Article Text
PDF
Abstract
Background/aims To describe the morphological features of choroidal neovascularisation (CNV) and to report the ability of optical coherence tomography angiography (OCT-A) to detect the presence of myopic CNV by means of this new technique.
Methods Myopic CNV cases were individuated from a pool of patients with pathological myopia consecutively presenting between October 2015 and March 2016. OCT-A images were assessed for classification of morphological features, and to estimate sensitivity and specificity.
Results Thirty-six eyes of 28 consecutive patients with myopic CNV were included. In 4 out of 36 eyes it was not possible to classify the CNV ‘shape’, ‘core’, ‘margin’ and ‘appearance’ because the vascular network was not clearly visualised due to the poor quality of the examination. CNV shape on OCT-A was rated as circular in 9 eyes and irregular in 23 eyes. CNV core was visible in 11 eyes. CNV margin was considered as well defined in 16 eyes and poorly defined in 16 eyes. CNV appearance showed an ‘interlacing’ aspect in 16 eyes and a ‘tangled’ aspect in the other 16 eyes. A total of 11 CNVs were defined as active, 9 of which (81.8%) were interlacing, while a total of 21 were inactive, 14 of which (66.7%) were tangled. OCT-A sensitivity turned out to be 90.48% and specificity was 93.75%.
Conclusions We describe the OCT-A features of myopic CNV secondary to pathological myopia and demonstrate its high sensitivity and specificity for neovascular detection. Qualitative evaluation of OCT-A characteristics may allow one to recognise different patterns, possibly corresponding to different degrees of neovascular activity.
- Imaging
- Macula
- Neovascularisation
- Diagnostic tests/Investigation
- Retina
Statistics from Altmetric.com
Introduction
High myopia is one of the leading causes of visual impairment in many developed countries. Pathological myopia is defined by an axial length of the eye greater than 26 mm and by a refractive error of −6 dioptres (D) or more, associated with complications of the posterior segment due to progressive and excessive elongation of the globe.1 Progressive posterior segment elongation is accompanied by degenerative changes, including the sclera, optic disc, choroid, Bruch's membrane, retinal pigment epithelium (RPE) and neural retina. These degenerative changes may lead to the development of macular lesions, such as myopic choroidal neovascularisation (CNV).2 The diagnosis of myopic CNV is confirmed by fluorescein angiography (FA) and structural spectral-domain optical coherence tomography (SD-OCT) B-scan.3 ,4 Myopic CNVs are usually type 2 neovascularisations or ‘classic’ on FA, with well defined hyperfluorescence on early frames and leakage of the dye on late frames.3 ,4 A structural SD-OCT B-scan shows a dome-shaped hyperreflective elevation above the RPE, often associated with discrete retinal changes, including oedema and neurosensory serous retinal detachment.
Optical coherence tomography angiography (OCT-A) is a recently introduced, non-invasive imaging and en face technique that provides three-dimensional (3D) images of dynamic blood perfusion within microcirculatory tissue beds. The imaging contrast of OCT-A images is based on the intrinsic optical scattering of signals backscattered by the moving of blood cells in a patient's blood vessels.5 OCT-A has the advantage not to require any dye injection and this permits better and more detailed visualisation of vascular networks without disturbance of dye leakage.
Our purpose was to report the ability of OCT-A to detect, in a non-invasive way, the presence of myopic CNV and to describe the morphological features of CNV by means of this new technique.
Methods
Myopic CNV cases were individuated from a pool of patients with pathological myopia consecutively presenting between October 2015 and March 2016 at the Retina and Imaging Unit, Department of Ophthalmology, University Vita-Salute, San Raffaele Hospital in Milan. This retrospective study followed the tenets of the Declaration of Helsinki for research involving human subjects. Informed consent was obtained from all subjects. Institutional Review Board approval was obtained for this retrospective study, and the retrospective review of patient information.
Inclusion criteria were age ≥18 years old; diagnosis of pathological myopia defined by a refractive error (spherical equivalent) >−8.00 D or an axial length >26.5 mm, accompanied by characteristic degenerative changes of the sclera, choroid, and retina; presence of CNV diagnosed by means of FA and structural SD-OCT B-scan (with or without metamorphopsia/persistent visual loss); adequate pupillary dilation; and fixation to permit high-quality OCT imaging. Ocular exclusion criteria consisted of any disease other than CNV secondary to pathological myopia (including retinal vascular diseases, vitreoretinal diseases, history of central serous retinopathy or macular dystrophies); history of ocular inflammation in the study eye; significant media opacities; and presence of large haemorrhage preventing the incidence of light from penetrating the retina.
