AIOS – Rema Mohan Award
Dr. Santosh Gopi Krishna Gadde, G16678, Ms.Neha Anegondi, Dr. Abhijit Sinha Roy, Dr. Naresh Kumar Yadav
Introduction
Diabetic mellitus (DM) is a chronic condition and is estimated to rise to 360 million by the year 2030 according to World Health Organization.1 Patients with DM are at high risk of developing diabetic retinopathy, a progressive condition that causes alterations in the retinal microvasculature.2,3 DM can lead to conditions such as retinal neovascularization, macular edema and retinal ischemia.2, 3 Optical coherence tomography angiography (OCTA) is a recent, non-invasive, dye less imaging technique and can be used for evaluating the retinal vasculature by capturing the dynamic motion of erythrocytes.4 OCTA has been shown to be a useful imaging modality for evaluation of ophthalmologic diseases such as DR, artery and vein occlusions, and glaucoma.5
Recent qualitative and quantitative studies on DR using OCTA have shown modifications in foveal avascular zone (FAZ) and retinal microvasculature compared to normal eyes.6-8 This study aimed at developing an automated method for segmentation and quantification of the foveal avascular zone (FAZ), vessel density, spacing between large vessels and spacing between small vessels in the OCTA images of the superficial and deep retinal vascular plexus. The study used local fractal analyses to quantify the local variations in both superficial and deep retinal vasculature of the OCTA scans. Correlations of levels of fasting blood sugar (FBS), post-prandial blood sugar (PPBS), blood pressure, body mass index (BMI), hemoglobin (Hb), glycosylated hemoglobin (HbA1c), low density lipoprotein (LDL) and high density lipoprotein (HDL) with the retinal vascular parameters were also analyzed.
Purpose:
To correlate retinal vascular features with severity and systemic indicators of diabetic retinopathy (DR) using optical coherence tomography angiography (OCTA).
Methods
This prospective, observational, cross-sectional study was approved by the institutional ethics committee of Narayana Nethralaya Super-specialty eye hospital, Bangalore, India. The research followed the tenets of the Declaration of Helsinki. Written informed consent was obtained from all the subjects before imaging. 209 eyes of 122 type 2 diabetes mellitus Indian patients with DR (age 38-80, female to male ratio=0.51) and 60 eyes of 31 normal Indian subjects (age 24-60, female to male ratio=0.82) were included in this study. Patients with macular edema, media opacities, refractive error more than ± 6 D, low vision, renal disease, hypertension, coexisting retinal diseases,large retinal non-perfusion around macula, intraocular surgery performed less than six weeks prior to imaging, under intravitreal injection treatment, and who underwent focal laser or Pan-Retinal Photocoagulation (PRP) laser treatment less than six weeks prior to imaging were excluded. Further, OCTA images with significant defects9 (motion artifacts, projection artifacts, vessel doubling and/or stretching defects) were also excluded. For every patient, a detailed medical and ophthalmic history was obtained including the duration of diabetes. Fasting blood sugar after a minimum of 8 hours overnight fasting (FBS in mg/dL), post-prandial blood sugar (PPBS in mg/dL), blood pressure (mm of Hg), body mass index [BMI=weight (in kg)/height2 (in m2)], hemoglobin (Hb in gm/dL), glycosylated hemoglobin (HbA1c as %), low density lipoprotein (LDL in mg/dL) and high density lipoprotein (HDL in mg/dL) was evaluated in each DR patient.
