Ophthalmic Surgery, Lasers and Imaging Retina

Clinical Science 

Parafoveal Retinal Vessel Density Assessment by Optical Coherence Tomography Angiography in Healthy Eyes

Malvika Arya, BS; Carl B. Rebhun, BA; A. Yasin Alibhai, MD; Xuejing Chen, MD, MS; Carlos Moreira-Neto, MD; Caroline R. Baumal, MD; Elias Reichel, MD; Andre J. Witkin, MD; Jay S. Duker, MD; SriniVas R. Sadda, MD; Nadia K. Waheed, MD, MPH

Abstract

BACKGROUND AND OBJECTIVE:

To assess variability in vessel density (VD) measurements across three optical coherence tomography angiography (OCTA) devices to identify a methodology that offers the least amount of variation in VD, and to assess the effect of averaging of multiple scans on VD variability.

PATIENTS AND METHODS:

Fifteen eyes of eight healthy individuals were imaged consecutively on three OCTA devices. Segmentations at the superficial, deep, and full retinal layers were generated. Repeat scans for each retinal layer were registered and averaged to generate one OCTA image. Two different automated thresholding techniques were used to calculate vessel area density (VAD) from binarized images and vessel skeleton density (VSD) from skeletonized images. Vessel length, a linear measure of the combined lengths of vessels, was calculated. Foveal avascular zone (FAZ) area was measured.

RESULTS:

All three OCTA devices were significantly different (P < .0001). This finding remained after averaging images (P < .0001). VSD was more repeatable within a device but less reproducible across devices. Conversely, VAD demonstrated less repeatability but greater reproducibility. Differences in VSD between devices were systematic and attributable to differences in resolution. Vessel length, unaffected by resolution, demonstrated no significant differences between the devices (P > .107). There was no significant difference in FAZ area across devices (P = .51). After averaging images, VD was significantly different from the single images for each device and plexus (P < .05) but remained within 1% of the value of a single scan.

CONCLUSIONS:

OCTA devices show variability in VD for healthy individuals. With greater repeatability, VSD appeared useful for following a patient on one device. VAD and vessel length seemed ideal for comparing vessel parameters between OCTA devices. After averaging multiple scans, VSD remained within 1% of a single scan, for which clinical significance remains to be determined. Caution is advised when comparing quantitative analyses across OCTA devices.

[Ophthalmic Surg Lasers Imaging Retina. 2018;49:S5–S17.]

Abstract

BACKGROUND AND OBJECTIVE:

To assess variability in vessel density (VD) measurements across three optical coherence tomography angiography (OCTA) devices to identify a methodology that offers the least amount of variation in VD, and to assess the effect of averaging of multiple scans on VD variability.

PATIENTS AND METHODS:

Fifteen eyes of eight healthy individuals were imaged consecutively on three OCTA devices. Segmentations at the superficial, deep, and full retinal layers were generated. Repeat scans for each retinal layer were registered and averaged to generate one OCTA image. Two different automated thresholding techniques were used to calculate vessel area density (VAD) from binarized images and vessel skeleton density (VSD) from skeletonized images. Vessel length, a linear measure of the combined lengths of vessels, was calculated. Foveal avascular zone (FAZ) area was measured.

RESULTS:

All three OCTA devices were significantly different (P < .0001). This finding remained after averaging images (P < .0001). VSD was more repeatable within a device but less reproducible across devices. Conversely, VAD demonstrated less repeatability but greater reproducibility. Differences in VSD between devices were systematic and attributable to differences in resolution. Vessel length, unaffected by resolution, demonstrated no significant differences between the devices (P > .107). There was no significant difference in FAZ area across devices (P = .51). After averaging images, VD was significantly different from the single images for each device and plexus (P < .05) but remained within 1% of the value of a single scan.

CONCLUSIONS:

OCTA devices show variability in VD for healthy individuals. With greater repeatability, VSD appeared useful for following a patient on one device. VAD and vessel length seemed ideal for comparing vessel parameters between OCTA devices. After averaging multiple scans, VSD remained within 1% of a single scan, for which clinical significance remains to be determined. Caution is advised when comparing quantitative analyses across OCTA devices.

[Ophthalmic Surg Lasers Imaging Retina. 2018;49:S5–S17.]

Introduction

Optical coherence tomography angiography (OCTA) is a noninvasive imaging technique that captures depthresolved angiographic data of the choroidal and retinal vasculature and superimposes its analysis on simultaneously acquired structural OCT images. Angiographic data are acquired by analyzing decorrelation signals generated by the movement of erythrocytes within blood vessels.1 The three-dimensional volumetric cube scan can be scrolled through and analyzed by generating automated or manually adjusted segmentation slabs of varying retinal thickness at different retinal levels thereby selectively depicting various vascular plexuses. OCTA has enhanced our understanding of type and location of various pathognomonic vascular changes in chorioretinal disorders, such as diabetic retinopathy, retinal arterial and venous occlusions, macular telangiectasia, and choroidal neovascularization.2–10

Current commercially available OCTA devices use either spectral-domain (SD) or swept-source (SS) technology, based on the illumination source and detection method of the OCT. SD-OCTA is currently commercially available, whereas SS-OCTA devices are only available for research protocols in the United States. SD-OCT employs a broad-bandwidth light source with a wavelength of 840 nm. In comparison, SS-OCT systems use a longer wavelength at 1,050 nm, which has deeper signal penetration into the choroid and thus has potential to allow for improved resolution of choroidal vascular signal beneath the retinal pigment epithelium. Compared with the SD-OCT acquisition speeds of around 70,000 A-scans per second, SS-OCT systems are also equipped with faster scan times (100,000 to 400,000 A-scans per second) that allow for more dense imaging but slightly lower axial resolution.11

Additionally, different OCTA devices use different algorithms to analyze the intensity and phase information of blood flow contained within the returning signal. For example, the RTVue XR Avanti with AngioVue (Optovue, Fremont, CA), an SD-OCT device, employs the intensity-based split-spectrum amplitude decorrelation angiography algorithm, whereas the Cirrus HD-OCT 5000 with AngioPlex (Carl Zeiss Meditec, Dublin, CA), also an SD-OCT device, employs the optical microangiography algorithm, which analyzes both phase and intensity to calculate the flow signal.12,13 The Plex Elite 9000 (Carl Zeiss Meditec) also uses the optical microangiography algorithm in its analysis but is an SS-OCTA device. Table 1 summarizes the specifications of each of these three devices.

