Sickle cell disease (SCD) is a condition of genetic etiology characterized by the presence of hemoglobin S.1,2 This pathological variant has the glutamate amino acid found at position 6 of the β chain replaced by hydrophilic valine.1,2 Although the heterozygous genotype confers at least some protection against malaria, homozygous S and subsequent pathogenic alleles in combination with S, such as C or β-thalassaemia, produce a systemic disorder that affects the cardiovascular, nervous, pulmonary, renal, musculoskeletal, and ocular systems.1,2
SCD is characterized by vaso-occulsion of a vascular bed when the abnormal erythrocytes are placed under physiological stress such as acidosis or hypoxia.3 In the eye, sickle cell retinopathy (SCR) is secondary to occlusion of the peripheral retina, typically in the temporal region, as a result of the higher oxygen demand.4 Ischemia resulting from vessel occlusion leads to upregulation of angiogenic factors, most notably vascular endothelial growth factor (VEGF).5 This promotes aberrant neovascularization, causing vision loss through tractional retinal detachment or vitreous hemorrhage.6
Current treatment modalities for SCR manifestations, including capillary drop-out and salmon patches, primarily comprise anti-VEGF therapy and scatter photocoagulation.7 The latter is guided by fluorescein angiography (FA).8 Areas of vascular leakage suggestive of neovascularization are visualized as hyperintense regions, whereas areas of nonperfusion are hypointense. At present, standard FA used in assessing SCR captures the retina at a field angle between 20° and 50°.8 Given the unique propensity of SCR to affect the retinal periphery, a wider field angle may be instrumental in assessing the degree and extent of SCR. In a case series of 12 eyes from six sickle cell patients, Cho and Kiss found that widefield angiography identified active peripheral neovascularization in six eyes, whereas conventional standard-field angiography and physical exam captured neovascularization in one eye.9 Herein, we seek to characterize, quantify, and compare the areas of vascular leakage in a sickle cell cohort using ImageJ software (NIH, Bethesda, MD) in both standard-field and widefield FA.
Patients and Methods
Study Design and Eligibility
This was a retrospective case series. Patient charts and images from January 2005 to January 2018 were accessed from the Kresge Eye Institute, Detroit, Michigan, electronic records if they had been treated for SCR with photocoagulation. In all instances, scatter laser photocoagulation had been delivered to the avascular retina with indirect power titrated to a gray-white burn with a distance between the spots of one-half to one laser burn units apart. Ethics approval was obtained from the Wayne State University Intuitional Review Board (REB# 061817MP2X). Research adhered to the tenets of the Declaration of Helsinki and Good Clinical Practice.
Those patients who had a pre-treatment and post-treatment FA either using conventional field dimensions or widefield scans and were 18 years or older at the time of treatment were included in this study. Patients were excluded from the study if they had received treatment for SCR outside of the study window, if they did not have a post-treatment FA, or if they had a subsequent retinal condition that would compromise FA evaluation such as diabetic retinopathy, a prior retinal detachment, retinal artery occlusion, or retinal vein occlusion. In instances where a patient had bilateral SCR and met the inclusion and exclusion criteria, only the eye with the highest Goldberg classification that received treatment was included in study analysis. Conventional FA (CFA) (TRC-50DX; Topcon, Tokyo, Japan) scans had a field angle of 35°, whereas ultra-wide-spectrum FA (UWF-FA) (Panoramic 200A; Optos, Dunfermline, United Kingdom) scans had a field angle of 200°.
Data Collection and Image Analysis
Age, sex, treated eye, treatment date, visual acuity (VA) before treatment, and VA after treatment were extracted from patient charts. Lesion area and intensity of late arteriovenous phase images on UWF-FA or CFA scans were analyzed using ImageJ software Fiji plugin. ImageJ is an open source image-processing package that has been previously applied to ophthalmic imaging.10 The software is publicly available for download at the National Institute of Health's website ( https://imagej.nih.gov/ij). After opening an image, ImageJ users may manually map a given area using the trace feature and quantify both the area and intensity within the tracing by selecting the “Analyze Histogram” feature of the program. Here, counts, averages, and traditional measures of dispersion for area and intensity are computed and tabulated by the software. With ImageJ, area is calculated by pixel number, which can be assigned a unit in relation to a known area in the image. Intensity is calculated per pixel on a ratio scale ranging from a theoretical value of 0 to 255 units. To calculate lesion intensity in this study, all pixel intensity scores for a given image were added together and then divided by the number of lesion pixels. The same calculation was applied to the optic disc in order to derive the disc intensity measure. Lesion intensity was then expressed as the mean lesion intensity minus the mean disc intensity measure.
