Evaluation of the macular microvasculature is essential for the diagnosis and management of various retinal conditions, and for retinal vascular diseases in particular. Among other parameters, changes in the macular vessel density (VD) and the size of the foveal avascular zone (FAZ) area are of particular relevance for both diagnosis and prognosis.1–8
Optical coherence tomography angiography (OCTA) is a noninvasive method of visualizing the different vascular layers in the retina that does not require intravenous injection of dye.9–11 With OCTA, the deep retinal plexus can be viewed separately from the superficial plexus.11 Another advantage of OCTA imaging is its capability for quantitative analysis of different layers of the retina and choroid. A decrease of the macular VD has been described in patients with vascular diseases like diabetic retinopathy, retinal vein occlusion, and retinal artery occlusion.12–14 The FAZ has been reported to be enlarged and irregular in patients with diabetic retinopathy and retinal vein occlusion,15,16 whereas a decrease in size was reported in those with macular telangiectasia, foveal hypoplasia, or albinism.17–19
Many groups, including ours, have shown the possibility of quantitative analysis of the superficial and deep retinal layer using OCTA.20–22 However, several factors (eg, the methods used for measurement and analysis, the presence of retinal pathology, and different types of artifacts and the device threshold for the detection of the blood vessels) may affect the accuracy and repeatability of the results.22,23 Differences between various OCTA instruments have not been fully characterized. In this study of normal subjects, we compared measurements of the FAZ area and capillary density obtained with two different OCTA devices.
Patients and Methods
Study Design and Population
In this prospective, comparative case series, we imaged 25 eyes of 14 healthy individuals. Three participants did not complete the imaging process on both devices for both eyes. The subjects were deemed normal based on the absence of any previous ocular or systemic disease or visual symptoms and the lack of any evidence of pathological changes on clinical examination and structural OCT of the macula. Patients with any vision complaints, refractive error greater than 2.5 diopters, or history of surgical intervention (including refractive surgery) were excluded.
This study was approved by the institutional review board of the University of California, Los Angeles, and conducted in accordance with the ethical standards stated in the Declaration of Helsinki, as well as in accordance with regulations set forth by the Health Insurance Portability and Accountability Act. Written informed consent was obtained from all examined healthy individuals before they participated in the study.
All OCTA scans were performed by a single experienced examiner using two devices: a swept-source OCTA (SS-OCTA) device (DRI OCT Triton plus; Topcon, Tokyo, Japan), with a wavelength of 1,050 nm and an acquisition speed of 100,000 A-scans per second, and a spectral-domain OCTA (SD-OCTA) device (RS-3000; Nidek, Gamagori, Japan), with a wavelength of 880 nm and an acquisition speed of 53,000 A-scans per second. The axial and transversal resolutions for both devices are 7 μm and 20 μm, respectively. Each Triton OCT scan cube consists of 320 clusters of four repeated B-scans, whereas each RS-3000 OCT scan cube consists of 256 clusters of four repeated B-scans.
With both devices, a 3 mm × 3 mm macular cube centered on the fovea was selected. En face images of the superficial retinal capillary layer (SRL) and deep retinal capillary layer (DRL) were generated using preset layer segmentation. Using the Triton, the SRL was segmented from 2.6 μm beneath the internal limiting membrane (ILM) to 15.6 μm beneath the interface of the inner plexiform layer and inner nuclear layer (IPL/INL). The DRL slab was generated from 15.6 μm beneath the IPL/INL to 70.2 μm beneath IPL/INL. The automated segmentation using RS-3000 defined the SRL slab to extend from 8 μm below the ILM to the inner boundary of the INL. The en face slab for the DRL extended from the inner boundary of the INL to 88 μm below. All images were reviewed by the examiner for correct segmentation.
As described in our previous publication,24 quantitative analyses were performed using the publicly available GNU Image Manipulation Program GIMP 2.8.14 (available in public domain at http://gimp.org).
For VD measurement, a standardized threshold was set for the images from both devices. The mean brightness of the small vessels around the FAZ area was taken as a threshold for the examination of the SRL and DRL. For this purpose, a circle of 1-mm diameter was centered on the fovea. After excluding the FAZ, the mean grey value of the pixels of the vessels was chosen as a threshold to binarize the images (Figure 1). Any brightness value higher than the threshold was interpreted as vessels. The image of the SRL and DRL was inverted using the selected threshold into a binary image, and the VD was assessed as the ratio of the area occupied by vessels over the total scanned area.