OCTA Device Specifications

Table 1:

OCTA Device Specifications

Variations in hardware and proprietary software of OCTA devices may lead to differences in their image analysis and output.14,15 Parafoveal vessel density measurement and size of the foveal avascular zone (FAZ) are quantifiable metrics of posterior pole vascularity that may change physiologically with advancing age, or pathologically with an underlying retinal vascular disorder.16 Various studies have assessed the repeatability and reproducibility of quantitative metrics of OCTA in healthy eyes with mixed results. Yanik et al. evaluated the diurnal repeatability of vessel density and FAZ area in the superficial and deep capillary plexuses on the RTVue XR Avanti.15 Venugopal et al. found similar repeatability of superficial plexus parafoveal vessel density in normal and glaucomatous eyes using the automated vessel density output on the RTVue XR Avanti SD-OCT.17 You et al. also used the same device to demonstrate good repeatability of superficial plexus vessel density in both healthy eyes and eyes with diabetic retinopathy.18 Corvi et al. assessed the reproducibility of vessel density, fractal dimension, and FAZ area across seven OCTA devices for the superficial and deep capillary plexuses.19 Similarly, Munk et al. quantitatively and qualitatively compared superficial vessel density across four OCTA devices.20 Magrath et al. found significant variability when comparing FAZ area and vessel density between two SD-OCTA devices for both the superficial and deep plexuses.14 Al-Sheikh et al. further compared not only vessel density, but skeletonized vessel density as well, for superficial and deep retinal layers between two OCTA devices.21 Uji et al. then assessed the impact of averaging multiple en face OCTA images on vessel density, skeletonized vessel density, vessel diameter index, and fractal dimension.22 The methodologies and results of these studies varied with the use of different thresholding techniques for binarization, utilization of either automated or manual vessel density calculation methods, assessment of different retinal layers, analysis of binarized vessel density or skeletonized vessel density, and analysis of varying quantitative OCTA metrics.

This study was an effort to level the playing field in terms of these differences and to combine the various methodologies of the aforementioned studies to assess repeatability and reproducibility across three OCTA devices. Thus, this study compared four different OCTA quantitative metrics, different thresholding techniques, three retinal segmentations (superficial, deep, and full retinal layers), two vessel density calculation techniques (binarized vessel density and skeletonized vessel density) with a uniform manual method of calculation, and the effect of image registration and averaging. The primary purposes of this study were to assess intradevice image repeatability on sequential scanning and to determine the comparability between different OCTA devices in their assessment of healthy eyes.

Patients and Methods

In this prospective, comparative study, healthy individuals without known ocular conditions or findings on examination were recruited to be imaged on three OCTA devices located at the New England Eye Center, Tufts Medical Center, Boston, Massachusetts. The study was approved by the Institutional Review Board of Tufts Medical Center, and the research adhered to the tenets of the Declaration of Helsinki and complied with the Health Insurance Portability and Accountability Act of 1996. Informed written consent was obtained before OCTA imaging.

Fifteen eyes of eight healthy individuals were included in this study. Each eye was scanned three times consecutively in random order, on the RTVue XR Avanti with AngioVue, the Cirrus HD-OCT 5000, and the Plex Elite 9000 OCTA devices. For brevity, from this point on, each device will be referred to as Avanti, Cirrus, and Plex, respectively. Macular 3 mm × 3 mm scans centered at the fovea were obtained. The scanning density (A-scans/cube) of the 3 mm × 3 mm scanning protocol differed between devices and was 304 × 304 on the Avanti, 245 × 245 on the Cirrus, and 300 × 300 on the Plex. Automated segmentation was used to generate slabs of the superficial capillary plexus (SCP), the deep capillary plexus (DCP), and full retinal layer (FRL) by every device, except for the Avanti on which manually adjusted segmentation was required to obtain the FRL slab. The SCP, by default in the various machines, was defined as 3 μm deep to anterior aspect of the inner limiting layer (ILM) to 16 μm deep to the anterior aspect of the inner plexiform layer (IPL) on the Avanti, and as ILM to IPL on the Cirrus and Plex devices. The DCP was defined as 16 μm deep to the anterior aspect of the IPL to 69 μm deep to the IPL on the Avanti, and as IPL to the outer plexiform layer on the Cirrus and Plex machines. The outer boundary of the FRL on the Avanti was 31 μm below the retinal pigment epithelium with the inner boundary manually adjusted to above the anterior aspect of the ILM, while the FRL on the Cirrus and Plex devices was from the ILM to 70 μm posterior to the retinal pigment epithelium.

The SCP, DCP, and FRL slabs were analyzed for each OCTA scan using Fiji Image J 1.0 software (National Institute of Health, Bethesda, Maryland, USA; available at http://rsb.info.nih.gov/ij/index.html). Each 3 mm × 3 mm en face OCTA image was cropped to 2 mm × 2 mm, centered at the fovea, to eliminate the manufacturer logos. Images were scaled to measurement in millimeters based on pixel density. Additionally, for each eye, the three repetitive scans obtained on each OCTA device were registered and their signals averaged to generate one image separately for each retinal layer. Linear and elastic registration of single unregistered images was performed using the StackReg ( http://bigwww.epfl.ch/thevenaz/stackreg/) and Bun-WarpJ ( https://imagej.net/BUnwarpJ) algorithms, respectively, to optimize the alignment of images for registration and subsequent averaging of their signals. Macros automating the above steps were created to execute the commands automatically.

To measure vessel density, en face images were converted to binary images (monochrome black and white images) from their original grayscale by thresholding techniques. Any brightness above the threshold value was interpreted as flow. Two different automated thresholding techniques, the mean method and the Otsu method, were applied because these methods best retained vascular details while minimizing background noise. The Otsu method is a global method of thresholding that applies a single threshold value for the entire image.23 Each pixel is designated either to foreground or to background based on its grayscale value after comparison with the threshold value. The Otsu method's algorithm finds a thresholding value for which the sum of weighted variances for foreground and background pixel spreads is at its minimum. Thus, the Otsu method is best applicable for uniformly contrasted images. The mean method is a local thresholding technique that generates a different threshold value for each pixel based on the grayscale information of its surrounding pixels within a local window.24 This thresholding technique is more suitable for nonuniformly illuminated images.

Binary images were used to calculate vessel area density (VAD) as the percentage of area representing blood flow (white pixels) divided by the total scan area. Because vessel thickness may vary from scan to scan and across different devices based on their scanning density, binary images were converted to their skeletonized versions, preserving vascular patterns, extent, and branching. During skeletonization, each vessel was reduced to the width of one pixel, irrespective of shape and caliber. The skeletonized images were then used to determine vessel skeleton density (VSD) as the ratio of total skeletonized vascular area over total scanned area. Figure 1 represents binarized and skeletonized images of the FRL slab of the same eye across all three devices by each automated thresholding method.

