Abstract
Purpose
To evaluate the relationship between retinal haemorrhages detected on Ultra-widefield (UWF) red channel images and perfusion status in eyes with acute central retinal vein occlusion (CRVO).
Methods
UWF fundus images were split into green and red channels using ImageJ software. The retinal haemorrhages were calculated quantitatively with both the green and red channel images, resulting in green channel haemorrhages (GCH) and red channel haemorrhages (RCH). The nonperfusion area (NPA) was also calculated from fluorescein angiography in each eye. The relationships between both the GCH and RCH with the NPA were investigated.
Results
Thirty-two eyes of 32 patients with acute CRVO (18 men, 14 women) were included. The mean GCH and RCH values were 10.4% ± 8.2% and 1.7% ± 1.7%, respectively. The mean NPA was 39.2% ± 28.8%. Significant correlations were seen between the GCH and NPA (r = 0.38; P = 0.022) and RCH and NPA (r = 0.44; P = 0.010, linear regression analysis). Multivariate analysis suggested that only the RCHs were correlated significantly with the NPA.
Conclusions
Retinal haemorrhages detected by UWF red channel imaging were less compared to green channel imaging and associated closely with retinal NPAs in eyes with acute CRVO.
UWF red channel imaging allowed us to identify ischaemia-related haemorrhage.
Subject terms: Retinal diseases, Vision disorders
Introduction
Central retinal vein occlusion (CRVO) is one of the most common causes of severe visual impairment and blindness [1–4]. Because of CRVO, venous tortuosity and dilation and retinal haemorrhages develop. Furthermore, this circulatory disturbance causes development of retinal capillary nonperfusion areas (NPAs) leading to retinal ischaemia. Fluorescein angiography (FA) is the gold standard for determining the retinal perfusion status. However, FA requires an intravenous injection of fluorescein dye that sometimes causes complications. Therefore, a noninvasive test to identify NPAs has been required. Recent studies to detect NPA using optical coherence tomography (OCT) angiography and en-face OCT images have been reported [5, 6]. However, they have the disadvantages that the angle of view is narrow and the images are too rough to evaluate especially in acute CRVO.
The relationship between retinal haemorrhages and retinal perfusion status in RVO has been reported [7]. Previous OCT studies have shown that development of retinal ischaemia depends on the shape of the haemorrhage or the layer in which the haemorrhage is located [8–10].
Recently, ultra-widefield (UWF) imaging has begun to be used widely to diagnose retinal diseases in clinical settings [11–13]. UWF fundus images are advantageous for screening because they can be obtained without contact and dilation. In contrast to other fundus colour images, the Optos system (Optos Panoramic 200, Marlborough, MA; Optos PLC, Dunfermline, Scotland, UK) acquires pseudocolor images by combining only red and green scanning lasers. The green laser image depicts the retina and its vasculature, and the red laser image highlights deeper structures, such as the retinal pigment epithelium (RPE) and choroid. The use of two different laser imaging systems might facilitate the evaluation of haemorrhages in the different retinal layers and properties. We recently reported that UWF red channel imaging allowed discrimination of ischaemic CRVO [14]. However, that study was qualitative. Therefore, the current study sought to quantitatively evaluate the relationship between retinal haemorrhage in green/red channels and NPA in eyes with acute CRVO.
Methods
Subjects
The institutional review board of the Yokohama City University School of Medicine, Kanagawa, Japan, approved the study protocol, which adhered to the tenets of the Declaration of Helsinki. The participants or their legal guardians provided informed consent.
The current study was a retrospective study of the clinical records of eyes diagnosed with acute CRVO from April 2016 to January 2019 in Yokohama City University Medical Centre. The diagnosis of acute CRVO was based on the characteristic fundus appearance, i.e., symptomatic CRVO with retinal haemorrhages affecting all four retinal quadrants and symptom duration less than 3 months before examination. Eyes were exclude that had grades 1 and 2 haemorrhages (no/small/isolated haemorrhages) graded by Hayreh and Zimmerman [7] and in the presence of laser photocoagulation scars, retinal artery occlusion, diabetic retinopathy, or hypertensive retinopathy. The cases in which the fundus was not clearly examined due to opacity of cornea, lens or vitreous were also excluded.