Standardized vessel density measurement. (A) The original angiography en face slab. (B) Binarized image before threshold adjustment. (C) The selected 1-mm diameter area after excluding the foveal avascular zone area. The mean value of the vessels in (D) was chosen as the threshold to binarize the image. (E) Demonstrating the binarized image after setting the standardized threshold. Note the change in vessel width compared to the original image. (F) The skeletonized image based on the standardized threshold image.
Since resolution and image quality could potentially affect density assessments, we also evaluated the extent of capillaries using a vessel length density (VLD) approach, as previously described.21 For this assessment, a skeletonized VD map was generated using 1-pixel centerline extraction of the binary blood vessel map, reducing all vascular structures to a one-pixel thickness. With this method, variations in vessel width, caliber, and shape were ignored (Figure 2).
Vessel length density measurement. (A, B) The superficial retinal layer using Triton. (C, D) The superficial retinal layer using the RS-3000 Advance. (E, F) The original and the skeletonized image of the deep retinal layer using the Triton. (G, H) The corresponding images using the RS-3000 Advance.
The area of the FAZ, defined as the area inside the central border of the capillary network, was outlined manually for both the SRL and DRL, in accordance with our previously described method.24 The measured area in pixels was converted to millimeters according to the scan size (3 mm × 3 mm). Measurement of the FAZ area was made by two independent certified readers (MA, KGF) at Doheny Image Reading Center. All measurements of the macular vasculature and FAZ area in the SRL and DRL were graded without reference to the corresponding scan of different layers from the same eye or from the other eye.
SPSS statistical software version 21 (SPSS, Chicago, IL) and MedCalc version 12 (MedCalc Software, Mariakerke, Belgium) were used for statistical analysis. Shapiro-Wilk test confirmed the normal distribution of the data. Coefficient of variation (CV) and Bland-Altman plots were used to assess the reproducibility of two measuring instruments using the same set of subjects.25,26 The CV, which is calculated as the ratio of the standard deviation over the mean, is a measure of dispersion of the data around the mean. This parameter allows variation between two data series to be compared, even if the means of the two series are dramatically different. Devices with a CV less than 10% are considered highly reliable, whereas a CV less than 5% indicates very high reliability.27 Bland-Altman plots are illustrated after logarithmic transformation of the results to better illustrate the relationship between the mean and the difference. In addition, the repeatability of the measurements was calculated using intraclass correlation coefficient (ICC). A t-test for paired samples was used to compare each layer within the group. A P value less than .05 was considered significant. ICC was calculated to assess intergrader agreement. In addition, the analysis for all significant differences was repeated to account for the correlation between two eyes of each participant. For this purpose, one eye of each participant was randomly selected.
Sample size power analysis performed using G*Power 3.0 software (Heinrich-Heine-Universität Düsseldorf, Düsseldorf, Germany) revealed a power of 67% to detect a difference between the two devices. To achieve a power of 80% with a confidence interval of 95%, effect size of 0.5, and an alpha error of 0.05, a total sample size of 43 eyes is required.
Twenty-five eyes of 14 healthy individuals were included in this study. The mean ± SD age was 37.71 ± 6.59 years (range: 27 years to 50 years). Six participants were male and eight were female.
The Table shows the mean VD ratio, VLD after skeletonization, and FAZ area using the Triton OCTA and RS-3000 OCTA. The mean difference in the VD between the two devices was 0.105 for the SRL and 0.096 for the DRL. The ICC was 0.04 and 0.252 for the SRL and DRL, respectively. The CV, which is a measure of the dispersion of data around the mean, was 19.5% and 16.9% for the SRL and DRL, respectively.
Vessel Density Measurements (Ratio), Vessel Length Density (mm−1), and FAZ Area (mm2) of the Superficial and Deep Retinal Layers Using the Two Different Optical Coherence Tomography Angiography Devices
After standardizing the threshold and skeletonization of the vessels, the mean difference in the VLD between the two devices was 0.593 mm−1 for the SRL and 0.760 mm−1 for the DRL. The CV was 3.49% for the SRL and 1.07% for the DRL. The ICC was 0.593 and 0.760 for the SRL and DRL, respectively (Table).