Full retinal layer (FRL) en face optical coherence tomography angiography (OCTA) scans of a normal eye imaged on the Avanti, Cirrus, and Plex devices with binarized and skeletonized images and corresponding vessel density values (%). (A1, B1, C1) FRL slabs of 3 mm × 3 mm OCTA scans. (A2, B2, C2) Binary images, generated from the mean-based thresholding method and used to calculate mean vessel area density (VAD). (A3, B3, C3) Skeletonized images of the corresponding binary images, used to calculate mean vessel skeleton density (VSD). (A4, B4, C4) Binary images, generated from the Otsu method and used to calculate Otsu VAD. (A5, B5, C5) Skeletonized images of the previous corresponding binary images, used to calculate Otsu VSD.

Figure 1.

Full retinal layer (FRL) en face optical coherence tomography angiography (OCTA) scans of a normal eye imaged on the Avanti, Cirrus, and Plex devices with binarized and skeletonized images and corresponding vessel density values (%). (A1, B1, C1) FRL slabs of 3 mm × 3 mm OCTA scans. (A2, B2, C2) Binary images, generated from the mean-based thresholding method and used to calculate mean vessel area density (VAD). (A3, B3, C3) Skeletonized images of the corresponding binary images, used to calculate mean vessel skeleton density (VSD). (A4, B4, C4) Binary images, generated from the Otsu method and used to calculate Otsu VAD. (A5, B5, C5) Skeletonized images of the previous corresponding binary images, used to calculate Otsu VSD.

Additionally, vessel length, a combined measure of the lengths of all the vessels in an image was calculated. VSD was used to calculate vessel length, as it is based on reduction of each vessel to a width of one pixel irrespective of its caliber. The formula used to calculate vessel length from the VSD method was as follows:

Vessel  length (mm) = VSD × total number of pixels in image × length of an individual pixel (mm)

Last, the FAZ area was measured for each scan of the SCP slab across all the devices. The manual tracing tool on ImageJ was used to outline the FAZ by a trained reader at the Boston Image Reading Center. Before tracing, the scale was set to millimeters using the pixel dimensions of the original images, outputting all FAZ area measurements in square millimeter for comparison across devices.

Statistical Analysis

Statistical analysis was performed using R Project (R Foundation for Statistical Computing, Vienna, Austria), SPSS Version 24 (IBM Corporation, Armonk, NY), and Microsoft Excel Version 15.39 (Microsoft Corporation, Redmond, WA). Four sets of vessel density measurements for 15 eyes across devices were obtained via thresholding by the mean and the Otsu methods for binarized and skeletonized images. For each of these four methods, paired t tests were performed between single and averaged images of each retinal layer independently within each device. One-way analysis of variance (ANOVA) tests with repeated measures were performed separately for single and averaged images of each retinal layer across devices for each vessel density measurement technique. FAZ area measurements across the three different devices were also compared by the one-way ANOVA test.

To assess intradevice repeatability and interdevice reproducibility of vessel density, coefficient of repeatability (CoR) was calculated for both binarized and skeletonized images at different retinal layers. Bland-Altman plots were constructed between repeat images of each eye for the calculation of CoR. A lower CoR value indicated a tighter dispersion of data and thus higher reproducibility.

Bland-Altman plots were generated to assess bias (the average difference) between single and averaged images in each retinal layer to evaluate the degree of change in vessel density after registration and averaging within a device. After comparison of the averaged vessel density value to that of each of the three original repeats, the greatest bias value, demonstrating maximum deviation of vessel density after averaging for that particular subset, was used for analysis. Figure 2 is a representative Bland-Altman plot demonstrating bias assessment in SCP images obtained on the Plex device for the mean VSD method. Also depicted is a representative Bland-Altman plot demonstrating CoR assessment in SCP images obtained on the Plex device for the mean VSD method.

Representative Bland-Altman plots demonstrating calculation of the coefficient of repeatability (CoR) and bias for the superficial capillary plexus (SCP) on the Plex device for the mean vessel skeleton density (VSD) method. (A) Calculation of bias between single scans and averaged scans. Comparisons were made between the averaged image and each of the three repeat scans, producing three plots for each plexus on each device. Each such plot generated a bias value. The greatest bias value among three comparisons (largest deviation in vessel density) was taken for each plexus for each device and analyzed as a representative of the average difference between a single optical coherence tomography angiography (OCTA) scan and an averaged OCTA scan. The highlighted bias of −0.17 is representative of the average difference between the averaged scan and repeat scan #3 for all 15 eyes and was taken as the representative value for analysis. (B) Calculation of CoR between repeat scans of all eyes. Comparisons were made between each of the three repeats, producing three plots for each plexus under each device. The greatest CoR value among three comparisons was taken for each plexus for each device and analyzed as a representation of the precision of values for that particular group. The highlighted CoR of 0.40 is represented as half of the distance between the upper limit and lower limit and was taken as the representative value for analysis.

Figure 2.

Representative Bland-Altman plots demonstrating calculation of the coefficient of repeatability (CoR) and bias for the superficial capillary plexus (SCP) on the Plex device for the mean vessel skeleton density (VSD) method. (A) Calculation of bias between single scans and averaged scans. Comparisons were made between the averaged image and each of the three repeat scans, producing three plots for each plexus on each device. Each such plot generated a bias value. The greatest bias value among three comparisons (largest deviation in vessel density) was taken for each plexus for each device and analyzed as a representative of the average difference between a single optical coherence tomography angiography (OCTA) scan and an averaged OCTA scan. The highlighted bias of −0.17 is representative of the average difference between the averaged scan and repeat scan #3 for all 15 eyes and was taken as the representative value for analysis. (B) Calculation of CoR between repeat scans of all eyes. Comparisons were made between each of the three repeats, producing three plots for each plexus under each device. The greatest CoR value among three comparisons was taken for each plexus for each device and analyzed as a representation of the precision of values for that particular group. The highlighted CoR of 0.40 is represented as half of the distance between the upper limit and lower limit and was taken as the representative value for analysis.

Results

Fifteen eyes of eight healthy individuals were analyzed in this study. One eye was not included because of poor fixation and subsequent motion artifact. The mean age of the participants was 28.3 ± 3.6 years. Six participants were male and two were female.

En face images were transformed into binary images by automated thresholding techniques to assess VAD for each retinal layer in each device, and CoR values were calculated as a measure of repeatability of images on repetitive scanning by the same device. A lower CoR value signified greater precision and thus higher reproducibility. Table 2 summarizes the CoR values for VAD in different retinal layers for each thresholding technique. CoR values for VAD were lower for all retinal layers across all devices by the mean thresholding technique, as compared to the Otsu method.