Lesion areas and, hence, the derived intensity scores were manually mapped out by two graders using the trace feature on ImageJ (Figure 1). Grader 2 assessed all images twice within a 72-hour washout period between reads. In addition to lesion area and intensity, Grader 1 assessed disc area and intensity for all images using the trace feature. Graders were blinded to one another's reads and their first read in the case of Grader 2. After manually determining lesion area and intensity, scores were standardized against the optic disc area and optic disc average intensity for each image to account for differences in exposure, media opacity, and distortions related to field angle.
Before and after ultra-wide-spectrum fluorescein angiography (UWF-FA) area assessment. (Top) Late arteriovenous UWF-FA scan (image on the left) and ImageJ area assessment by Grader 1 (images on the right) pre-treatment. (Bottom) Late arteriovenous UWF-FA scan (image on the left) and ImageJ area assessment by Grader 1 (images on the right) post-treatment.
Patient demographic data were summarized as means and standard deviations for continuous variables or proportions and percentages for categorical variables. Dependent t-tests were used to assess the change in area and intensity pre- to post-treatment upon standardizing the measures against the disc scores. Post-hoc, an independent t-test was used to assess pre- to post-area change between wide- and standard-field FAs. Intra- and interrater reliability for area and intensity measures using CFA and UWF-FA scans individually were calculated using intraclass correlations (ICC; 2, 1) with 95% confidence intervals (CIs). Here, graders are assumed to be selected at random from a population of similar graders. All statistical tests throughout analysis were two-sided, and a P value of less than .05 was considered significant. As this study was exploratory in nature, we did not adjust for multiple testing. Analysis was performed on SPSS software (version 22.0; IBM, Armonk, NY).
Eleven eyes from 11 patients met the study inclusion and exclusion criteria. Of these, five patients (45.5%) were female, and the mean age of participants at the time of treatment was 46.5 years ± 16.3 years. Subsequent study participants' demographic information is present in Table 1. Pre- and post-treatment disc areas were measured at 143 × 101 ± 111 × 101 pixels and 132 × 101 ± 103 × 101 pixels, respectively. The total disc intensity was 229 × 103 ± 197 × 103 units pre-treatment and 170 × 103 ± 140 × 103 units post-treatment, respectively. Grader 1 measured pre- and post-treatment lesion area to be 6.24 ± 5.55 disc areas (DA) and 2.57 ± 2.62 DA, respectively, whereas grader 2 measured the pre-treatment area to be 7.10 ± 5.88 DA and the post-treatment area to be 3.73 ± 3.59 DA during the first read. Grader 2′s second read lesion areas were 7.12 ± 6.03 DAs pre-treatment and 3.77 ± 3.60 DAs post-treatment. The average lesion intensity relative to disc intensity pre-treatment was measured as 33.2 ± 54.4 units by Grader 1 and 17.8 ± 58.2 units by Grader 2 during the first read, and 4.1 ± 57.8 units by Grader 2 during the second read. Post-treatment, the average intensity relative to the optic disc was recorded as 10.2 ± 88.0 units by Grader 1, −5.7 ± 80.9 units by Grader 2 during the first read, and −7.3 ± 83.7 units by Grader 2 during the second read. Since the intensity was recorded relative to total disc intensity, a negative mean intensity value means that the average lesion intensity was less than that of the disc intensity. Subsequent lesion parameters are documented in Table 2.
Intensity and Area Measures
For read 1, lesion area decreased by 3.39 ± 3.56 DAs pre- to post-treatment (P < .001). The decrease in lesion area was observed for both UWF-FA images, which decreased by 3.83 ± 3.65 DAs (P = .003) and CFA which decreased by 2.81 ± 3.55 DAs (P = .034). For UWF-FA images, pre- to post-lesion intensity standardized against disc intensity declined −23.3 ± 72.3 units (P = .154). For CFA images the pre- to post-standardized intensity change was −14.0 ± 58.6 units (P = .468). The difference in area decline pre- to post-treatment between wide- and standard-field FAs was not significant (P = .652). When imaging fields were combined, pre- to post-lesion intensity standardized against disc intensity declined 19.3 ± 65.4 units (P = .104). The inter-rater reliability for change in lesion area measured as a proportion of DA was 0.88 (95% CI, 0.61–0.97; P < .001), whereas the inter-rater reliability for change in intensity relative to disc intensity was 0.90 (95% CI, 0.67–0.97; P < .001). The intra-rater reliability for change in lesion area measured as a proportion of DA was 0.97 (95% CI, 0.89–0.99 (P < .001). The intra-rater reliability for change in intensity pre- to post-treatment relative to disc intensity was 0.57 (95% CI, 0.00–0.86; P < .001).