Examinations
All patients underwent comprehensive ophthalmologic examinations, including visual acuity (VA) measurement, slit-lamp biomicroscopy, and UWF colour fundus photography and fluorescein angiography (FA) using the Optos system. The VA was measured in decimal units and converted to the logarithm of the minimum angle of resolution (logMAR) units for statistical analyses.
Quantifying retinal haemorrhages and NPAs
For each case, the UWF colour image and UWF FA image were evaluated. For evaluating the retinal haemorrhages, the UWF colour image was split into grayscale images of blue, green, red channel images using “split colour channel” in ImageJ software (ImageJ, V.2.0.0-rc-69/1.52i, NIH, Bethesda, MD) (Fig. 1A). Since the UWF colour image consisted of green and red wave laser image, nothing was reflected in blue channel image. In determining the area of green channel haemorrhage and red channel haemorrhage, we followed the method used in the past to determine the area of non-perfusion, “wand tool” or “freehand selections” in Image J (Fig. 1B, A, B) [15, 16]. The haemorrhage region was determined by two researchers (ST, YT), and the agreement rate was checked.
Fig. 1. Methods for quantification of green and red channel retinal haemorrhages.
1A. The methods of taking out the green and red laser wave image. A UWF colour image (A) was split into grayscale images of blue (B), green (C), red (D) channel images using ImageJ software. Since the UWF image consisted of green and red wave laser image, nothing was reflected in blue channel image. 1B Quantification of retinal haemorrhages and NPA. A The green channel image focuses on the retina and vasculature. Retinal haemorrhage is marked with black in the green channel image. An area in which the image could be evaluated clearly was selected and applied to other images of the same patient (region of interest (ROI), white line). The haemorrhage pixels within the ROI are 136,730 and pixels of ROI are 3,033,264. The GCH was 4.5% because the green channel haemorrhage ratio (GCH) was defined as the pixels of the haemorrhages divided by the number of pixels in ROI. B The red channel image focuses on the choroid. Retinal haemorrhage is marked with black in the red channel image. The ratio of red channel haemorrhage (RCH) was also calculated, resulting in 1.4%. C An image from UWF fluorescein angiograms with the NPA selected (yellow lesion). D A binarized and marginal image of (A) and (C) to evaluate the location agreements with retinal haemorrhage and NPA. The haemorrhage in the NPA is light grey.
For the evaluation of NPA, one FA image was selected during the arteriovenous phase (between 45 s and 2 min) to clearly assess of capillary perfusion status. “Wand tool” in ImageJ was used to select regions of nonperfusion as describe previous reports (Fig. 1B, C) [15, 16].
OPTOS imaging can capture a wide field of retina, however the range will vary depending on eyelid, eyelash, and mydriasis conditions. An area in which the image could be evaluated retina clearly was selected in each patient (region of interest [ROI]) (Fig. 1B, A–D. White line). The same ROI was applied to red channel, green channel and FA images, in the same patient. Based on previous reports [17], the ratio between the green channel haemorrhages (GCH), red channel haemorrhages (RCH), and NPA was obtained, i.e., the pixels of the haemorrhages or NPA divided by the number of pixels in ROI. The green/red ratio were also determined, which is the GCH divided by the RCH in the same case.
To evaluate correlations in location with red channel haemorrhages and NPAs, the red channel image and FA image were merged (Fig. 1B, D). More specifically, after setting the opacity of red channel image to 50%, the red channel image and FA image were superimposed using “Overlay” in the ImageJ. Then, the position of the optic nerve head was visually confirmed to have no significant deviation. Following this, the two images were binarized by “Threshold” and were combined with “marge channels”. The overlapped areas of the haemorrhage and NPA were displayed as light grey. This enabled to count the number of pixels within/outside of the RCH region and NPA/PA for statistical analysis.