Comparison of the Mean and Otsu Binarized Vessel Area Densities by CoR Values

Table 2:

Comparison of the Mean and Otsu Binarized Vessel Area Densities by CoR Values

Binary images generated by both thresholding methods were skeletonized. VSD was calculated for each retinal layer and CoR values were determined. Table 3 displays the CoR values for VSD assessments in different retinal layers for each thresholding technique. CoR values for VSD were lower for all retinal layers across all devices by the mean method as compared to the Otsu method. Because the mean thresholding technique demonstrated better repeatability and reproducibility for both VAD and VSD, it was used for further analysis of data in this study. All trends and findings discussed below were also calculated using the Otsu technique, and findings were similar to the values calculated using the mean technique.

Comparison of the Mean and Otsu Vessel Skeleton Densities by CoR Values

Table 3:

Comparison of the Mean and Otsu Vessel Skeleton Densities by CoR Values

CoR values were lowest for the DCP layer, suggesting superior reproducibility of this plexus on repetitive scanning in normal eyes. Furthermore, CoR values of mean VSD were lower than those of mean VAD, indicating greater repeatability and reproducibility of mean VSD as compared to mean VAD.

The mean VAD (%) and the mean VSD (%) values for each retinal layer are summarized in Table 4. A systematic interdevice trend in VAD and VSD values was noted for all eyes across all devices for all retinal layers by both thresholding methods. Figure 3 illustrates changes in mean VSD from the Avanti to the Cirrus to the Plex devices. All devices were significantly different from each other (P < .0001) for both single and averaged images. Similarly, for the mean VAD method, significant differences were noted between the devices (P < .05 for each) for all retinal layers as shown in Figure 4. However, compared with the mean VSD values, the mean VAD values showed numerically less variation across the devices (within 8%).

Mean VAD and Mean VSD Values of Various Retinal Layers in the Three Different OCTA Devices

Table 4:

Mean VAD and Mean VSD Values of Various Retinal Layers in the Three Different OCTA Devices

Variation in mean vessel skeleton density (VSD) measurements (%) for all eyes for single and averaged scans, and mathematically averaged mean VSD measurements (%) across three optical coherence tomography angiography (OCTA) devices. (A1–A3) Systematic decrease in mean VSD values from Avanti to Cirrus to Plex for all eyes in the superficial capillary plexus (SCP) (A1), deep capillary plexus (DCP) (A2), and full retinal layer (FRL) (A3) layers in single OCTA scans. (B1–B3) Systematic decrease in mean VSD values across devices for all eyes in the SCP (B1), DCP (B2), and FRL (B3) layers following averaging of repeat OCTA scans. (C1–C3) Comparison of the mean VSD means between the Avanti, Cirrus, and Plex devices in single OCTA scans for the SCP (C1), DCP (C2), and FRL (C3) layers. (D1–D3) Comparison of the mean VSD means between the Avanti, Cirrus, and Plex devices after averaging of repeat OCTA scans for the SCP (D1), DCP (D2), and FRL (D3) layers.**P < .001

Figure 3.

Variation in mean vessel skeleton density (VSD) measurements (%) for all eyes for single and averaged scans, and mathematically averaged mean VSD measurements (%) across three optical coherence tomography angiography (OCTA) devices. (A1–A3) Systematic decrease in mean VSD values from Avanti to Cirrus to Plex for all eyes in the superficial capillary plexus (SCP) (A1), deep capillary plexus (DCP) (A2), and full retinal layer (FRL) (A3) layers in single OCTA scans. (B1–B3) Systematic decrease in mean VSD values across devices for all eyes in the SCP (B1), DCP (B2), and FRL (B3) layers following averaging of repeat OCTA scans. (C1–C3) Comparison of the mean VSD means between the Avanti, Cirrus, and Plex devices in single OCTA scans for the SCP (C1), DCP (C2), and FRL (C3) layers. (D1–D3) Comparison of the mean VSD means between the Avanti, Cirrus, and Plex devices after averaging of repeat OCTA scans for the SCP (D1), DCP (D2), and FRL (D3) layers.

**P < .001

Variation in mean vessel area density (VAD) measurements (%) for all eyes for single and averaged scans, and mathematically averaged mean VAD measurements (%) across three optical coherence tomography angiography (OCTA) devices. (A1–A3) Mean VAD values from Avanti to Cirrus to Plex for all eyes in the superficial capillary plexus (SCP) (A1), deep capillary plexus (DCP) (A2), and full retinal layer (FRL) (A3) layers in single OCTA scans, displaying no uniform trend. (B1–B3) Mean VAD values across devices for all eyes in the SCP (B1), DCP (B2), and FRL (B3) layers following averaging of repetitive OCTA scans. (C1–C3) Comparison of the mean VAD means between the Avanti, Cirrus, and Plex devices in single OCTA scans for the SCP (C1), DCP (C2), and FRL (C3) layers. (D1–D3) Comparison of the mean VAD means between the Avanti, Cirrus, and Plex devices after averaging of repetitive OCTA scans for the SCP (D1), DCP (D2), and FRL (D3) layers.**P < .05ns = no significant difference.

Figure 4.

Variation in mean vessel area density (VAD) measurements (%) for all eyes for single and averaged scans, and mathematically averaged mean VAD measurements (%) across three optical coherence tomography angiography (OCTA) devices. (A1–A3) Mean VAD values from Avanti to Cirrus to Plex for all eyes in the superficial capillary plexus (SCP) (A1), deep capillary plexus (DCP) (A2), and full retinal layer (FRL) (A3) layers in single OCTA scans, displaying no uniform trend. (B1–B3) Mean VAD values across devices for all eyes in the SCP (B1), DCP (B2), and FRL (B3) layers following averaging of repetitive OCTA scans. (C1–C3) Comparison of the mean VAD means between the Avanti, Cirrus, and Plex devices in single OCTA scans for the SCP (C1), DCP (C2), and FRL (C3) layers. (D1–D3) Comparison of the mean VAD means between the Avanti, Cirrus, and Plex devices after averaging of repetitive OCTA scans for the SCP (D1), DCP (D2), and FRL (D3) layers.

**P < .05

ns = no significant difference.

Statistically significant differences were noted in the mean VAD and the mean VSD values between single and averaged images in all retinal layers across all devices, as shown in Figure 5, except for the FRL in the Avanti with mean VSD, where no such difference was seen. Although statistically significantly different, values after averaging were numerically similar to those of the single scans. In an effort to quantify these differences between averaged and single scans, bias was calculated (Table 5). Plotting of the differences between single and averaged images for a given retinal layer estimated intradevice change in vessel density values after averaging. A negative bias signified that percentage vessel density decreased after averaging and vice versa. For the mean VSD method, the average difference in vessel density after averaging was less than 1% of the mean VSD value of a single scan. This trend was noted across all three OCTA devices.