Each patient underwent a comprehensive ophthalmologic examination, including measurement of best-corrected visual acuity (BCVA), dilated slit lamp anterior segment and fundus biomicroscopy, FA (when deemed necessary by the treating physician), SD-OCT (Spectralis+HRA; Heidelberg Engineering, Heidelberg, Germany) and OCT-A. OCT-A was performed through AngioPlex CIRRUS HD-OCT model 5000 (Carl Zeiss Meditec, Inc., Dublin, USA), with a scanning area of 3×3 mm, centred on the foveal area. AngioPlex uses optical microangiography (OMAG), a recently developed imaging technique that produces 3D images of dynamic blood perfusion within micro-circulatory tissue beds at an imaging depth up to 2.0 mm.6 AngioPlex CIRRUS HD-OCT model 5000 contains A-scan rate of 68 000 scans per second, using a superluminescent diode (SLD) centred on 840 nm. A three by three angio cube contains 245 B-scan slices repeated up to four times at each B-scan position. Each B-scan is made up of 245 A-scans, each A-scan is 1024 pixels deep.
The automatic segmentation provided by the OCT-A software was manually adjusted by two expert retina specialists (LQ and CG) by modulating the segmentation lines for correct visualisation of the capillary plexus, outer retinal layers and choriocapillaris to better identify the CNV plane. The OCT-A images were assessed for classification of morphological features, including CNV shape, CNV core, CNV margin and overall appearance. CNV shape was classified as ‘circular’ or ‘irregular’. CNV core was classified as ‘visible’ or ‘not visible’; if the core was visible, CNV was classified as ‘central core’ or ‘eccentric core’. CNV margin on OCT-A was classified as ‘well defined’ or ‘poorly defined’ on the basis of its appearance and its borders. CNV was classified as mainly ‘tangled’ or mainly ‘interlacing’, on the basis of the overall appearance, when >50% of the lesion was characterised by one of these features. To assess the activity of the lesions using traditional imaging, we considered as active those lesions which presented well defined hyperfluorescence in the early phase of FA with leakage in the late phase (reference examination); in addition, active lesions appeared as elevation of the RPE, presenting subretinal or intraretinal hypo-reflective or hyper-reflective exudation on structural SD-OCT B-scan along with overlying fuzzy area and absence of external limiting membrane visibility.7 ,8 Lesions presenting staining of a CNV scar on FA and a well defined profile with hyper-reflective borders on structural SD-OCT B-scan were defined as inactive. Disagreement regarding interpretation of the different features was resolved by open adjudication.
To estimate the sensitivity and specificity of OCT-A, an additional cohort of patients with pathological myopia with no evidence of CNV at FA and structural SD-OCT B-scan were merged with the OCT-A study group as negative controls (negative control group). Ocular exclusion criteria consisted of any disease other than pathological myopia (including retinal vascular diseases, vitreoretinal diseases, history of central serous retinopathy or macular dystrophies), and previous treatments, including anti-vascular endothelial growth factor (VEGF) intravitreal injections, photodynamic therapy, laser photocoagulation or vitrectomy, in the study eye. OCT-A of the two cohorts were randomly presented and independently evaluated by two expert retina specialists (AC and FC) to confirm the presence/absence of CNV. If an eye was determined to have CNV on FA and structural SD-OCT B-scan, OCT-A showing an abnormal neovascular network was considered as true positive; if CNV was not visualised on OCT-A, the exam was considered as false negative. If FA and structural SD-OCT B-scan did not demonstrate a CNV (negative control group), OCT-A with no evidence of CNV was considered as true negative; if a CNV was detected, the exam was taken as false positive. Both readers evaluated OCT-A without visualisation of the corresponding structural SD-OCT B-scan. Disagreement was resolved by open adjudication.
Statistical analysis was performed using SPSS software V.21 (SPSS, Inc., Chicago, Illinois, USA). All data were expressed as mean±SD. Categorical variables were analysed using χ2 test while continuous variables were analysed using Student's t test. The correlation between readers was tested using Pearson's correlation: a correlation coefficient (r)>0.75 was considered as ‘good to excellent’, 0.50–0.75 as ‘moderate to good’, 0.25–0.50 as ‘fair’ and 0.00–0.25 as ‘little or no’ relationship. The significance level for all testing was set at p<0.05.