All normal and DR subjects underwent imaging on the Angiovue™ spectral domain OCTA system (Optovue Inc., Fremont, CA) by a single operator using the Angiovue software. The OCT had a high scan speed of 70,000 A-scans per second. Analyses were performed on a scan area of 3 mm × 3 mm generated from the superficial and deep retinal vascular plexus around the fovea for all eyes. In addition to OCTA imaging, all eyes underwent fundus photography to grade the eyes according to the severity of the disease. The DR eyes were classified as mild non-proliferative diabetic retinopathy (NPDR), moderate NPDR, severe NPDR and proliferative diabetic retinopathy (PDR) using the ETDRS classification.10
To analyze the local variations and complexity in the vasculature of the OCTA scans generated from superficial and deep retinal vascular plexus, local fractal analysis was applied.11,12 Box counting method was used to calculate the fractal dimension given by:
In eq. (1), Np was the number of boxes of magnification (p) required to include a structure in the image. A modification was made to box counting method by using a moving window of size (2w+1) × (2w+1) pixels. The moving window was used to calculate the local fractal dimension of each pixel in the OCTA image of the superficial and deep retinal vascular plexus using eq. (2):
In eq. (2), I was the original OCTA image and N was the new image formed after substituting the center pixel of each window by the computed fractal dimension of the window. A window size of 3 × 3 pixels was used to calculate the local fractal dimension as it was observed to be the most accurate for measurement of the FAZ area.12 To assess the FAZ area, the OCTA image was generated by projecting the entire inner retina into one enface angiogram. Then the area of the FAZ was segmented by binarizing the image, followed by boundary detection and connected component labeling (MathWorks Inc., Natick, MA). Using eq. (2), each pixel in the OCTA image was assigned a fractal dimension value. However, the magnitude of fractal dimension of each pixel varied with distribution of the vascular network around the pixel. Thus, a pixel in a larger vessel had a higher fractal dimension than a pixel in a smaller vessel or a non-vessel region. Thus, a normalized fractal dimension ratio was computed for each pixel by taking the ratio of its local fractal dimension with the maximum computed local fractal dimension in the image (N).12
The normalized ratio was represented as a 2-D contour map and represented the probability of presence of a given pixel in the OCTA image within a vessel or a non-vessel region.12 A normalized ratio closer to 1 indicated a vessel and a ratio closer to 0 indicated a non-vessel region. A scoring system was developed based on visual comparison of the normalized ratio map with the OCTA image.12 The pixels within the large vessels had a normalized ratio of 0.9-1.0 while the pixels within the small vessels had a ratio between 0.7-0.9.12 Pixels in regions which were devoid of or had minimal vascular features had a normalized ratio between 0.0-0.3.12 By visual examination, these regions were observed to occur between or around the large vessels. In some regions, these regions were also observed between widely spaced small vessels. In general, these were termed as “spaces between large vessels”. The only exception to this classification was the FAZ, which was common to all the images. Pixels in regions around small vessels, which may be branching out from a large vessel or surrounding small vessels, had a normalized ratio between 0.3-0.7.12 These were termed as “spaces between small vessels” and were generally observed in a cluster of closely packed vessels. The vessel density was computed as percentage by counting all the pixels with a normalized ratio between 0.7-1.0 and then dividing by the total number of pixels in the OCTA image.12 Similarly, the spacing between large vessels and the spacing between the small vessels were expressed as percentage of the total number of pixels in the OCTA image.12 Figure 1 shows the description of the scoring system. Vessel density, spacing between the large vessels and between the small vessels were computed for OCTA images of the superficial and deep retinal vascular plexuses. Figures 2A and 2B show examples of OCTA images of the superficial and deep retinal vascular plexus in moderate NPDR.Figures 2C and 2D show their respective normalized ratio contour maps. Figures 3A to 3D show the examples of OCTA images of the superficial retinal vascular plexus in NPDR (mild, moderate, severe) and PDR eyes. Figures 3E to 3H show their respective normalized ratio contour maps. All the above methods were implemented using MATLAB v7.10 (Mathworks Inc., MA, USA).
Statistical analysis
All analyzed variables were reported as mean ± standard error of the mean. The analyzed variables were FAZ area (mm2), vessel density (%), spacing between large vessels (%) and spacing between small vessels (%). One-way analysis of variance (ANOVA) was performed between the normal eyes and DR grades for each analyzed variable. Also, the analyzed vascular parameters were correlated with all the systemic indicators. Area under the curve, sensitivity and specificity of the vascular parameters to differentiate DR eyes from normal eyes was performed with receiver operating characteristics (ROC) curve. A p-value < 0.05 was considered statistically significant and was Bonferroni-corrected for multiple group comparisons.All statistical analyses were performed in MedCalc v15.8 (MedCalc Inc., Ostend, Belgium)
Results
The number of eyes was 60, 35, 95, 57 and 22 in normal, mild NPDR, moderate NPDR, severe NPDR and PDR groups, respectively. Table 1 shows the systemic characteristics of all the DR subjects included in the study. Duration of DM (p=0.01), HbA1c (p=0.03), FBS (p=0.02) and PPBS (p=0.03) were significantly altered among the grades of DR (Table 1). Table 2 shows the FAZ area, vessel density, spacing between large vessels and spacing between small vessels in the superficial and deep retinal layers between the normal eyes and the DR grades. Normal eyes had a lower FAZ area (p<0.001), higher vessel density (p<0.001), lower spacing between large (p<0.001) and spacing between small vessels (p<0.001) as compared to DR grades (Table 2).