The study at hand sought to evaluate changes in neovascular surface area and intensity in a SCR cohort pre- to post-scatter photocoagulation treatment for both UWF-FA and CFA images using ImageJ software. Both lesion intensity and area decreased pre- to post-treatment in UWF-FA and CFA images. However, only the change in lesion area was significant, ranging from −3.34 ± 2.43 DAs to −3.66 ± 3.72 (P < .001) depending on the grader and the read. Neovascular lesion area declined by 3.83 ± 3.65 DAs (P = .003) for UWF-FA and 2.81 ± 3.55 DAs (P = .034) for CFA scans. Inter-rater reliability was good (0.75 to 0.90)11 for both lesion area and intensity at 0.88 (95% CI, 0.61–0.97) and 0.90 (95% CI, 0.67–0.97), respectively. Intra-rater reliability was excellent (P > .90)11 for area measures at 0.97 (95% CI, 0.89–0.99) but only moderate (between 0.50 and 0.75)11 for intensity measures at 0.57 (95% CI, 0.00–0.86).
Both the relatively lower intra-rater reliability of intensity measures compared to area measures and nonsignificance of pre- to post-treatment intensity measures may in part be explained by the intensity measure gradient. ImageJ captures intensity on a scale of 0 to 255 arbitrary units. This contrasts to area measures that may be calculated to the pixel. Indeed the average lesion area ranged from 847 × 101 ± 128 × 102 to 109 × 102 ± 178 × 102 pixels based on grader and read, meaning the sensitivity of area measures was several orders of magnitude greater than intensity measures. Moreover, lesion intensity has a steep gradient in the late arteriovenous phase around the lesion margins, where intensity can drop off from 200 or more arbitrary units to 50 or less arbitrary units significantly skewing the average lesion intensity measure. Area measures do not have a steep marginal gradient, as the addition of marginal area by a grader adds, at most, a few pixels to a measure that is several thousand pixels large. Thus, intensity measures are more susceptible to error resulting from grader read variability than area measures. This problem was compounded by low sample size, which likewise makes average measures more susceptible to outliers.
In 1971, Goldberg published a SCR classification for proliferative disease using fundoscopic examination.12 The Goldberg criterion have remained the gold standard for monitoring SCR disease extent and progress for nearly 50 years despite questionable inter-rater reliability and sensitivity to change.13 The advent of optical coherence tomography (OCT) has brought with it the possibility of using the novel modality to capture SCR changes. Chow et al. first reported OCT changes including macular thinning, foveal thinning, and temporal splaying in a SCD cohort.14 Subsequent work by Mathew et al. in 208 eyes of 107 patients with SCD has found that temporal inner and outer subfields from the Early Treatment Diabetic Retinopathy Study grid were significantly thinner when compared to age-matched controls.15 Although these findings hold promise for using OCT in the setting of SCR, the vision-threatening changes that evolve with disease progression are generally in the peripheral retina. A quantitative approach to analyzing peripheral FA findings would help to determine response to treatment and predict future risk for bleeding. ImageJ may be applied to either widefield or standard-field FA images to monitor disease progression by capturing lesion area in a sensitive and reliable manner. Collectively, these findings suggest that lesion area may be a sensitive measure of SCR disease extent and progression in FA scans.
- Serjeant GR. The emerging understanding of sickle cell disease. Br J Haematol. 2001;112(1):3–18. doi:10.1046/j.1365-2141.2001.02557.x [CrossRef] PMID:11167776
- Ballas SK, Lieff S, Benjamin LJ, et al. Investigators, Comprehensive Sickle Cell Centers. Definitions of the phenotypic manifestations of sickle cell disease. Am J Hematol. 2010;85(1):6–13. PMID:19902523
- Bowling B, ed. Kanski's Clinical Ophthalmology: A Systematic Approach. 8th ed. London: Elsevier; 2016.