Results
Study population and main clinical findings
Thirty-six eyes of 28 consecutive patients (5 men and 23 women, mean age 64±14.7 years (range, 26–84 years)) diagnosed with myopic CNV were enrolled. Eight patients presented bilateral CNV. Six out of 36 eyes presented newly diagnosed CNV, while 30 out of 36 eyes had received previous anti-VEGF injections (Lucentis, Genentech/Novartis, Inc., in all cases) for actively leaking CNV. Demographic data and angiographic features of the study group are listed in table 1.
Demographic characteristics and optical coherence tomography angiography features of patients with choroidal neovascularisation secondary to pathological myopia
Considering the dye angiography and structural SD-OCT B-scan criteria we found 11 out of 36 eyes with active CNVs and 21 out of 36 eyes with inactive CNVs.
Thirty-two eyes of 32 subjects with pathological myopia (27 women and 5 men; mean age 56.2±14.4 years, range 26–84 years) without evidence of CNV on FA and structural SD-OCT B-scan were enrolled as the negative control group.
OCT-A features of myopic CNV
In the analysis for classification of morphological features, 4 out of 36 eyes were excluded because the vascular network was not clearly visualised due to poor quality of the examination. The correlation between readers for appearance, shape, margin and location of core was ‘good to excellent’, while for visibility of core, it was ‘moderate to good’ (table 2).
Pearson correlation between readers for optical coherence tomography angiography features of patients with choroidal neovascularisation secondary to pathologic myopia
CNV shape on OCT-A was rated as circular in 9 out of 32 eyes, irregular in 23 out of 32 eyes. CNV core was visible in 11 out of 32 eyes, and was not visible in 21 out of 32 eyes; in the 11 eyes with a visible core, the core position was considered as central in 8 CNVs and as eccentric in 3 CNVs. CNV margin was considered as well defined in 16 out of 32 eyes and poorly defined in 16 out of 32 eyes. Moreover, CNV overall appearance was classified as mainly tangled in 16 out of 32 eyes and mainly interlacing in 16 out of 32 eyes. Examples of shape, core, margin and overall appearance are represented in figures 1⇓⇓⇓–5.
Fluorescein angiography (FA), structural spectral domain optical coherence tomography (SD-OCT) B-scan and OCT angiography (OCT-A) of choroidal neovascularisation (CNV) in the right eye of patient #6 with pathological myopia. Early and late phases of FA (A, B) reveal dye staining suggestive of inactive type 2 (classic) CNV. OCT-A (3×3 mm) segmentation of outer retinal layers (C) and corresponding structural SD-OCT B-scan show an irregular, well defined, tangled CNV (arrowhead points out the eccentric core).
Fluorescein angiography (FA), structural spectral domain optical coherence tomography (SD-OCT) B-scan and OCT angiography (OCT-A) of choroidal neovascularisation (CNV) in the left eye of patient #6 with pathological myopia. Early and late phases of FA (A, B) reveal dye staining suggestive of inactive type 2 (classic) CNV. OCT-A (3×3 mm) segmentation of outer retinal layers (C) and corresponding structural SD-OCT B-scan show an irregular, well defined, tangled CNV. The core is not visible.
Fluorescein angiography (FA), structural spectral domain optical coherence tomography (SD-OCT) B-scan and OCT angiography (OCT-A) of choroidal neovascularisation (CNV) in the right eye of patient #11 with pathological myopia. Early and late phases of FA (A, B) reveal dye staining suggestive of inactive type 2 (classic) CNV. OCT-A (3×3 mm) segmentation of outer retinal layers (C) and corresponding structural SD-OCT B-scan show an irregular, poorly defined, tangled CNV (arrowhead points out the central core).
Fluorescein angiography (FA), structural spectral domain optical coherence tomography (SD-OCT) B-scan and OCT angiography (OCT-A) of choroidal neovascularisation (CNV) in the right eye of patient #7 with pathological myopia. Early and late phases of FA (A, B) reveal dye leakage suggestive of active type 2 (classic) CNV. OCT-A (3×3 mm) segmentation of outer retinal layers (C) and corresponding structural SD-OCT B-scan show a circular, well defined, interlacing CNV (arrowhead points out the central core).
Fluorescein angiography (FA), structural spectral domain optical coherence tomography (SD-OCT) B-scan and OCT angiography (OCT-A) of choroidal neovascularisation (CNV) in the right eye of patient #14 with pathological myopia. Early and late phases of FA (A. B) reveal dye leakage suggestive of active type 2 (classic) CNV. OCT-A (3×3 mm) segmentation of outer retinal layers (C) and corresponding structural SD-OCT B-scan show a circular, well defined, interlacing CNV (arrowhead points out the central core).