Intradevice comparison between single images and averaged images. (A1–A3) Comparison of mean vessel area density (VAD) measurements (%) at the superficial capillary plexus (SCP) (A1), deep capillary plexus (DCP) (A2), and full retinal layer (FRL) (A3) layers for all devices. (B1–B3) Comparison of mean vessel skeleton density (VSD) measurements (%) at the SCP (B1), DCP (B2), and FRL (B3) layers for all devices. The averaged scans were statistically significantly different from their single counterparts, but the numerical differences were minimal. Mean VSD of the averaged images was less than 1% of the mean VSD value of a single original scan.**P < .05ns = no significant difference.

Figure 5.

Intradevice comparison between single images and averaged images. (A1–A3) Comparison of mean vessel area density (VAD) measurements (%) at the superficial capillary plexus (SCP) (A1), deep capillary plexus (DCP) (A2), and full retinal layer (FRL) (A3) layers for all devices. (B1–B3) Comparison of mean vessel skeleton density (VSD) measurements (%) at the SCP (B1), DCP (B2), and FRL (B3) layers for all devices. The averaged scans were statistically significantly different from their single counterparts, but the numerical differences were minimal. Mean VSD of the averaged images was less than 1% of the mean VSD value of a single original scan.

**P < .05

ns = no significant difference.

Comparison of Unregistered and Registered Vessel Density Values for Mean VAD and Mean VSD Methods via Bias Values

Table 5:

Comparison of Unregistered and Registered Vessel Density Values for Mean VAD and Mean VSD Methods via Bias Values

The large variation in VSD across devices was thought to be a result of the differences in image size and resolution of the exported images. This is demonstrated in Figure 6. Because of these systematic differences in VSD, as a result of differences in the resolution of the output image, vessel length was calculated for each plexus for interdevice comparison. Vessel length is a linear combined measure of the lengths of all the vessels in an image, and it is not affected by differences in image resolution. The results of vessel length values and comparison across devices are summarized in Table 6 for individual layers. For the FRL, no significant difference was noted in vessel length between the devices. However, significant differences were noted between the Avanti and Cirrus and the Avanti and Plex devices for the superficial and deep layers, while no significant difference was seen between the Cirrus and Plex for either of these two layers.

Pixel size variation in skeletonized images of the Avanti, Cirrus, and Plex optical coherence tomography angiography devices. With a universal decrease of vessel width to one pixel, the difference in area occupied by one pixel decreases from the Avanti to the Cirrus to the Plex device as the resolution, respectively, increases. Please note, these are not enlargements of the original images but of the skeletonized images, and hence the pixelated appearance.

Figure 6.

Pixel size variation in skeletonized images of the Avanti, Cirrus, and Plex optical coherence tomography angiography devices. With a universal decrease of vessel width to one pixel, the difference in area occupied by one pixel decreases from the Avanti to the Cirrus to the Plex device as the resolution, respectively, increases. Please note, these are not enlargements of the original images but of the skeletonized images, and hence the pixelated appearance.

Vessel Length and Comparison Across OCTA Devices

Table 6:

Vessel Length and Comparison Across OCTA Devices

Last, FAZ area was measured for the SCP layer only, as this layer offered the clearest visualization and delineation of the FAZ. The average FAZ areas were 0.31 mm2 on the Avanti, 0.33 mm2 on the Cirrus, and 0.30 mm2 on the Plex. No statistically significant differences were seen across the devices (P = .51), as shown in Table 7.

Foveal Avascular Zone Area Measurements

Table 7:

Foveal Avascular Zone Area Measurements

Discussion

Accuracy, repeatability, and reproducibility are vital to the interpretation of quantitative data obtained from any ancillary test. This is true for OCTA imaging in eyes with chorioretinal vascular disorders as well. Standardization of methods to analyze OCTA data is therefore necessary to more accurately interpret these images in patients. Values generated by analysis software may vary across different OCTA devices, and it is important to quantify these differences to better understand the reproducibility across different devices and therefore the ability to use them interchangeably in clinical trials and in clinical practice. This study attempts to more fully understand the reproducibility of various quantification methods in repeated OCTA imaging in healthy eyes, and to assess the reproducibility of the values generated between devices.

In most devices, the first step to quantification of vessel density is the conversion of a gray-scale image to a binarized black and white image. In our study comparing two thresholding methods, the reproducibility of vessel density by the mean thresholding technique was superior in comparison to the Otsu method, as reflected by the mean method's lower CoR values across all devices. Furthermore, CoR values were lower for VSD measurements in comparison to VAD measurements. This observation corroborates with Al-Sheikh et al.'s findings of skeletonized vessel length density being a superior parameter of vascular flow quantification because of its higher reproducibility, allowing for better standardization between devices.21 The mean thresholding method generates a local threshold value for a pixel based on grayscale information of its surrounding pixels, possibly leading to more accurate vascular maps, whereas the Otsu method assesses a single thresholding value for the entire image. This study suggests using the mean VSD method as a quantitative parameter for monitoring consecutive OCTA scans on the same device because of its high intradevice repeatability.

In terms of interdevice reproducibility, VSD values varied significantly for different OCTA devices, irrespective of thresholding technique applied. VSD measurements were highest for the Avanti and lowest for the Plex, with a similar trend noted for each eye. VSD decreased from the Avanti (17% to 18%) to the Cirrus (10% to 12%) to the Plex (5% to 6%). This observation was attributed to differences in pixel dimension and image export resolution between devices. Figure 6 highlights pixel size variation in skeletonized images of all three OCTA devices. Because VSD is calculated based on vessels that have been reduced to a width of one pixel, VSD values generated by the Plex images were less than those generated by the Cirrus and the Avanti images, because of the smaller dimension of an individual pixel on the Plex in comparison to either of the other two devices. Because of the variability in VSD noted across the devices, a conversion factor based on pixel dimension may be calculated and applied to improve quantitative interdevice comparability. Another option would be to normalize the resolution of the image obtained using image processing techniques, with the concern that reducing resolution may result in loss of information in the images. Confirming this assertion requires further investigation.

Although mean VSD was found to be a highly repeatable parameter of parafoveal vessel density measurement within a single device, the mean VAD measurements, though displaying statistical differences, seemed more comparable across the three devices. In the current study, the mean VAD method demonstrated less repeatability than the mean VSD method, but greater comparability across devices because of more numerically similar values. Therefore, in monitoring a patient with consecutive imaging on the same device, the mean VSD method appears to be a more reliable quantitative parameter, whereas mean VAD may be preferred for monitoring the same patient on different devices.