Among the DR grades, vessel density was similar (p>0.05) in both superficial and deep retinal layers (Table 2).Also, the FAZ area did not significantly change (p=0.82) among the DR grades. In the superficial layer and among the DR grades (table 2), PDR (24.1±1.21) and severe NPDR (24.1±0.9) had a significantly higher spacing (p=0.04) between the large vessels as compared to mild (21.1±1.01) and moderate (21.9±0.56). However (Table 2), mild NPDR (39.6±0.56) had a higher spacing between small vessels (p=0.001) as compared to moderate NPDR (37.9±0.21), severe NPDR (37.3±0.44) and PDR (36.8±0.52). Among the DR grades (Table 2), the spacing between the large vessels (p=0.34) and the spacing between the small vessels (p=0.19) were similar in the OCTA scans of the deep retinal vascular plexus.
In the superficial layer, the spacing between the large vessels correlated positively with HbA1c (r=0.25, p=0.03), FBS (r=0.23, p=0.02) and PPBS (r=0.26, p=0.03). Also, the vessel density correlated negatively with HbA1c (r=-0.28, p=0.006), FBS (r=-0.27, p=0.009)and PPBS (r=-0.28, p=0.009). However, vessel density and spacing between the large vessels showed no correlation with other systemic indicators (p>0.05). The FAZ area and spacing between the small vessels showed no correlation with systemic indicators (p>0.05). In the deep layer, the systemic indicators showed no correlation with retinal vascular parameters (p>0.05). Table 3 summarizes the results of the ROC analyses. Area of the FAZ (Table 3) had the lowest sensitivity and specificity among all the vascular parameters in superficial and deep retinal vascular plexus. Vessel density had relatively higher sensitivity (94.7%) and specificity (85.5%), particularly when the deep retinal plexus was used as a differentiator (Table 3). Interestingly, spacing between large vessels achieved the highest sensitivity and specificity in both superficial (92.7% and 92.9%, respectively) and deep (96% and 92.7%, respectively) retinal vascular plexus. The area under the ROC curve of spacing between the large vessels also was significantly better than the same for vessel density in both superficial (p=0.001) and deep retinal plexus (p=0.01).
Discussion
OCTA can demonstrate clinically relevant changes in the retinal vasculature of DR patients qualitatively such as distortion and enlargement of foveal avascular zone, retinal capillary dropouts and micro-aneurysms comparable to FFA.13-15 In this study, local fractal based method was used to quantify the retinal vascular parameters in the OCTA images of superficial and deep retinal layers and were correlated with systemic indicators to explore the possibility of using these vascular parameters as metrics to gauge severity of DR. In addition to vessel density, the study defined “spaces between large vessels” and “spaces between small vessels”.12
The superficial and deep retinal vascular plexuses merged at the edge of FAZ in the OCTA images (Figure 2). Thus, segregating the FAZ into the superficial and deep could be anatomically misleading. So, the FAZ was assessed by projecting the entire inner retina into one enface angiogram. The terms “large vessel” and “small vessel” were specific to the distribution of normalized local fractal dimension values in a given layer (superficial and deep). Normalization was performed by dividing the local fractal values of a layer by the maximum local fractal value of the same layer. Thus, the absolute value of local fractal dimension was specific to each layer but the normalized value was automatically constrained between 0 and 1 by definition. By visual comparison of the OCTA image with the normalized local fractal dimension values, a range was defined for classification of vessels into large and small. Further, no distinction was made between arterioles and venules.12,30In this study, the analysis was based on normalized values only. Dimensionality of large vessels in the superficial layer was different from the dimensionality of large vessels and/or capillaries in the deep layer.