The tangled network was defined on OCT-A by a loose lacing appearance with long filamentous vessels and few large branches with a thick vessel wall. This appearance was often associated with absence of neovascular activity (14 out of 16 tangled CNVs: 87.5%) on conventional imaging, including FA and structural SD-OCT B-scan (figures 1⇑–3). However, 9 out of 16 CNVs (56.25%) classified as interlacing showed activity on FA and structural SD-OCT B-scan. Of note, the five patients with newly diagnosed CNV showed an interlacing appearance on OCT-A (figures 4 and 5).
A total of 11 CNVs were defined as active, 9 of which (81.8%) were interlacing and 2 (18.2%) were tangled, while a total of 21 were inactive, 7 of which (33.3%) were interlacing and 14 (66.7%) were tangled.
The features most frequently observed (irregular in shape, with poorly defined margin and without visible core) were all collectively detected in 9 out of 32 eyes (28.1%).
A ‘fair’ but significant correlation was found between the presence of neovascular activity and interlacing appearance (r=0.46; p=0.008), and interlacing appearance and newly diagnosed CNV (r=0.43; p=0.014).
Assessment of CNV detection on OCT-A
In the analysis of sensitivity and specificity, 4 out of 36 eyes with CNV (study group) were excluded because, due to poor quality of the examination, it was not possible to clearly visualise the presence of the vascular network. All these excluded eyes presented extensive RPE/chorioretinal atrophy, being possibly responsible for the poor quality of OCT-A images. Moreover, 11 out of 36 eyes with CNV (study group) were excluded from the analysis of sensitivity and specificity because of absence of FA performed on the same day as OCT-A. In the end, a total of 21 eyes of 17 patients (mean age: 57.8±14.5; 3 men/14 women) were included to estimate sensitivity and specificity. Thirty-two eyes without evidence of CNV on FA and structural SD-OCT B-scan were enrolled as a negative control group. There was no significant difference regarding all demographic data considered between the two groups (age, p=0.6; gender, p=0.9).
The total pool of patients, which included the study group and the negative control group, was evaluated. The correlation between readers was ‘good to excellent’ (r=0.79; p<0.0001). Both readers correctly identified on OCT-A myopic CNVs in 19 out of 21 eyes, and correctly excluded 30 out of 32 eyes with pathological myopia without CNV; in 2 out of 32 eyes both readers incorrectly identified the presence of CNV on OCT-A and this may be due to large atrophy secondary to high myopia which could be misdiagnosed as CNV. Thus, OCT-A sensitivity turned out to be 90.48% and specificity was 93.75% (p<0.0001), with positive predictive value and negative predictive value equal to 0.9048 and 0.9375 respectively.
Discussion
This study reports the OCT-A features of CNV in patients affected by pathological myopia. In many cases, myopic CNVs appeared as irregular in shape, with poorly defined margins, and without a visible core. Moreover, OCT-A showed that in most cases active myopic CNVs appeared to be mainly interlacing, while in inactive CNVs the neovascular network appeared to be mainly tangled. These data suggest that OCT-A could be considered as a potentially useful tool in characterising CNV by its morphology and detecting CNV activity in eyes with pathological myopia.
OCT-A also recognised the absence of neovascularisations in myopic eyes without evidence of CNV on FA and structural SD-OCT B-scan. This analysis suggests that OCT-A seems particularly useful in grossly visualising neovascular networks, as an adjunct to the well established imaging techniques of the retina and choroid, such as structural SD-OCT B-scan and FA. A structural SD-OCT B-scan provides a cross-sectional, 3D image of tissues, demonstrating the structure of retinal and choroidal layers, thus allowing one to observe haemorrhage or fluid with different signal strengths. FA directly detects blood flow and vascular networks of the retina, thus showing the CNV thanks to dye staining and leakage. It is noteworthy that OCT and dye angiography techniques have some inherent limits compared with OCT-A. OCT cannot visualise fine vessels nor provide functional information of retinal microcirculation. However, FA does not provide 3D images and needs an intravenous dye injection that can cause side effects, from nausea to serious anaphylactic reactions, thus limiting its application. OCT-A compensates for some of these limits and provides the vessel arrangement of CNV through different techniques (OMAG was used in the current study), which distinguishes CNV from other structures by detecting flowing red blood cells.