Statistical analysis
Linear regression analysis investigated the correlations both between the GCH and NPA and between the RCH and NPA. Multivariate analysis was followed by model selection using the second-order bias-corrected Akaike information criterion (AICc) index. The AIC is an established statistical measure to evaluate the relationships among variables, and the AICc is a corrected version of the AIC that provides an accurate estimate despite a small sample size. The chi-square test examined the location agreement between the RCH and NPA. All statistical analyses were performed using R 3.4.3 (R Foundation for Statistical Computing, Vienna, Austria).
Results
Thirty-two eyes of 32 patients with acute CRVO (18 males, 14 females; mean age, 67.1 ± 10.3 years) were studied. The mean logMAR VA was 1.03 ± 0.74 The mean duration of symptoms before examination was 5.9 ± 3.3 weeks (Table 1). FA detected the NPAs in all examined eyes (mean NPA, 39.2% ± 28.7%). The mean GCHs and RCHs were 10.4% ± 8.2% and 1.7% ± 1.7%, respectively. The mean ratio of the red/green haemorrhage in the same case was 0.16 ± 0.11. The intraclass correlation coefficient (ICC) of GCH and RCH were 0.858 [95% confidence interval (CI)]: 0.710–0.934, (P < 0.001), 0.945 [95% confidence interval] (CI): 0.891–0.973, (P < 0.001), respectively.
Table 1.
Clinical characteristics and result of the nonperfusion and haemorrhage ratio.
| Mean ± SD | Range | |
|---|---|---|
| Eyes | 32 | – |
| Gender, M:F | 18:14 | – |
| Age (years) | 67.1 ± 10.3 | 49.0–88.0 |
| LogMAR VA | 1.03 ± 0.74 | –0.08 to 2.30 |
| Duration of symptoms (weeks) | 5.9 ± 3.3 | 1.0–12.0 |
| Ratio of nonischemic:ischaemic | 4: 28 | – |
| NPA (%) | 39.2 ± 28.7 | 0.7–90 |
| GCH (%) | 10.4 ± 8.2 | 1.5–40.0 |
| RCH (%) | 1.7 ± 1.7 | 0.0–7.2 |
| Red/green haemorrhage ratio | 0.16 ± 0.10 | 0.00–0.41 |
M male, F female, SD standard deviation, LogMAR VA logarithm of the minimum angle of resolution visual acuity, NPA nonperfusion area, GCH green channel haemorrhages, RCH red channel haemorrhages.
Univariate analysis suggested that there were significant correlations between the GCH and NPA (r = 0.38; P = 0.022) and between RCH and NPA (r = 0.44; P = 0.010, linear regression analysis) (Fig. 2). As a result of AICc model selection, the optimal model for NPA identified only RCH as significantly correlated with NPA (Table 2).
Fig. 2. Correlation between the RCHs, GCHs, and NPA.
There are significant correlations between the GCHs and NPA (A) (r = 0.38; P = 0.022) and the RCHs and NPA (B) (r = 0.44; P = 0.010, linear regression analysis).
Table 2.
Relationship between GCH, RCH and NPA.
| Univariate analysis | Optimal model | |||||
|---|---|---|---|---|---|---|
| Variable | Coefficient | SE | P Value | Coefficient | SE | P Value |
| GCH | 1.37 | 0.59 | 0.028 | NS | NS | NS |
| RCH | 7.76 | 2.83 | 0.01 | 7.76 | 2.83 | 0.01 |
SE standard error of the mean, NS not significant.
A significant relationship was seen between the number of pixels within/outside of the RCH region and NPA/PA (P < 0.0001, chi-square test, Table 3), which suggested a significant association in location agreement between the RCH and NPA. The representative cases also showed that location agreement between the RCH and NPA (Supplementary Fig. 1).
Table 3.
