Grading of macular perfusion in retinal vein occlusion using en-face swept-source optical coherence tomography angiography: a retrospective observational case series

The traditional approach for evaluation of macular perfusion following retinal vein occlusion (RVO) consisted of assessment of the superficial capillary plexus (SCP) using fundus fluorescein angiography (FFA), which relayed information on the morphological alteration of the perifoveal capillary network due to ischemia. These changes were used as indicators of visual prognosis after RVO episode according to the area of the macula involved [1, 2]. However, evidence from studies on animal models and normal human subjects revealed that the retinal microcirculation targeted in RVO is far more complex to be truly represented by conventional FFA as the sole imaging modality. The reason is that the inner retina is nourished by a complex vascular network composed of superficial, intermediate, and deep capillary plexuses arranged at different axial levels [3‐5]. Moreover, the vascular structure of the deep capillary plexus (DCP) that involves direct communication with major veins and the lack of vascular smooth muscles render the DCP most vulnerable to hemodynamic disturbances following RVO and the ensuing hypoperfusion compared to the SCP [6‐8]. Hypoperfusion of the DCP could be detrimental to the visual function due to the pivotal role of the DCP in nourishment of the watershed zone that is located between the inner nuclear layer (INL) and the outer plexiform layer (OPL) and that contains neuronal synapses transmitting visual signal from the photoreceptors [9]. Given that conventional FFA lacks the ability of depth-resolved imaging, it cannot provide useful information on the perfusion profile of capillary plexuses located at different axial depths. The reason is that the relatively fast and profuse leakage of fluorescein dye from the choriocapillaris, along with scattering of light from the nerve fiber layer and deeper retinal layers create fuzzy background fluorescence that impedes detailed imaging of the perfusion profile of the DCP. In fact, FFA reveals only the SCP perfusion. These limitations have been elaborated in several studies that focused on imaging the retinal vascular plexuses using optical coherence tomography angiography (OCTA) technology as alternative to conventional FFA [6, 10‐13]. Another important limitation of FFA is that it involves intravenous dye injection which renders it an unrealistic and possibly hazardous follow-up imaging modality given the chronic course of RVO and the need for repeated evaluation [14]. The recently - introduced swept-source optical coherence tomography angiography (SS-OCTA) instated major advances in imaging the retinal vascular plexuses in cases of RVO, and particularly defrayed the drawbacks of conventional FFA. Firstly, SS-OCTA relies on motion contrast techniques to detect moving erythrocytes in relation to static tissue to generate images of the retinal vascular plexuses obviating the need for dye injection. Secondly, SS-OCTA is integrated in the swept-source structural optical coherence tomography (SS-OCT). The SS-OCT incorporates a long wavelength (1050 nm) scanning light, reduced sensitivity roll-off feature, and ultra-high-speed image acquisition. These implements enable deeper penetration with minimal light scattering, hence superior axial resolution and segmentation of different retinal layers. The result is generation of ultra-high-definition images of the SCP and the DCP in separate layers non-invasively [5, 15‐19]. The aim of the present study is to evaluate the efficacy of SS-OCTA in grading macular perfusion in cases of RVO and compare the findings with SS-OCT and conventional FFA.

This is a retrospective observational case series in which we reviewed the clinical data, FFA, SS-OCT, and SS-OCTA images of all consecutive patients diagnosed with different types of RVO in a private practice from September 2016 to March 2018, and compared the findings.