Bland-Altman plots for the VD and VLD are shown in Figure 3. The VD showed that the differences are proportional to the mean with a wide range of distribution between the two devices, especially in the low and high VD for both the SRL and DRL. The SRL showed a wider range of distribution than the DRL. The VLD showed an equal distribution of the values close to the mean line for both the SRL and DRL.
Bland-Altman plots illustrating 95% limits of agreement between the two devices. (A, B) The vessel density (ratio) in the superficial and deep retinal layer (SRL and DRL). (C, D) The vessel length density in the SRL and DRL. (E, F) Image illustrates the foveal avascular zone area in the SRL and DRL.
For the FAZ area, the mean difference between the two devices was 0.001 mm2 in the SRL and 0.010 mm2 in the DRL. The CV was 2.17% for the SRL and 5.74% the DRL. The ICC was 0.997 and 0.953 for the SRL and DRL, respectively (Table). The level of agreement is illustrated in the Bland-Altman plots in Figure 3.
Comparing all eyes, the deep FAZ was significantly larger than the superficial FAZ area. The mean difference between the two layers was 0.086 mm2 ± 0.073 mm2 using the Triton OCTA (paired t-test; P < .001) and 0.098 mm2 ± 0.064 mm2 using the RS-3000 OCTA (paired t-test; P < .001).
Repeated analysis on one eye per participant revealed the same level of agreement and significance for all measurements.
Measurement of the FAZ area showed good intergrader agreement overall, though better for the Triton and for the SRL (ICC Triton SRL 0.972, Triton DRL 0.88; RS-3000 SRL 0.866, RS-3000 DRL 0.736).
In this study, VD and FAZ area measurements in normal eyes using two different OCTA devices were compared after adjusting the image threshold to obtain a uniform reading. We observed that both OCTA devices in this study yield statistically similar FAZ area results and VD measurements when expressed as VLD.
Previous groups studied the difference between SS-OCTA and SD-OCTA in various choroidal diseases. Novais et al. reported a higher rate of neovascular membrane detection with SS-OCTA compared to SD-OCTA.28 They also described a larger neovascular membrane area in eyes with choroidal neovascularization measured with SS-OCTA enface images than SD-OCTA, especially in type 1 choroidal neovascularization, suggesting that SS-OCTA may provide a more accurate representation of the neovascular membrane based on the increased penetration as well as reduced signal attenuation of the retinal pigment epithelium.29 Other studies reported VD measurements in the superficial and deep retinal layers using different instruments, methods of measurement, and regions of interest.7,30 One challenge of capillary density measurements based on quantifying the area occupied by vessels, however, is that these measurements can be affected by differences in instrument software, resolution, or image quality, which can make the width of vessels appear larger on lower resolution or blurred images.31 This can artificially change the density measurements. Skeletonization of vessels and choosing only the central single pixel of the vessel allows one to remove the dependence on vessel width and to express VD as the total VLD. Although this metric has its own limitations in that it will invariably underestimate the true physical density in tissue, it offers the possibility of allowing better standardization between devices. We observed exactly this type of behavior in our cohort. Whereas the standard mean VD demonstrated a CV between devices of 16% to 19%, the CV was only 1% to 3.5% for VLD metric. The ICC for the VD was low, with values between 0.07 and 0.25, whereas the skeletonized images showed a high agreement between 0.59 and 0.76. This suggests that this metric may be the preferred parameter for future studies, which involve multiple devices. Despite the relatively small cohort size, the high similarity of our measurements after excluding one eye per participant gives us confidence that the differences in our results are reliable.