Mean VSD systematically differed across the devices for all eyes, an observation that may be accounted for by resolution differences among the devices. To assess this hypothesis and to account for differences in image resolution, vessel length was calculated. For this quantitative metric, no significant difference was noted between the devices for the FRL. This finding suggests that images from these three devices may be comparable if using FRL vessel length as a quantitative parameter for assessment. However, for the SCP and DCP, the Avanti was significantly different from the Cirrus and Plex, while both of these Zeiss devices were not significantly different from each other. These findings may be attributable to differences in automated segmentation and projection artifact suppressing algorithms. The Cirrus and Plex use the same automated segmentation algorithm for the superficial and deep layers, while the Avanti uses different segmentation boundaries for these same layers. By taking the full retina slab on all three devices, this variation in automated segmentation between devices is nullified, and the differences in vessel length diminish. Thus, when quantitatively comparing images from different OCTA devices for a single patient, the inherent differences in automated segmentation algorithms must be kept in mind as a confounding factor. Evaluation of the full retina may be a better strategy in these situations.

Another finding of this study was that regardless of the method of vessel density analysis, the swept-source device images generated the most repeatable values (most precise results), as inferred by its lower CoR values. This observation may be attributable to the higher sampling density in the swept-source device as compared to the spectral-domain nature of the Avanti and Cirrus devices, which allows for more precise imaging and less extrapolation. Another reason could be better penetration of the longer wavelength signal in the swept-source device compared with that of the spectral-domain device. However, given that this study was conducted in young, healthy individuals with no evidence of cataracts or vitreous opacities, this explanation seems less likely.

As another method to assess interdevice comparability, FAZ measurements were compared across devices in this study. Previous studies have reported higher consistency of FAZ measurements in the superficial retinal layer because of better delineation of its margins, as compared to that in the deep retinal layer, where FAZ margins may be blurred by projection artifacts.14,25–28 FAZ areas across all three devices showed no statistically significant difference in this study. Mean FAZ areas in SCP scans ranged from 0.30 mm2 to 0.33 mm2 across the devices, which was in agreement with findings of previous studies. Thus, FAZ area measurement has excellent repeatability within a device as well as reproducibility across devices.

In this study, in an effort to minimize interdevice variability, multiple repeat scans were averaged and differences in vessel density values between single images and averaged images were assessed.29 Uji et al. calculated the skeletonized vessel density of single images and averaged images obtained on the Cirrus device for superficial and deep retinal layers.22 After assessing averages of different numbers of frames, their findings demonstrated that vessel density values of averaged images were significantly different from those of single images. They concluded that vessel density decreased with an increasing number of averaging frames. Two to five frame averages for the superficial retinal layer and three to six frame averages for the deep retinal layer were associated with significant variation in vessel density measurements. Our study concurred with the above study and showed that after averaging of three frames, vessel density measurements for each retinal plexus were significantly different from those of the single image values. However, the mean VSD value following averaging seemed to remain within ± 1% of the mean VSD value of a single scan. Therefore, the clinical utility of averaging of multiple repetitive scans is unclear, keeping in mind the discomfort of the patient in undergoing multiple consecutive scans and the time-consuming process to coregister these scans. Although we question the clinical significance of a 1% difference in vessel density, the findings in our study suggest that perhaps a single volumetric scan of a given plexus might provide reliable and reproducible data, as opposed to obtaining multiple scans for registration and averaging, at least for clinical evaluation. Of course, these data are from a study of healthy individuals and may not necessarily be generalizable to patients with significant retinal pathology in which variations in the flow speeds and ability to detect vessels may, indeed, make averaging useful. In addition, the scans used in this study were of excellent quality with high signal strength. The signal strength of images from each device was 10 on the Cirrus, at least eight on the Plex, and at least 62 on the Avanti. Registering and averaging scans may be of greater relevance in real-world clinical practice in which scans may not always be of optimal quality. Further studies in diseased eyes are needed to evaluate this.

In conclusion, after assessing two thresholding techniques for measurement of vessel density, the mean VSD parameter appeared to provide a more repeatable intradevice assessment of parafoveal vessel density. However, because of disparities in vessel density values across devices from differences in resolution and greater variability of mean VSD than mean VAD across devices, the mean VAD or vessel length measurement method seemed to provide superior interdevice reproducibility. The swept-source OCTA device used in this study demonstrated the highest repeatability. Additionally, following the registration and averaging of three repeat scans, vessel density was significantly different from that of the original single scans. However, these differences were noted to be less than 1%, for which clinical significance remains to be determined. Overall, significant variability exists among vessel density measurements between the Avanti, Cirrus, and Plex devices. Because of interdevice variability in various aspects ranging from light source to proprietary algorithms to pixel dimensions and resolution, cautious interpretation is advised when comparing quantitative analyses across different OCTA devices. However, although more studies are needed to validate this, the results of our study seem to suggest that, with careful planning and using selected parameters, interdevice comparisons may be possible in a clinical trial or clinical setting.