Similarly, a range was also defined for spacing between the large vessels and spacing between the small vessels.12 These were simply treated as a capillary free zone in normal eyes. Since they were free from capillaries, it directly implied that there was no perfusion there. If normal eyes were considered as baseline, then the increase in spacing in disease eyes as compared to normal eyes indicated loss of capillaries. This in turn indicated a decrease in capillary perfusion.A previous study using the same technique and definition of ranges for normalized local fractal dimension was validated with quantitative numbers of vessel density in living animal eyes.12 However, new insights into macular capillary changes may result in modification of the defined range in future.22,30
Normal eyes had a significantly lower FAZ area as compared to DR patients in both superficial and deep retinal layers. These findings were consistent with previous studies on DR using OCTA and FFA.8,16-18 However, the FAZ area was similar in both superficial and deep retinal plexus among the DR grades. Also, DR eyes had lower vessel density in both superficial and deep retinal layers as compared to normal eyes. Similar findings were reported in a recent study which showed lower capillary density in DR eyes as compared to normal eyes.19 The vessel density across the DR grades was similar (p>0.05). However, PDR and severe NPDR had a higher spacing between the large vessels as compared to mild and moderate NPDR. Moreover, mild NPDR had higher spacing between the small vessels as compared to moderate NPDR, severe NPDR and PDR. The perifoveal inter-capillary area was increased in DR as compared to normal eyes.18 Also in previous studies, DR eyes showed changes in the retinal vasculature and thereby an increase in retinal non-perfusion areas.20 Histological studies on diabetic retina have showed localized regions of non-perfusion and progressive loss of retinal capillary cells.21 These give a possible explanation of the progressive increase in the spacing between the large vessels with increase in the severity of DR.
With progression, a few smaller capillary free zones may become enlarged due to vessel dropout and may get converted to large vacant spaces as a result. From Table 2, decrease in spacing between small vessels was accompanied by a corresponding increase in spacing between large vessels with DR severity. This may explain why mild NPDR had higher spacing between small vessels than the more severe grades.In a recent paper, similar increase was observed in total avascular area in disease eyes and was termed as “capillary non-perfusion”. 22 The ROC analyses confirmed that spacing between large vessels (maximum area under the curve=0.99±0.01 in deep retinal vascular plexus; Table 3) could be a better indicator of adverse vascular changes in DR and changes in the vasculature following treatment instead of just vessel density (maximum area under the ROC curve=0.94±0.02 in deep retinal vascular plexus in Table 3).
HbA1c is used for the diagnosis of DM and for gauging progression of DR.23-25 In a study on relationship of HbA1c with risk of development and progression of retinopathy, HbA1c was a major predictor of progression of retinopathy.25 In this study, spacing between large vessels and vessel density in the superficial layer showed a significant correlation with HbA1c levels. HbA1c levels were elevated (> 9%) in severe NPDR and PDR as compared to mild and moderate NPDR (Table 1). Diabetic patients with elevated blood glucose levels had a greater severity of DR than those with lower glucose levels.26 In this study, spacing between the large vessels and vessel density correlated significantly with FBS and PPBS as well. FBS and PPBS levels were higher in severe NPDR and PDR as compared to mild and moderate NPDR (Table 1). Previous studies showed a positive correlation among HbA1c, FBS and PPBS, thereby indicating that individuals with lower HbA1c levels have a lower chance of developing DR.26, 27_ENREF_22 In this study, systemic data was included to assess their correlations with the OCTA parameters. Every ophthalmologist may not have access to OCTA but they do have access to the systemic data of DR patients. Therefore, we assessed whether the levels of systemic DM markers were indicative of the magnitudes of OCTA parameters such as vessel density. From the analyses, systemic data correlated with the OCTA parameters but the correlation was not strong enough (low r value) for use as a predictive tool. Although the correlationswere weak, longitudinal assessment of progression of disease in the same eye may lead to stronger correlations between these parameters. The presentstudy can possibly serve as a baseline for such progression studies in future.