Several studies have reported the FAZ area in healthy subjects using various imaging modalities.1,4,7,21,22,30,32–36 The mean FAZ area was 0.35 mm2 to 0.43 mm2 in FA, 0.30 mm2 to 0.32 mm2 by adaptive optics instruments, and 0.26 mm2 to 0.35 mm2 by OCTA. The mean FAZ area in our study was 0.33 mm2 using both the Triton and the RS-3000 OCTA devices, which is within the range of the previous reports for OCTA. Similar to the previous reports, the FAZ area was greater in the DRL. Our results show that the FAZ area can be quantified in both layers; however, agreement at the level of the superficial layer is higher than agreement at the deep retinal layer, which is consistent with previous studies. Shahlaee et al. investigated the interobserver reliability for FAZ area measurement and found higher agreement for the SRL rather than the DRL.37 Practically, the border of the FAZ was better defined and visualized in the SRL than in the deeper layers. One potential explanation for this is projection artifacts from the superficial layer, which may interfere with delineation of the FAZ border at the DRL. Another explanation is the slight interdevice difference in the preset segmentation of the retinal layers. Though both devices describe using similar conventions for the boundaries of the SRL and DRL, the SRL was slightly thicker and the DRL was slightly thinner on the Triton compared to the RS-3000. Regardless, this point should be considered when evaluating FAZ area in ischemic diseases, specifically in diseases such as diabetic retinopathy or macular telangiectasia, which affect mostly the DRL.
Our study has several limitations. The small sample size may make small differences between groups less detectable. We did not compare macular measurements from patients with different retinal diseases, and thus we cannot say what the magnitude of differences between instruments would be in those settings. Despite these limitations, to our knowledge, this is the first study reporting a comparison of the macular retinal vasculature between two different OCTA devices, including both spectral-domain and swept-source technologies. Future studies are needed to confirm our results and explore clinical relevance of this finding.
In summary, we report on a method for quantitative measurement using OCTA that appears to make comparison of vessel density between instruments more feasible. In addition, we demonstrated that FAZ area measurement is comparable between different devices, though with less agreement for the DRL. These findings may be of value in the design of future studies that may look to incorporate multiple OCTA devices.
- Arend O, Wolf S, Jung F, et al. Retinal microcirculation in patients with diabetes mellitus: Dynamic and morphological analysis of perifoveal capillary network. Br J Ophthalmol. 1991;75(9):514–518. doi:10.1136/bjo.75.9.514 [CrossRef]
- Chen X, Rahimy E, Sergott RC, et al. Spectrum of retinal vascular diseases associated with paracentral acute middle maculopathy. Am J Ophthalmol. 2015;160(1):26–34.e1. doi:10.1016/j.ajo.2015.04.004 [CrossRef]
- Gass JD. A fluorescein angiographic study of macular dysfunction secondary to retinal vascular disease. VI. X-ray irradiation, carotid artery occlusion, collagen vascular disease, and vitritis. Arch Ophthalmol. 1968;80(5):606–617. doi:10.1001/archopht.1968.00980050608006 [CrossRef]
- Kim DY, Fingler J, Zawadzki RJ, et al. Noninvasive imaging of the foveal avascular zone with high-speed, phase-variance optical coherence tomography. Invest Ophthalmol Vis Sci. 2012;53(1):85–92. doi:10.1167/iovs.11-8249 [CrossRef]
- Mendis KR, Balaratnasingam C, Yu P, et al. Correlation of histologic and clinical images to determine the diagnostic value of fluorescein angiography for studying retinal capillary detail. Invest Ophthalmol Vis Sci. 2010;51(11):5864–5869. doi:10.1167/iovs.10-5333 [CrossRef]
- Pemp B, Schmetterer L. Ocular blood flow in diabetes and age-related macular degeneration. Can J Ophthalmol. 2008;43(3):295–301. doi:10.3129/i08-049 [CrossRef]
- Tam J, Martin JA, Roorda A. Noninvasive visualization and analysis of parafoveal capillaries in humans. Invest Ophthalmol Vis Sci. 2010;51(3):1691–1698. doi:10.1167/iovs.09-4483 [CrossRef]