- Emerson GG, Lutty GA. Effects of sickle cell disease on the eye: clinical features and treatment. Hematol Oncol Clin North Am. 2005;19(5):957–973, ix. doi:10.1016/j.hoc.2005.07.005 [CrossRef] PMID:16214655
- Kim SY, Mocanu C, Mcleod DS, et al. Expression of pigment epithelium-derived factor (PEDF) and vascular endothelial growth factor (VEGF) in sickle cell retina and choroid. Exp Eye Res. 2003;77(4):433–445. doi:10.1016/S0014-4835(03)00174-X [CrossRef] PMID:12957143
- Brazier DJ, Gregor ZJ, Blach RK, Porter JB, Huehns ER. Retinal detachment in patients with proliferative sickle cell retinopathy. Trans Ophthalmol Soc U K. 1986;105(Pt 1):100–105. PMID:3459293
- Siqueira RC, Costa RA, Scott IU, Cintra LP, Jorge R. Intravitreal bevacizumab (Avastin) injection associated with regression of retinal neovascularization caused by sickle cell retinopathy. Acta Ophthalmol Scand. 2006;84(6):834–835. doi:10.1111/j.1600-0420.2006.00779.x [CrossRef] PMID:17083555
- Rabiolo A, Parravano M, Querques L, et al. Ultra-wide-field fluorescein angiography in diabetic retinopathy: a narrative review. Clin Ophthalmol. 2017;11:803–807. doi:10.2147/OPTH.S133637 [CrossRef] PMID:28490862
- Cho M, Kiss S. Detection and monitoring of sickle cell retinopathy using ultra wide-field color photography and fluorescein angiography. Retina. 2011;31(4):738–747. doi:10.1097/IAE.0b013e3181f049ec [CrossRef] PMID:21836403
- Schindelin J, Rueden CT, Hiner MC, Eliceiri KW. The ImageJ ecosystem: an open platform for biomedical image analysis. Mol Reprod Dev. 2015;82(7–8):518–529. doi:10.1002/mrd.22489 [CrossRef] PMID:26153368
- Koo TK, Li MY. A Guideline of Selecting and Reporting Intraclass Correlation Coefficients for Reliability Research. J Chiropr Med. 2016;15(2):155–163. doi:10.1016/j.jcm.2016.02.012 [CrossRef] PMID:27330520
- Goldberg MF. Classification and pathogenesis of proliferative sickle retinopathy. Am J Ophthalmol. 1971;71(3):649–665. doi:10.1016/0002-9394(71)90429-6 [CrossRef] PMID:5546311
- Han IC, Zhang AY, Liu TYA, et al. Utility of ultra-widefield retinal imaging for the staging and management of sickle cell retinopathy. Retina. 2019;39(5):836–843. doi:10.1097/IAE.0000000000002057 [CrossRef] PMID:29384996
- Chow CC, Genead MA, Anastasakis A, Chau FY, Fishman GA, Lim JI. Structural and functional correlation in sickle cell retinopathy using spectral-domain optical coherence tomography and scanning laser ophthalmoscope microperimetry. Am J Ophthalmol. 2011;152(4):704–711.e2. doi:10.1016/j.ajo.2011.03.035 [CrossRef] PMID:21726848
- Mathew R, Bafiq R, Ramu J, et al. Spectral domain optical coherence tomography in patients with sickle cell disease. Br J Ophthalmol. 2015;99(7):967–972. doi:10.1136/bjophthalmol-2014-305532 [CrossRef] PMID:25595176
| Male||6 (54.5%)|
| Female||5 (45.5%)|
| OD||6 (54.5%)|
| OS||5 (45.5%)|
| UWF-FA||6 (54.5%)|
| CFA||5 (45.5%)|
|Sickle Cell Genotype|
| SC||8 (72.7%)|
| Other||3 (27.3%)|
|Goldberg Classification at Baseline|
| I||0 (0.0%)|
| II||0 (0.0%)|
| III||9 (81.8%)|
| IV||2 (18.2%)|
| V||0 (0.0%)|
|Variable||Mean ± SD|
|Age||46.5 ± 16.3 years|
|Baseline VA, logMAR (Snellen)||0.446 (20/55.9) ± 0.633 (20/85.9)|
|Post-treatment VA, logMAR (Snellen)||0.412 (20/51.6) ± 0.532 (20/68.1)|
Intensity and Area Measures
|Variable||Reader 1, Read 1 (Mean ± SD)||Reader 2, Read 1 (Mean ± SD)||Reader 2, Read 2 (Mean ± SD)|
|Pre-treatment lesion area*||847 × 101 ± 128 × 102||109 × 102 ± 178 × 102||108 × 102 ± 182 × 102|
|Pre-treatment lesion intensityʈ||170 × 104 ± 269 × 104||192 × 104 ± 328 × 104||168 × 104 ± 330 × 104|
|Post-treatment lesion area*||369 × 101 ± 458 × 101||661 × 101 ± 953 × 101||680 × 101 ± 960 × 101|
|Post-treatment lesion intensityʈ||560 × 103 ± 672 × 103||837 × 103 ± 113 × 104||843 × 103 ± 112 ×104|
|Pre-treatment lesion area by DA||6.24 ± 5.55||7.10 ± 5.88||7.12 ± 6.03|
|Pre-treatment lesion intensity minus DI||33.2 ± 54.4||17.8 ± 58.2||4.1 ± 57.8|
|Post-treatment lesion area by DA||2.57 ± 2.62||3.73 ± 3.59||3.77 ± 3.60|
|Post-treatment lesion intensity minus DI||10.2 ± 88.0||−5.7 ± 80.9||−7.3 ± 83.7|
|Change in lesion area||−3.66 ± 3.72||−3.37 ± 2.29||−3.34 ± 2.43|
|Change in lesion intensity||−23.0 ± 69.3||−23.5 ± 22.7||−11.4 ± 25.9|