Numbers of pixels in NPA and PA of RCH.
| NPA | PA | Total | |
|---|---|---|---|
| RCH | 1,312,175 | 509,067 | 1,821,242 |
| Outside RCH | 45,057,445 | 66,217,154 | 111,274,599 |
| Total | 46,369,620 | 66,726,221 | 113,095,841 |
A significant (P < 0.0001, χ2 test) relationship is seen between the number of pixels within/outside the RCH and NPA/PA.
PA perfusion area.
Univariate analysis of the association between haemorrhages and VA suggested significant correlations between the GCH and VA (r = 0.39; P = 0.031) and between the RCH and VA (r = 0.54; P = 0.001, linear regression analysis) (Supplementary Fig. 2). However, the AICc model identified only RCH as being significantly correlated with VA (Sup Table. 1).
Finally, we investigated the association between symptom duration and haemorrhage. Univariate analysis suggested significant correlations between duration and the GCH (r = –0.37; P = 0.03) and between duration and RCH (r = –0.47; P = 0.005); however, no significant correlations were seen between duration and the RCH/GCH ratio (r = –0.17; P = 0.29, linear regression analysis).
Discussion
In this retrospective study of eyes with acute CRVO, we evaluated the retinal haemorrhages in green and red wave laser images obtained by UWF imaging. Our results suggested that the haemorrhagic area detected by the red channel images is correlated significantly with the NPA. Delineation of retinal vascular nonperfusion is of particular clinical value in eyes with CRVO, because it is the most important factor predictive of development of neovascular complications such as vitreous haemorrhages or neovascular glaucoma [1, 17]. Since the colour UWF imaging can be obtained noninvasively, it is possible that the precise prediction of the ischaemic status by UWF red channel images might be important for determining the treatment strategy for CRVO.
The Optos colour fundus photographs are created by overlapping the scanned laser images of the green and red wavelengths. Each wavelength reaches a different retinal layer. The green channel image depicts the retina and its vasculature and the red channel images depict the deeper retina, RPE, and choroid. Previous reports have suggested that green channel imaging enables detection of all retinal haemorrhages in RVO [8, 9]. Since the present study showed that the red/green haemorrhage ratio was 0.16 in CRVO eyes, it seems that only certain types of haemorrhages are detectable in the red channel images. One possibility is that those haemorrhages are too thick to be transmitted by the red wave. However, this was beyond the scope of our current study and further investigation is needed using OCT.
Retinal haemorrhages are related to VA and NPA, which are factors involved in the severity of CRVO. Hayreh and Zimmerman investigated the prevalence of retinal haemorrhages in acute CRVO and reported that 16% had severe haemorrhages in non-ischaemic CRVO in contrast to 40% in ischaemic CRVO [7]. The current study also found a correlation between retinal haemorrhages and NPA or VA as previously reported. However, we showed for the first time that the RCHs were associated closely with the NPA or VA compared to the GCH. Our current results suggested that the RCHs might be useful for precisely evaluating the ischaemic status in eyes with CRVO. As mentioned previously, the RCHs might be deeper and thicker compared to the GCHs. Au et al., who showed that retinal haemorrhages in the nerve fibre layer were well correlated with a greater risk of ischaemia and neovascularization [10], proposed that retinal haemorrhages are first found at the level of the deep retinal capillary plexus and then the increased venous pressure subsequently causes haemorrhage in the intermediate capillary plexus, superficial capillary plexus and ultimately radial peripapillary capillary plexi. However, in BRVO eyes, the flame-shaped haemorrhages in the nerve fibre layer tend to be in the non-ischaemic macula [8]. Those investigators also demonstrated in OCT images that flame-shaped haemorrhages are in the inner retinal layer and non-flame-shaped haemorrhages are in the outer retina. The relationship between the shape of the retinal haemorrhage and ischaemia is still debated, but the presence of thick haemorrhages in the deep to surface layers certainly suggests increased venous pressure in those areas. Therefore, the haemorrhages that appeared on the red channel image may be related to retinal NPAs.