- Yu S, Pang CE, Gong Y, et al. The spectrum of superficial and deep capillary ischemia in retinal artery occlusion. Am J Ophthalmol. 2015;159(1):53–63.e1-2. doi:10.1016/j.ajo.2014.09.027 [CrossRef]
- Savastano MC, Lumbroso B, Rispoli M. In vivo characterization of retinal vascularization morphology using optical coherence tomography angiography. Retina. 2015;35(11):2196–2203. doi:10.1097/IAE.0000000000000635 [CrossRef]
- Jia Y, Tan O, Tokayer J, et al. Split-spectrum amplitude-decorrelation angiography with optical coherence tomography. Opt Express. 2012;20(4):4710–4725. doi:10.1364/OE.20.004710 [CrossRef]
- Spaide RF, Klancnik JM Jr, Cooney MJ. Retinal vascular layers imaged by fluorescein angiography and optical coherence tomography angiography. JAMA Ophthalmol. 2015;133(1):45–50. doi:10.1001/jamaophthalmol.2014.3616 [CrossRef]
- Bonini Filho MA, Adhi M, de Carlo TE, et al. Optical coherence tomography angiography in retinal artery occlusion. Retina. 2015;35(11):2339–2346. doi:10.1097/IAE.0000000000000850 [CrossRef]
- Matsunaga DR, Yi JJ, De Koo LO, Ameri H, Puliafito CA, Kashani AH. Optical coherence tomography angiography of diabetic retinopathy in human subjects. Ophthalmic Surg Lasers Imaging Retina. 2015;46(8):796–805. doi:10.3928/23258160-20150909-03 [CrossRef]
- Shahlaee A, Hong BK, Ho AC. Optical coherence tomography angiography features of branch retinal vein occlusion. Retin Cases Brief Rep. 2017;11(1):90–93. doi:10.1097/ICB.0000000000000297 [CrossRef]
- Conrath J, Giorgi R, Raccah D, Ridings B. Foveal avascular zone in diabetic retinopathy: Quantitative vs qualitative assessment. Eye (Lond). 2005;19(3):322–326. doi:10.1038/sj.eye.6701456 [CrossRef]
- Parodi MB, Visintin F, Della Rupe P, Ravalico G. Foveal avascular zone in macular branch retinal vein occlusion. Int Ophthalmol. 1995;19(1):25–28. doi:10.1007/BF00156415 [CrossRef]
- McAllister JT, Dubis AM, Tait DM, et al. Arrested development: High-resolution imaging of foveal morphology in albinism. Vision Res. 2010;50(8):810–817. doi:10.1016/j.visres.2010.02.003 [CrossRef]
- Provis JM, Hendrickson AE. The foveal avascular region of developing human retina. Arch Ophthalmol. 2008;126(4):507–511. doi:10.1001/archopht.126.4.507 [CrossRef]
- Spaide RF, Klancnik JM Jr, Cooney MJ. Retinal vascular layers in macular telangiectasia type 2 imaged by optical coherence tomographic angiography. JAMA Ophthalmol. 2015;133(1):66–73. doi:10.1001/jamaophthalmol.2014.3950 [CrossRef]
- Chidambara L, Gadde SG, Yadav NK, et al. Characteristics and quantification of vascular changes in macular telangiectasia type 2 on optical coherence tomography angiography. Br J Ophthalmol. 2016;100(11):1482–1488. doi:10.1136/bjophthalmol-2015-307941 [CrossRef]
- Kuehlewein L, Tepelus TC, An L, Durbin MK, Srinivas S, Sadda SR. Noninvasive visualization and analysis of the human parafoveal capillary network using swept source oct optical microangiography. Invest Ophthalmol Vis Sci. 2015;56(6):3984–3988. doi:10.1167/iovs.15-16510 [CrossRef]
- Gadde SG, Anegondi N, Bhanushali D, et al. Quantification of vessel density in retinal optical coherence tomography angiography images using local fractal dimension. Invest Ophthalmol Vis Sci. 2016;57(1):246–252. doi:10.1167/iovs.15-18287 [CrossRef]
- Bradley PD, Sim DA, Keane PA, et al. The evaluation of diabetic macular ischemia using optical coherence tomography angiography. Invest Ophthalmol Vis Sci. 2016;57(2):626–631. doi:10.1167/iovs.15-18034 [CrossRef]
- Al-Sheikh M, Akil H, Pfau M, Sadda SR. Swept-source OCT angiography imaging of the foveal avascular zone and macular capillary network density in diabetic retinopathy. Invest Ophthalmol Vis Sci. 2016;57(8):3907–3913. doi:10.1167/iovs.16-19570 [CrossRef]
- Shoukri MM, Colak D, Kaya N, Donner A. Comparison of two dependent within subject coefficients of variation to evaluate the reproducibility of measurement devices. BMC Med Res Methodol. 2008;8:24. doi:10.1186/1471-2288-8-24 [CrossRef]
- Quan H, Shih WJ. Assessing reproducibility by the within-subject coefficient of variation with random effects models. Biometrics. 1996;52(4):1195–1203. doi:10.2307/2532835 [CrossRef]
- Westgard JO, Klee GG. Quality Assurance. In: Burtis K, ed. Fundamentals of Clinical Chemistry. 2nd ed. Philadelphia, PA: WB Saunders Company; 1996:211–223.