Here we demonstrated high sensitivity (90.48%) and specificity (93.75%) for neovascular detection with OCT-A in pathological myopia. Several studies were carried out to research the potential value of OCT-A in a variety of ocular diseases. Miyata et al9 recently reported the OCT-A detection of myopic CNVs demonstrating successful detection in 94.1% of cases. The detection rate by OCT-A seemed higher than that in previous studies on OCT-A for exudative age-related macular degeneration (AMD). De Carlo et al10 evaluated various CNVs (related to neovascular AMD, angioid streaks, central serous chorioretinopathy) and reported a sensitivity of 50% for CNV detection by OCT-A. Moult et al11 reported that 84.2% of CNVs were detected by OCT-A in patients with neovascular AMD. Kuehlewein et al12 evaluated type 1 CNV in patients with AMD, reporting an OCT-A detection rate of 75%. As suggested by Miyata et al, the higher detection rate of CNV in pathological myopia can be attributed to the location and size of CNVs.9 Type 2 CNVs are located above the RPE and imaging can be more clearly obtained. Of note, the high sensitivity and specificity in our series for myopic CNV detection by OCT-A may be due to the absence of intraretinal/subretinal fluid and preserved retinal morphology in myopic CNV, which limits the potential artefacts and thus represents an obvious advantage for OCT-A interpretation. A high sensitivity and specificity suggests that the OCT-A can be considered a good test for screening and follow-up of myopic CNV.
However, in our series, despite the obvious advantage of OCT-A (a dye-less, no time-consuming 3D method), in 4 out of 36 eyes (11%) it was not possible to classify the shape, margin and overall appearance because the neovascular network was not clearly visualised, despite its evidence on FA/structural SD-OCT. This was fewer than reported in the paper by Miyata et al,9 in which the quality of images was not sufficient for CNV assessment in 6 out of 23 eyes (26%). Of note, in our series, the four eyes for which the neovascular network was not clearly visualised all presented extensive RPE/chorioretinal atrophy areas, possibly being responsible for the poor quality and interpretation of OCT-A images.
Imaging neovascular membranes with OCT-A has several limitations. First, poor fixation could severely limit the quality of the images, resulting in noisy images with potentially significant motion artefacts. Low-quality image acquisition decreases the likelihood of detecting small or poorly perfused CNVs. Projection artefacts may also limit accurate evaluation: the light beam that encounters the superficial retinal plexus may pass through the moving blood cells (45% of photons are estimated to pass across the vessel), and the projection of the scanned vessel may appear on the reconstruction of the deeper retinal plexus.13 Many myopic lesions, such as lacquer cracks, RPE atrophy and retinoschisis may disturb the actual appearance of CNV on OCT-A. These features of pathological myopia lead to increased obstacles for doctors and patients undergoing OCT-A examination.
We acknowledge there are several limitations of this study. The series we analysed is relatively small. Moreover, this is a retrospective study and a prospective analysis will be useful for expanding the findings provided in this report. In addition, longitudinal assessment of myopic CNV and its features on OCT-A have not been investigated; however, this was beyond the scope of the current study. Finally, in our study we excluded eyes with the presence of a large haemorrhage preventing the incidence of light from penetrating the retina; this design could improve the sensitivity and specificity in favour of OCT-A by excluding the cases for which FA has better performance.
In this study, while we described different CNV patterns and their correlation with neovascular activity, we do not claim a new classification of the CNV because of the obvious limitations of the current analysis (such as the retrospective design and the sample of the study for each category not being appropriately represented).
In conclusion, we described the OCT-A features of myopic CNV secondary to pathological myopia and demonstrated its high sensitivity and specificity for neovascular detection. Qualitative evaluation of OCT-A characteristics may allow one to recognise different patterns possibly corresponding to different degrees of neovascular activity. For this reason, although FA and structural SD-OCT B-scan remain the diagnostic gold standard for the neovascular activity of CNV, OCT-A could be an alternative useful tool that clinicians could use to evaluate patients with high myopia, especially those patients with a history of allergic reaction. However, given that, in contrast to FA and structural SD-OCT B-scan, OCT-A does not provide information on leakage of the dye and retinal oedema/neurosensory detachment, to date, OCT-A should not be considered as a standalone test, but rather as an additional tool to better characterise myopic CNV and help the treatment decision.
References
Footnotes
Contributors GQ and FB have contributed equally to this study, and should be considered as equivalent authors.
Competing interests None declared.
Patient consent Obtained.
Ethics approval Ethics Committee San Raffaele.
Provenance and peer review Not commissioned; externally peer reviewed.