The current study investigated if the RCHs and GCHs changed over time after onset. It is possible that the RCHs would increase with time because the haemorrhage moved to the deeper retinal layers. However, the results showed that the amount of the GCHs and RCHs decreased with symptom duration and the ratio of the RCHs to the GCHs was not correlated with the symptom duration. This suggests that the RCHs are not caused by haemorrhagic migration but reflect the severity of the CRVO.
Our study had several limitations, including its small sample size and cross-sectional nature. Furthermore, we manually assessed the haemorrhages and NPAs based on previous reports [15, 16]. Some haemorrhages were microscopic, and it was difficult to assess them in totality. If we could automate the process, it would be more objective.
There is no precise method for measuring misalignment, when evaluating correlations in location with red channel haemorrhages and NPAs, However, the red channel and FA images position of the optic nerve head was visually confirmed to have no significant deviation. Because of this, it can be said that both images were taken at the same position.
Although the status of the retinal perfusion and retinal haemorrhages changes over time in eyes with CRVO, longitudinal changes were not examined. The evaluation of the RCH changes may deepen the relationship between ischaemia and haemorrhage.
In conclusion, retinal haemorrhages detected by UWF red channel imaging were less compared to green channel image, and its associated closely with retinal NPAs in eyes with acute CRVO. UWF red channel imaging allow us to identify ischaemia-related haemorrhage.
Summary
What was known before
Optos system acquires pseudocolor images by combining only red and green scanning lasers. The green laser image depicts the retina and its vasculature, and the red laser image highlights deeper structures, such as the retinal pigment epithelium (RPE) and choroid. The use of two different laser imaging systems might facilitate the evaluation of haemorrhages in the different retinal layers and properties.
What this study adds
Retinal haemorrhages detected by UWF red channel imaging were less compared to green channel imaging and associated closely with retinal NPAs in eyes with acute CRVO.
Supplementary information
Author contributions
ST, YT and TI were responsible for writing the report, conducting the search, extracting and analysing data, interpreting results, updating reference lists. KoN, RA, KO, ShoK, KeN and ShiK contributed to extracting data. YY, MI and KK provided feedback on the report.
Data availability
The datasets generated during and/or analysed during the current study are available from the corresponding author (ST) on reasonable request.
Competing interests
The authors declare no competing interests.
Footnotes
Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
These authors contributed equally: Shin Tanaka, Yui Tanaka.
Supplementary information
The online version contains supplementary material available at 10.1038/s41433-022-02337-3.
References
- 1.Natural history and clinical management of central retinal vein occlusion. The central vein occlusion study group. Arch Ophthalmol. 1997;115:486–91. doi: 10.1001/archopht.1997.01100150488006. [DOI] [PubMed] [Google Scholar]
- 2.McIntosh RL, Rogers SL, Lim L, Cheung N, Wang JJ, Mitchell P, et al. Natural history of central retinal vein occlusion: an evidence-based systematic review. Ophthalmology. 2010;117:1113–23. doi: 10.1016/j.ophtha.2010.01.060. [DOI] [PubMed] [Google Scholar]
- 3.Hayreh SS, Podhajsky PA, Zimmerman MB. Natural history of visual outcome in central retinal vein occlusion. Ophthalmology. 2011;118:119–33. doi: 10.1016/j.ophtha.2010.04.019. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Baseline and early natural history report. The central vein occlusion study. Arch Ophthalmol. 1993;111:1087–95. doi: 10.1001/archopht.1993.01090080083022. [DOI] [PubMed] [Google Scholar]