The study included patients with different types of RVO [central retinal vein occlusion (CRVO), hemicentral retinal vein occlusion (HRVO), and branch retinal vein occlusion (BRVO)]. If the occlusion affected one of the major tributaries of the central retinal vein, the condition was designated major BRVO; whereas whenever the occlusion involved a smaller macular tributary, the condition was designated macular BRVO [20]. Patients with major BRVO had to have concomitant macular involvement to be eligible for the study. All patients participating in the study had to have sufficient resolution of retinal hemorrhages to allow for good quality FFA, SS-OCT, and SS-OCTA. Patients were allowed to enter the study regardless of whether or not they received single or multiple therapies for macular edema or neovascularization secondary to RVO, whether anti-vascular endothelial growth factor (anti-VEGF) agents, intravitreal triamcinolone acetonide (IVTA), and/or laser photocoagulation. Patients with concomitant retinal vascular disease known to affect the retinal vascular plexuses as diabetic retinopathy, retinal vasculitis, combined retinal arterial and venous occlusions, and those with media opacities that were dense enough to impede sufficient image quality for reliable interpretation were excluded from the study.

FFA images were obtained using Topcon TRC 50DX fundus camera (Topcon Corporation, Tokyo, Japan). FFA frames were evaluated for retinal capillary non-perfusion (CNP) following an episode of RVO. RVO was designated ischemic whenever FFA revealed CNP area ≥ 5 disc diameters (DD), and ≥ 10 DD for BRVO and for CRVO and HRVO, respectively [21, 22]. The perifoveal capillary arcade was evaluated in the early venous phase of the angiogram for the presence of signs of macular ischemia. These included disruption of the perifoveal continuum of intercapillary plexus, enlargement of inter-capillary pillars, telangiectasia, capillary pruning and rarefaction, and presence of areas of capillary non-perfusion (CNP). The macula was considered ischemic if ischemic changes of the perifoveal capillary arcade were detected in ≥2 quadrants on FFA.

SS-OCT images were acquired using the DRI OCT Triton machine version 10.11 (Topcon Corporation, Tokyo, Japan); whereas SS-OCTA images were acquired using an integrated blood flow detection algorithm; OCTARA (Optical Coherence Tomography Angiography Ratio Analysis). The SS-OCT scans were evaluated for the presence of macular edema. The central macular thickness (CMT) value was obtained from 1.00 mm diameter circle on the ETDRS grid. Morphological alteration in the retinal microstructure were assessed in a 1.00 mm area centered around the fovea for the presence of disorganized retinal inner layers (DRIL), disruption of the retinal outer layers (DROL) causing loss of the continuity of the external limiting membrane, and/or the photoreceptors inner segment/outer segment (IS/OS) layer. For assessment of DRIL and DROL, the B-scan of maximum affection was selected and two lines were extended temporally and nasally from the fovea up to a distance of 1000 μm in each direction, using the DRI OCT Triton linear measurement tool. DRIL and DROL were considered significant if they involved ≥50% of the 1.00 mm area around the fovea.

OCTA acquisition protocol in the macular region consisted of a 6 × 6 mm area or 9 × 9 mm area centered onto the fovea. Whenever higher resolution was needed, a 4.5 × 4.5 mm area, or a 3 × 3 mm area were used. By default, the integrated software (IMAGEnet 6 ophthalmic data system) deploys automated segmentation to delineate the SCP and the DCP. Images of the SCP were obtained using two slabs located between the internal limiting membrane (ILM) with an offset of 2.6 μm and the inner plexiform layer (IPL) with an offset of 15.6 μm, whereas DCP images were obtained using two slabs located between the IPL with an offset of 15.6 μm and the outer plexiform layer (OPL) with an offset of 70.2 μm. In cases of significant disorganization of retinal layers due to cystoid macular edema (CME) or sizeable sub-retinal fluid, the integrated automated segmentation feature failed to detect the correct boundaries of the SCP and the DCP due to posterior displacement of flow signals, and we had to resort to manual adjustment of the segmentation slab. To perform manual segmentation the operator manually places two segmentation lines at sequential depths guided by corresponding SS-OCT images to reveal the maximum extent of the SCP and the DCP as identified by their respective morphological criteria. The SS-OCTA software generates color-coded flow density maps of the SCP and the DCP, each layer separately. In this map, vessel density in a given area is inferred from the decorrelation motion contrast signal provided by SS-OCTA, where high-flow is represented by increased vessel density and vice-versa [23]. Different vessel densities are then given color-codes and numeric percentage values that reflect the percentage area occupied by blood vessels, where bright red color represents areas of highest density, hence high numeric percentage; whereas dark blue represents areas of no detectable vessels, hence low or zero numeric percentage. Intermediate color shades represent variable grades of vessel density [24].