- 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]
- Copete S, Flores-Moreno I, Montero JA, Duker JS, Ruiz-Moreno JM. Direct comparison of spectral-domain and swept-source OCT in the measurement of choroidal thickness in normal eyes. Br J Ophthalmol. 2014;98(3):334–338. doi:10.1136/bjophthalmol-2013-303904 [CrossRef]
- Popovic Z, Knutsson P, Thaung J, et al. Noninvasive imaging of human foveal capillary network using dual-conjugate adaptive optics. Invest Ophthalmol Vis Sci. 2011;52(5):2649–2655. doi:10.1167/iovs.10-6054 [CrossRef]
- Ghasemi Falavarjani K, Al-Sheikh M, Darvizeh F, Sadun AA, Sadda SR. Retinal vessel calibre measurements by optical coherence tomography angiography. Br J Ophthalmol. 2016Nov16. pii: bjophthalmol-2016–309678. doi:10.1136/bjophthalmol-2016-309678 [CrossRef]. [Epub ahead of print]
- Bresnick GH, Condit R, Syrjala S, Palta M, Groo A, Korth K. Abnormalities of the foveal avascular zone in diabetic retinopathy. Arch Ophthalmol. 1984;102(9):1286–1293. doi:10.1001/archopht.1984.01040031036019 [CrossRef]
- John D, Kuriakose T, Devasahayam S, Braganza A. Dimensions of the foveal avascular zone using the Heidelberg retinal angiogram-2 in normal eyes. Indian J Ophthalmol. 2011;59(1):9–11. doi:10.4103/0301-4738.73706 [CrossRef]
- Wu LZ, Huang ZS, Wu DZ, Chan E. Characteristics of the capillary-free zone in the normal human macula. Jpn J Ophthalmol. 1985;29(4):406–411.
- Nelson DA, Burgansky-Eliash Z, Barash H, et al. High-resolution wide-field imaging of perfused capillaries without the use of contrast agent. Clin Ophthalmol. 2011;5:1095–1106.
- Chui TY, VanNasdale DA, Elsner AE, Burns SA. The association between the foveal avascular zone and retinal thickness. Invest Ophthalmol Vis Sci. 2014;55(10):6870–6877. doi:10.1167/iovs.14-15446 [CrossRef]
- Shahlaee A, Pefkianaki M, Hsu J, Ho AC. Measurement of foveal avascular zone dimensions and its reliability in healthy eyes using optical coherence tomography angiography. Am J Ophthalmol. 2016;161:50–55.e1. doi:10.1016/j.ajo.2015.09.026 [CrossRef]
Vessel Density Measurements (Ratio), Vessel Length Density (mm−1), and FAZ Area (mm2) of the Superficial and Deep Retinal Layers Using the Two Different Optical Coherence Tomography Angiography Devices
|Triton Mean ± SD||RS-3000 Mean ± SD||Difference||ICC||CV|
|SRL||0.435±0.007||0.329 ± 0.018||0.105||0.04||19.592|
|DRL||0.453 ± 0.005||0.356 ± 0.037||0.096||0.252||16.904|
|SRL||25.446 ± 0.714||25.271 ± 1.543||0.174||0.593||3.49|
|DRL||30.759 ± 0.791||30.243 ± 1.230||0.516||0.76||1.07|
|FAZ area (mm2)|
|SRL||0.334 ± 0.119||0.333 ± 0.117||0.0015||0.997||2.171|
|DRL||0.421 ± 0.114||0.432 ± 0.091||0.0107||0.953||5.74|