- 5.Hajdu D, Told R, Angeli O, Weigert G, Pollreisz A, Schmidt-Erfurth U, et al. Identification of microvascular and morphological alterations in eyes with central retinal non-perfusion. PLoS One. 2020;15:e0241753. doi: 10.1371/journal.pone.0241753. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Sakimoto S, Gomi F, Sakaguchi H, Akiba M, Kamei M, Nishida K. Analysis of retinal nonperfusion using depth-integrated optical coherence tomography images in eyes with branch retinal vein occlusion. Invest Ophthalmol Vis Sci. 2015;56:640–6. doi: 10.1167/iovs.14-15673. [DOI] [PubMed] [Google Scholar]
- 7.Hayreh SS, Zimmerman MB. Fundus changes in central retinal vein occlusion. Retina. 2015;35:29–42. doi: 10.1097/IAE.0000000000000256. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Muraoka Y, Uji A, Tsujikawa A, Murakami T, Ooto S, Suzuma K, et al. Association between retinal hemorrhagic pattern and macular perfusion status in eyes with acute branch retinal vein occlusion. Sci Rep. 2016;6:28554. doi: 10.1038/srep28554. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Muraoka Y, Uji A, Tsujikawa A, Murakami T, Ooto S, Suzuma K, et al. Association between retinal hemorrhagic patterns and perfusion status in eyes with acute central retinal vein occlusion. Retina. 2017;37:500–8. doi: 10.1097/IAE.0000000000001196. [DOI] [PubMed] [Google Scholar]
- 10.Au A, Hilely A, Scharf J, Gunnemann F, Wang D, Chehaibou I, et al. Relationship between nerve fiber layer hemorrhages and outcomes in central retinal vein occlusion. Invest Ophthalmol Vis Sci. 2020;61:54. doi: 10.1167/iovs.61.5.54. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Nagiel A, Lalane RA, Sadda SR, Schwartz SD. Ultra-widefield fundus imaging: a review of clinical applications and future trends. Retina. 2016;36:660–78. doi: 10.1097/IAE.0000000000000937. [DOI] [PubMed] [Google Scholar]
- 12.Prasad PS, Oliver SC, Coffee RE, Hubschman JP, Schwartz SD. Ultra wide-field angiographic characteristics of branch retinal and hemicentral retinal vein occlusion. Ophthalmology. 2010;117:780–4. doi: 10.1016/j.ophtha.2009.09.019. [DOI] [PubMed] [Google Scholar]
- 13.Neubauer AS, Kernt M, Haritoglou C, Priglinger SG, Kampik A, Ulbig MW. Nonmydriatic screening for diabetic retinopathy by ultra-widefield scanning laser ophthalmoscopy (Optomap) Graefes Arch Clin Exp Ophthalmol. 2008;246:229–35. doi: 10.1007/s00417-007-0631-4. [DOI] [PubMed] [Google Scholar]
- 14.Sato A, Asaoka R, Tanaka S, Nagura K, Tanaka Y, Arasaki R, et al. Using ultra-widefield red channel images to improve the detection of ischemic central retinal vein occlusion. PLoS One. 2021;16:e0260383. doi: 10.1371/journal.pone.0260383. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Croft DE, van Hemert J, Wykoff CC, Clifton D, Verhoek M, Fleming A, et al. Precise montaging and metric quantification of retinal surface area from ultra-widefield fundus photography and fluorescein angiography. Ophthalmic Surg Lasers Imaging Retin. 2014;45:312–7. doi: 10.3928/23258160-20140709-07. [DOI] [PubMed] [Google Scholar]
- 16.Wykoff CC, Brown DM, Croft DE, Major JC, Jr, Wong TP. Progressive retinal nonperfusion in ischemic central retinal vein occlusion. Retina. 2015;35:43–47. doi: 10.1097/IAE.0000000000000277. [DOI] [PubMed] [Google Scholar]
- 17.Tsui I, Kaines A, Havunjian MA, Hubschman S, Heilweil G, Prasad PS, et al. Ischemic index and neovascularization in central retinal vein occlusion. Retina. 2011;31:105–10. doi: 10.1097/IAE.0b013e3181e36c6d. [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Data Availability Statement
The datasets generated during and/or analysed during the current study are available from the corresponding author (ST) on reasonable request.