue line), and a hyperreflective appearance in the nasal area (green line). The discontinuous yellow lines indicate the GA margins. (C) Spectral-domain OCT centered at the fovea shows progressive choroidal thinning extending from the temporal to nasal macula and is illustrated in the shaded yellow area in (D).

We found no relationship between the retinal thickness at the fovea and estimated NIR intensity within areas of GA. As the photoreceptor outer segments are largely absent within areas of GA, our methodology did not allow us to assess the effect of healthy retina with intact outer segments on NIR intensity. Our results do indicate that, within areas of GA, the remaining retinal layers have little effect on NIR intensity. 

We acknowledge several limitations in the present study. We did not control for other factors that may have influenced NIR intensity within GA lesions, including medial clarity, refractive error, axial length, lens status, or the ratio of choroidal intravascular to extravascular volume. Since, to our knowledge, no standardized methodology for quantifying NIR intensity exists, we adapted methodology from prior studies quantifying FAF images.¹³ Also, our use of a correction factor based on the mean gray value obtained over a large retinal vein has not been validated in a cohort of normal eyes. However, the relationship between NIR intensity within the GA area with subfoveal choroidal thickness was found to be highly statistically significant. 

In summary, NIR is clearly a useful imaging modality that provides high-quality fundus images that are often acquired simultaneously with SD-OCT. While NIR appears sensitive for the detection of GA in most eyes with AMD, our results showed that variations in choroidal thickness may reduce its utility in some patients having thicker choroids. While mean choroidal thickness in eyes with GA is thinner than that of age-matched normal controls,²⁸ within our cohort, we found a wide range of choroidal thickness (17–533 μm). The relationship between NIR intensity and choroidal thickness helps explain why in different patients NIR intensity within GA shows greater variability than that seen with FAF, even when their clinical findings appear similar. Recognition of this relationship may help clinicians better interpret NIR in eyes with GA and may prompt them to look more closely at choroidal thickness when evaluating these images. Further analysis of additional ocular features that may influence NIR intensity is warranted. Additionally, whether variations in choroidal thickness have a similar effect on NIR intensity with areas of macular atrophy caused by other diseases such as macular dystrophies remains to be explored. 

Acknowledgments

Supported by the LuEsther T. Mertz Retinal Research Center, Manhattan Eye, Ear, and Throat Hospital, New York, New York, United States, and The Macula Foundation, Inc., New York, New York, United States. 

Disclosure: R. Dolz-Marco, Alcon (F), Allergan (F), Bayer (F), Heidelberg Engineering (F), Novartis (F), Thea (F); O. Gal-Or, None; K.B. Freund, Genentech (C), Optos (C), Optovue (C), Heidelberg Engineering (C) Bayer HealthCare (C)