Much of our understanding on the relationship between visual function and macular ischemia in patients with RVO is derived from FFA. The conclusion that visual dysfunction is in lockstep with ischemic changes of the FAZ as reflected by FFA is at most a conjecture, with discrepant reports describing the relationship between the grade of visual dysfunction and status of the FAZ following RVO [25‐27]. The reason is that FFA, due to its inherent limitations, fails to include critical information on structural changes of the retinal micro-structure and of the DCP, which are cardinal determinants of the visual outcome in patients with RVO [28‐31]. Accordingly, absolute reliance on FFA for evaluation of macular ischemia following RVO episode could pose major compromise of our perception of the perfusion profile of the macula. In the present study, we focused on quantifying the pathological changes in the retinal vascular plexuses secondary to RVO using SS-OCTA and to devise a perfusion profile of the macula that would accurately represent the ischemic insult inflicted. Furthermore, we correlated our findings with the anatomic changes evident on SS-OCT and FFA and their relation to the visual function. Our results showed a statistically significant discrepancy between the grade of ischemia in SCP and DCP on SS-OCTA, in the sense that more ischemic quadrants were present at the level of the DCP relative to the SCP, which showed less ischemic damage. Moreover, perifoveal capillary ischemia and reduced vessel density per quadrant were more pronounced in the DCP relative to the SCP. These results confirm reports from several authors that ischemic damage in RVO is preferentially main-seated in the DCP and often precedes SCP ischemia [4, 7, 10, 29‐33]. In comparison with our results, Coscas et al. [6] reported in a series of 54 patients with RVO, more extensive ischemic damage in the DCP relative to the SCP (84.3% vs. 58.8%, respectively). Cardoso et al. [13], demonstrated in a reliability analysis of 81 eyes with RVO that capillary changes associated with ischemia were primarily located in the DCP. Chung et al. [34], reported in a study recruiting 12 patients with RVO, patchy CNP areas involving the DCP that were not detected on the SCP using OCTA. Further corroborating evidence of more selective DCP affection in RVO was demonstrated in a retrospective series including 17 eyes by Samara et al. [8], and in two series by Suzuki et al. [35, 36] that included 12 eyes with BRVO and CRVO and 28 eyes with BRVO, respectively. These studies detected abnormally enlarged FAZ that was more pronounced in the DCP in eyes with RVO compared to the fellow unaffected eye. Moreover, Suzuki et al. [35], detected reduced retinal flow area that was most marked in the DCP in eyes with RVO compared to the unaffected fellow eye. Similarly, Adhi et al. [9], reported in a prospective case series of 23 patients with RVO, decreased vascular perfusion in the DCP that surpassed in frequency of occurrence and severity of involvement that of the SCP. Another important aspect of the present study, is the analysis of the correlation between SS-OCT anatomical features with ischemic changes revealed on SS-OCTA. Increased CMT and DROL were the most sensitive SS-OCT indices for DCP ischemia. On the other hand, DRIL was the least reliable SS-OCT index for SCP or DCP, though it had a direct proportional correlation with the severity of perifoveal capillary ischemia. Henceforth, we could propose a high-risk SS-OCT profile that includes increased CMT, DROL and DRIL and that portends significant ischemia in the DCP and in the perifoveal capillary arcade, respectively. Our conclusion on the positive correlation between increased CMT and DCP ischemia is ratified by Spaide’s theory on the cause-effect relationship between compromised blood flow within the DCP and reduced clearance of interstitial fluid from the retina and that eventually leads to macular edema [37]. Further corroborating evidence is provided from the work of Tsuboi et al. [38], who demonstrated a statistically significant positive correlation between flow-voids; equivalent to diminished vessel density in the SCP and the DCP and persistent macular edema in BRVO patients. In comparison, our results revealed that FFA was not reliable in assessment of ischemia. FFA representation of macular perfusion was significantly