References

  1. 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:10.1186/s40942-015-0005-8 [CrossRef]
  2. Coscas GJ, Lupidi M, Coscas F, Cagini C, Souied EH. Optical coherence tomography angiography versus traditional multimodal imaging in assessing the activity of exudative age-related macular degeneration: a new diagnostic challenge. Retina. 2015;35:2219–2228. doi:10.1097/IAE.0000000000000766 [CrossRef]
  3. de Carlo TE, Bonini Filho MA, Chin AT, et al. Spectral-domain optical coherence tomography angiography of choroidal neovascularization. Ophthalmology. 2015;122:1228–1238. doi:10.1016/j.ophtha.2015.01.029 [CrossRef]
  4. Novais EA, Adhi M, Moult EM, et al. Choroidal neovascularization analyzed on ultrahigh-speed swept-source optical coherence tomography angiography compared to spectral-domain optical coherence tomography angiography. Am J Ophthalmol. 2016;164:80–88. doi:10.1016/j.ajo.2016.01.011 [CrossRef]
  5. Arya M, Cole ED, Sabrosa AS, et al. Visualization of choroidal neovascularization using two commercially available spectral domain optical coherence tomography angiography devices [published online ahead of print August 1, 2018]. Retina. doi:10.1097/IAE.0000000000002241 [CrossRef].
  6. Kashani AH, Chen CL, Gahm JK, et al. Optical coherence tomography angiography: a comprehensive review of current methods and clinical applications. Prog Retin Eye Res. 2017;60:66–100. doi:10.1016/j.preteyeres.2017.07.002 [CrossRef]
  7. Adhi M, Brewer E, Waheed NK, Duker JS. Analysis of morphological features and vascular layers of choroid in diabetic retinopathy using spectral-domain optical coherence tomography. JAMA Ophthalmol. 2013;131:1267–1274. doi:10.1001/jamaophthalmol.2013.4321 [CrossRef]
  8. Agemy SA, Scripsema NK, Shah CM, et al. Retinal vascular perfusion density mapping using optical coherence tomography angiography in normals and diabetic retinopathy patients. Retina. 2015;35:2353–2363. doi:10.1097/IAE.0000000000000862 [CrossRef]
  9. Choi W, Waheed NK, Moult EM, et al. Ultrahigh speed swept source optical coherence tomography angiography of retinal and choriocapillaris alterations in diabetic patients with and without retinopathy. Retina. 2017;37:11–21. doi:10.1097/IAE.0000000000001250 [CrossRef]
  10. Kashani AH, Lee SY, Moshfeghi A, Durbin MK, Puliafito CA. Optical coherence tomography angiography of retinal venous occlusion. Retina. 2015;35:2323–2331. doi:10.1097/IAE.0000000000000811 [CrossRef]
  11. Fujimoto J, Swanson E. The development, commercialization, and impact of optical coherence tomography. Invest Ophthalmol Vis Sci. 2016;57:OCT1–OCT13. doi:10.1167/iovs.16-19963 [CrossRef]
  12. Huang D, Jia Y, Gao SS, Lumbroso B, Rispoli M. Optical coherence tomography angiography using the Optovue device. Dev Ophthalmol. 2016;56:6–12. doi:10.1159/000442770 [CrossRef]
  13. Rosenfeld PJ, Durbin MK, Roisman L, et al. ZEISS Angioplex spectral domain optical coherence tomography angiography: technical aspects. Dev Ophthalmol. 2016;56:18–29. doi:10.1159/000442773 [CrossRef]
  14. Magrath GN, Say EAT, Sioufi K, Ferenczy S, Samara WA, Shields CL. Variability in foveal avascular zone and capillary density using optical coherence tomography angiography machines in healthy eyes. Retina. 2017;37:2102–2111. doi:10.1097/IAE.0000000000001458 [CrossRef]
  15. Yanik OO, Demirel S, Ozmert E, Batioglu F. Repeatability of automated vessel density and superficial and deep foveal avascular zone area measurements using optical coherence tomography angiography: diurnal findings. Retina. 2017;38(6):1238–1245. doi:10.1097/IAE.0000000000001671 [CrossRef]
  16. Iafe NA, Phasukkijwatana N, Chen X, Sarraf D. Retinal capillary density and foveal avascular zone area are age-dependent: quantitative analysis using optical coherence tomography angiography. Invest Ophthalmol Vis Sci. 2016;57:5780–5787. doi:10.1167/iovs.16-20045 [CrossRef]
  17. Venugopal JP, Rao HL, Weinreb RN, et al. Repeatability of vessel density measurements of optical coherence tomography angiography in normal and glaucoma eyes. Br J Ophthalmol. 2018;102:352–357. doi:10.1136/bjophthalmol-2017-310637 [CrossRef]
  18. You Q, Freeman WR, Weinreb RN, et al. Reproducibility of vessel density measurement with optical coherence tomography angiography in eyes with and without retinopathy. Retina. 2017;37:1475–1482. doi:10.1097/IAE.0000000000001407 [CrossRef]
  19. Corvi F, Pellegrini M, Erba S, Cozzi M, Staurenghi G, Giani A. reproducibility of vessel density, fractal dimension, and foveal avascular zone using 7 different optical coherence tomography angiography devices. Am J Ophthalmol. 2018;186:25–31. doi:10.1016/j.ajo.2017.11.011 [CrossRef]
  20. Munk MR, Giannakaki-Zimmermann H, Berger L, et al. OCT-angiography: a qualitative and quantitative comparison of 4 OCT-A devices. PLoS One. 2017;12:e0177059. doi:10.1371/journal.pone.0177059 [CrossRef]
  21. Al-Sheikh M, Falavarjani KG, Tepelus TC, Sadda SR. Quantitative comparison of swept-source and spectral-domain OCT angiography in healthy eyes. Ophthalmic Surg Lasers Imaging Retina. 2017;48:385–391. doi:10.3928/23258160-20170428-04 [CrossRef]
  22. Uji A, Balasubramanian S, Lei J, Baghdasaryan E, Al-Sheikh M, Sadda SR. Impact of multiple en Ffce image averaging on quantitative assessment from optical coherence tomography angiography images. Ophthalmology. 2017;124:944–952. doi:10.1016/j.ophtha.2017.02.006 [CrossRef]
  23. Nobuyuki O. A threshold selection method from gray-level histograms. IEEE Trans Sys Man Cybern. 1979;9(1):62–66. doi:10.1109/TSMC.1979.4310076 [CrossRef]
  24. Glasbey CA. An analysis of histogram-based thresholding algorithms. CVGIP: Graphical Models and Image Processing. 1993;55:532–537.
  25. Adhi M, Filho MA, Louzada RN, et al. Retinal capillary network and foveal avascular zone in eyes with vein occlusion and fellow eyes analyzed with optical coherence tomography angiography. Invest Ophthalmol Vis Sci. 2016;57:OCT486–OCT494. doi:10.1167/iovs.15-18907 [CrossRef]
  26. Spaide RF, Fujimoto JG, Waheed NK. Image artifacts in optical coherence tomography angiography. Retina. 2015;35:2163–2180. doi:10.1097/IAE.0000000000000765 [CrossRef]
  27. Shiihara H, Sakamoto T, Yamashita T, et al. Reproducibility and differences in area of foveal avascular zone measured by three different optical coherence tomographic angiography instruments. Sci Rep. 2017;7:9853. doi:10.1038/s41598-017-09255-5 [CrossRef]
  28. Garrity ST, Iafe NA, Phasukkijwatana N, Chen X, Sarraf D. Quantitative analysis of three distinct retinal capillary plexuses in healthy eyes using optical coherence tomography angiography. Invest Ophthalmol Vis Sci. 2017;58:5548–5555. doi:10.1167/iovs.17-22036 [CrossRef]
  29. Mo S, Phillips E, Krawitz BD, et al. Visualization of radial peripapillary capillaries using optical coherence tomography angiography: the effect of image averaging. PLoS One. 2017;12:e0169385. doi:10.1371/journal.pone.0169385 [CrossRef]