The study had a few limitations. In this study, normal subjects (mean age: 38.7 ± 1.68 years; range: 20-60 years) were significantly younger (p<0.001) than DR patients. However, the FAZ area and OCTA parameters in normal eyes were unaffected by age (p>0.05). A recent OCTA study on normal Indian eyes (age: 20-67 years) using local fractal based method demonstrated that vessel density and FAZ did not change significantly with age.12Based on these observations, age difference between normal and DR eyes was not considered in the analyses. However, other studies have showed a significant decrease in vessel density with age in Chinese and Caucasian eyes.28,29Therefore, a larger cohort study may be needed to reassess these correlations in Indian eyes. A recent study showedthat the vessels in the superficial plexus superimposed on the deep plexus in 68% of the cases.30 A study on image artifacts in OCTA described this artifact as “projection artifact”.9In this study, images of the deep retinal vascular plexus with significant projections of the large vessels from the superficial layer were excluded. However, it was difficult to completely avoid projection artifacts and these might still have some confounding effects on the analyses.Some recent techniques aim to reduce these artifacts, which can only help in improving the analyses with local fractal dimension in the near future.31Despite the upcoming techniques used to compensate for axial eye motions, transverse motions from fixation changes remain a major cause of artifacts in OCTA.9Additionally, the OCTA software can introduce artifacts such as loss of detail, doubling of vessels, stretching defects and false flow artifact.9 In this study, OCTA images with significant motion artifacts or stretching defects or doubling of vessels were excluded from this study. However, other artifacts which were not apparent could have some confounding effect on the results of this study.
OCTA is proving to be a very useful modality in evaluating vascular changes in healthy as well diseased eyes.32, 33 However, a lot of improvement and understanding is still needed to correctly interpret the data and its clinical significance.9Thus, spacing between the large vessels, which correlated positively with HbA1c, FBS and PPBS, may be a sensitive marker to quantify the progression of DR. This may be used in combination with HbA1c, FBS and PPBS for identifying patients who are at high risk of developing DR early and for grading of DR.Further studies with long term follow-up are needed to confirm these findings.
Acknowledgements
The present imaging study was done as part of a collaborative characterization of a large cohort of diabetic retinopathy patients with Genentech, San Francisco, CA, USA.
References
1.Wild S, Roglic G, Green A, Sicree R, King H. Global prevalence of diabetes: estimates for the year 2000 and projections for 2030. Diabetes care 2004;27:1047-1053.
2.Klein R, Klein BE, Moss SE, Cruickshanks KJ. The Wisconsin Epidemiologic Study of diabetic retinopathy. XIV. Ten-year incidence and progression of diabetic retinopathy. Arch Ophthalmol 1994;112:1217-1228.
3.Rema M, Pradeepa R. Diabetic retinopathy: an Indian perspective. Indian J Med Res 2007;125:297-310.
4.Spaide RF, Klancnik JM, Jr., Cooney MJ. Retinal vascular layers imaged by fluorescein angiography and optical coherence tomography angiography. JAMA Ophthalmol 2015;133:45-50.
5.Talisa E de Carlo AR, Nadia K Waheed,Jay S Duker. A review of optical coherence tomography angiography (OCTA). IntJ of Retina and Vitreous 2015.
6.Couturier A, Mane V, Bonnin S, et al. Capillary Plexus Anomalies in Diabetic Retinopathy on Optical Coherence Tomography Angiography. Retina 2015;35:2384-2391.
7.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 LasersImagingRetina 2015;46:796-805.
8.Takase N, Nozaki M, Kato A, Ozeki H, Yoshida M, Ogura Y. Enlargement of Foveal Avascular Zone in Diabetic Eyes Evaluated by En Face Optical Coherence Tomography Angiography. Retina 2015;35:2377-2383.
9.Spaide RF, Fujimoto JG, Waheed NK. Image Artifacts in Optical Coherence Tomography Angiography. Retina 2015;35:2163-2180.
10.Early Treatment Diabetic Retinopathy Study Research Group. Grading diabetic retinopathy from stereoscopic color fundus photographs–anextension of the modified Airlie House classification: ETDRS report number 10. Ophthalmology 1991;98:786-806.