proportionate to SCP perfusion on SS-OCTA but in most cases it could not reveal the entire extent of macular ischemia as revealed by SS-OCTA depiction of DCP. Our findings are in line with two studies of the retinal vascular layers by Mendis et al. [39], and Spaide et al. [10], who compared FFA versus confocal laser scanning microscopy, and FFA versus OCTA respectively in evaluating the retinal vascular layers. Both studies concluded that FFA was unable to provide complete information on the DCP. In accordance with our findings, Chung et al. [34], reported poor agreement between FFA and OCTA in revealing CNP following RVO with clear superiority of OCTA over FFA (91.67% vs. 58.33%), particularly in DCP non-perfusion and in detection of perifoveal capillary ischemia. Similarly, two case series including 28 eyes and 10 eyes with BRVO by Suzuki et al. [36], and Rispoli et al. [40], respectively, and one case report of ischemic BRVO by Kuehlewein et al. [41], found that OCTA was superior to FFA in detection of CNP areas and in delineating them with higher-resolution, and in detection of microvascular abnormalities and FAZ dimensions. It is worthy of note that the latter authors used the SS-OCTA technology in their study. In terms of versatility of the studied imaging modalities in predicting visual outcome after RVO, our findings revealed that increased CMT, DRIL, and DROL on SS-OCT, and SCP and DCP ischemia on SS-OCTA were significant predictors of poor visual outcome; whereas FFA features that included macular edema and macular ischemia did not correlate with the visual outcome of those patients. These findings are in line with Wakabayashi et al. [7], who studied the correlation between OCTA and BCVA in a series of 85 eyes with BRVO. These authors found that DCP ischemia is the most significant determinant of BCVA. Furthermore, the authors reported significant correlation between DCP ischemia on OCTA and worsening macular edema and disrupted photoreceptors layer on OCT. Our results are congruous with those of Samara et al. [8], who detected that logMAR BCVA correlated positively with foveal thickness on OCT, and negatively with vascular density in DCP on OCTA. The present study has several confounding factors, of which the most salient is its retrospective design, which allowed entry of patients with important epidemiologic and phenotypic variation of RVO. This uneven stratification might have posed an information bias that could affect our interpretation of the correlation between the tested parameters in more than one aspect. Firstly, almost 50% of recruited patients had CRVO and HRVO, which implicate more extensive DCP ischemia on SS-OCTA, more diffuse structural changes on SS-OCT and worse impact on visual acuity than BRVO. Secondly, recruited patients had different stages of RVO whether acute or chronic. Accordingly, those patients were going through different stages of microvascular remodeling following the retinal venous occlusive event with fairly wide-variation in time course and quantity. Thirdly, patients were allowed to enroll in the study regardless of whether they were treatment-naïve or had received single or multiple lines of therapy, which could have important impact on the vessel density in the SCP and DCP, and on the structural changes on SS-OCT. Finally, the current study lacked concurrent evaluation of the perfusion of the peripheral retina, which has an important impact on macular edema. Rather the study focused on grading the perfusion in the macular area and the posterior pole. A more comprehensive evaluation of RVO would include correlation between the current findings and the grade of non-perfusion in the retinal periphery using wide-field OCTA or a montage of multiple 9 × 9 mm or 12 × 12 mm scans.

The present study was able to affirm the efficacy of SS-OCTA in qualitative assessment of macular perfusion status and its ability to perform differential layer segmentation of the retinal vascular plexuses to define the extent and location of maximum ischemic insult following RVO, hence endorsing its role as more representative grader of ischemic damage in the posterior pole than conventional FFA. In addition, we were able to correlate the anatomic features on SS-OCT, vascular perfusion on SS-OCTA and visual function and to identify significant predictors on SS-OCT and SS-OCTA for visual outcome after RVO.