OCTA Device Specifications

SpecificationsRTVue XR AvantiCirrus HD-OCT 5000Plex Elite 9000
OCT TechnologySpectral-domainSpectral-domainSwept-source
Wavelength (nm)∼840∼840∼1,060
Acquisition Speed (A-scans/s)70,00068,000100,000
A-scan Depth (mm)∼323
Available Default OCTA Scan Protocols (mm)3 × 3, 6 × 6, 8 × 8 centered on macula; 3 × 3, 4.5 × 4.5 centered on ONH3 × 3, 6 × 6 centered on macula3 × 3, 6 × 6, 9 × 9, 9 × 15, 12 × 12 centered on macula (montage available)
Multimodal ImagingNoNoNo
Axial Resolution (µm)556.3
Transverse Resolution (µm)151520
Acquisition ScanTwo co-registered acquisition scans with eye trackingOne acquisition scan with eye-trackingOne acquisition scan with eye-tracking
B-scans/cube304 (repeated two times at each position in X and Y direction)245 (repeated four times at each position for 3 mm × 3 mm), 350 (repeated two times at each position for 6 mm × 6 mm)300 (repeated three times at each position for 3 mm × 3 mm), 500 (repeated two times for 6 mm × 6 mm, 9 mm × 9 mm, and 12 mm × 12 mm)
Acquisition Time (s)∼3∼3.5∼2.6 for 3 mm × 3 mm ∼5 for 6 mm × 6 mm
AlgorithmSplit-spectrum amplitude decorrelation angiographyOptical microangiographyOptical microangiography

Comparison of the Mean and Otsu Binarized Vessel Area Densities by CoR Values

OCTA DeviceFull Retina LayerSuperficial Capillary PlexusDeep Capillary Plexus
Mean VAD (%)Otsu VAD (%)Mean VAD (%)Otsu VAD (%)Mean VAD (%)Otsu VAD (%)
Avanti2.236.662.335.551.734.40
Cirrus3.328.002.867.142.624.83
Plex1.633.211.643.191.642.63

Comparison of the Mean and Otsu Vessel Skeleton Densities by CoR Values

OCTA DeviceFull Retina LayerSuperficial Capillary PlexusDeep Capillary Plexus
Mean VSD (%)Otsu VSD (%)Mean VSD (%)Otsu VSD (%)Mean VSD (%)Otsu VSD (%)
Avanti2.023.111.692.841.541.79
Cirrus1.492.681.202.280.961.41
Plex0.450.630.400.490.300.36

Mean VAD and Mean VSD Values of Various Retinal Layers in the Three Different OCTA Devices

OCTA DeviceMean VAD (%) ± SDMean VSD (%) ± SD
Full Retina LayerSuperficial Capillary PlexusDeep Capillary PlexusFull Retina LayerSuperficial Capillary PlexusDeep Capillary Plexus
Avanti34.89 ± 0.9530.93 ± 0.9936.41 ± 0.5018.57 ± 0.7317.24 ± 0.7018.14 ± 0.45
Cirrus37.23 ± 0.9434.26 ± 0.6133.41 ± 1.2513.07 ± 0.4412.51 ± 0.4011.38 ± 0.40
Plex44.30 ± 1.1141.31 ± 0.8536.16 ± 1.517.29 ± 0.316.75 ± 0.236.08 ± 0.26

Comparison of Unregistered and Registered Vessel Density Values for Mean VAD and Mean VSD Methods via Bias Values

OCTA DeviceMean VADMean VSD
Full Retinal LayerSuperficial Capillary PlexusDeep Capillary PlexusFull Retinal LayerSuperficial Capillary PlexusDeep Capillary Plexus
Avanti2.361.512.590.410.680.41
Cirrus1.301.542.42−0.69−0.69−0.51
Plex1.181.411.31−0.26−0.17−0.39

Vessel Length and Comparison Across OCTA Devices

Mean Vessel Length (mm) ± SD
OCTA DeviceFull Retinal LayerSuperficial Capillary PlexusDeep Capillary Plexus
Avanti7476.93 ± 293.316943.24 ± 280.887304.19 ± 181.41
Cirrus7473.18 ± 249.427156.37 ± 230.846509.94 ± 226.44
Plex9927.18 ± 419.419201.78 ± 312.968283.44 ± 351.76
Compared DevicesFull Retinal LayerSuperficial Capillary PlexusDeep Capillary Plexus
Avanti-Cirrusns (P = .107)P < .0001P = .001
Avanti-Plexns (P = .244)P < .0001P = .042
Cirrus-Plexns (P = 1)ns (P = 1)ns (P = .444)

Foveal Avascular Zone Area Measurements

OCTA DeviceMean FAZ Area (mm2) ± SDP Value
Avanti0.312 ± 0.073.51
Cirrus0.333 ± 0.063
Plex0.305 ± 0.071
Authors

From the New England Eye Center, Tufts Medical Center, Boston, MA (MA, CBR, AYA, XC, CRB, ER, AJW, JSD, NKW); Hospital de Olhos do Parana, Curitiba, Brazil (CMN); and Doheny Eye Institute, David Geffen School of Medicine at UCLA, Los Angeles (SRS).

Presented at the 2017 Retina Research Scholar Honoree Program Symposium in New Orleans.

Dr. Duker did not participate in the editorial review of this manuscript.

Supported in part by a grant from the Macula Vision Research Foundation.

Dr. Baumal is a speaker for Optovue and Carl Zeiss Meditec and has served on the advisory board for Genentech. Dr. Duker is a consultant for Allergan, Aura Biosciences, Bayer, Helio Vision, Novartis, Roche, Santen and Thrombogenics, and receives research support from Carl Zeiss Meditec and Optovue. He also serves on the Board of Directors for Eleven Oncology and EyePoint Pharmaceuticals and is a Hemera Biosciences stockholder. Dr. Sadda is a consultant for Optos, CenterVue, and Heidelberg Engineering and receives research support from Carl Zeiss Meditec and Optos. Dr. Sadda has also served as a consultant for Novartis, Genentech/Roche, Allergan and Amgen. Dr. Waheed is a consultant for Optovue and receives research support from Carl Zeiss Meditec, Topcon Medical Systems, and Nidek.

Address correspondence to Nadia K. Waheed, MD, MPH, Boston Image Reading Center, New England Eye Center at Tufts Medical Center, 260 Tremont Street, Biewend Building, 9 - 11th Floor, Boston, MA 02116; email: nadiakwaheed@gmail.com.

Received: June 28, 2018
Accepted: August 01, 2018

10.3928/23258160-20180814-02

Sign up to receive

Journal E-contents