11.Taud H, Parrot J-F. Measurement of DEM roughness using the local fractal dimension. Géomorphologie 2005;10:11.
12.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;57:246-252.
13.Freiberg FJ, Pfau M, Wons J, Wirth MA, Becker MD, Michels S. Optical coherence tomography angiography of the foveal avascular zone in diabetic retinopathy. Graefes Arch Clin Exp Ophthalmol 2015.
14.Hwang TS, Jia Y, Gao SS, et al. Optical Coherence Tomography Angiography Features of Diabetic Retinopathy. Retina 2015;35:2371-2376.
15.Ishibazawa A, Nagaoka T, Takahashi A, et al. Optical Coherence Tomography Angiography in Diabetic Retinopathy: A Prospective Pilot Study. Am J Ophthalmol 2015;160:35-44 e31.
16.Di G, Weihong Y, Xiao Z, et al. A morphological study of the foveal avascular zone in patients with diabetes mellitus using optical coherence tomography angiography. Graefes Arch Clin Exp Ophthalmol 2015.
17.de Carlo TE, Chin AT, Bonini Filho MA, et al. Detection of Microvascular Changes in Eyes of Patients with Diabetes but Not Clinical Diabetic Retinopathy Using Optical Coherence Tomography Angiography. Retina 2015;35:2364-2370.
18.Conrath J, Giorgi R, Raccah D, Ridings B. Foveal avascular zone in diabetic retinopathy: quantitative vs qualitative assessment. Eye 2005;19:322-326.
19.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.
20.Hammes HP, Feng Y, Pfister F, Brownlee M. Diabetic retinopathy: targeting vasoregression. Diabetes 2011;60:9-16.
21.Cai J, Boulton M. The pathogenesis of diabetic retinopathy: old concepts and new questions. Eye 2002;16:242-260.
22.Hwang TS, Gao SS, Liu L, et al. Automated Quantification of Capillary Nonperfusion Using Optical Coherence Tomography Angiography in Diabetic Retinopathy. JAMA ophthalmology 2016;1-7.
23.Krishnamurti U, Steffes MW. Glycohemoglobin: a primary predictor of the development or reversal of complications of diabetes mellitus. Clinical chemistry 2001;47:1157-1165.
24.Cho NH, Kim TH, Woo SJ, et al. Optimal HbA1c cutoff for detecting diabetic retinopathy. Acta Diabetol 2013;50:837-842.
25.The relationship of glycemic exposure (HbA1c) to the risk of development and progression of retinopathy in the diabetes control and complications trial. Diabetes 1995;44:968-983.
26.Frank RN. Diabetic retinopathy and systemic factors. Middle East Afr J Ophthalmol 2015;22:151-156.
27.Ketema EB, Kibret KT. Correlation of fasting and postprandial plasma glucose with HbA1c in assessing glycemic control; systematic review and meta-analysis. Arch Public Health 2015;73:43.
28.Shahlaee A, Samara WA, Hsu J, et al. In Vivo Assessment of Macular Vascular Density in Healthy Human Eyes Using Optical Coherence Tomography Angiography. Am J Ophthalmol 2016.
29.Yu J, Jiang C, Wang X, et al. Macular perfusion in healthy Chinese: an optical coherence tomography angiogram study. Invest Ophthalmol Vis Sci 2015;56:3212-3217.
30.Bonnin S, Mane V, Couturier A, et al. New Insight into the Macular Deep Vascular Plexus Imaged by Optical Coherence Tomography Angiography. Retina 2015;35:2347-2352.
31.Zhang M, Hwang TS, Campbell JP, et al. Projection-resolved optical coherence tomographic angiography. Biomed Opt Express 2016;7:816-828.
32.Mastropasqua R, Di Antonio L, Di Staso S, et al. Optical Coherence Tomography Angiography in Retinal Vascular Diseases and Choroidal Neovascularization. J Ophthalmol 2015;2015:343515.
33.Jia Y, Morrison JC, Tokayer J, et al. Quantitative OCT angiography of optic nerve head blood flow. Biomed Opt Express 2012;3:3127-3137.