CORRELATION OF EN FACE OPTICAL COHERENCE TOMOGRAPHY ANGIOGRAPHY AVERAGING VERSUS SINGLE-IMAGE QUANTITATIVE MEASUREMENTS WITH RETINAL VEIN OCCLUSION VISUAL OUTCOMES

Pseudoaveraging for denoising of OCT angiography: a deep learning approach for image quality enhancement in healthy and diabetic eyes

BACKGROUND

This study aimed to develop a deep learning (DL) algorithm that enhances the quality of a single-frame enface OCTA scan to make it comparable to 4-frame averaged scan without the need for the repeated acquisitions required for averaging.

METHODS

Each of the healthy eyes and eyes from diabetic subjects that were prospectively enrolled in this cross-sectional study underwent four repeated 6 × 6 mm macular scans (PLEX Elite 9000 SS-OCT), and the repeated scans of each eye were co-registered to produce 4-frame averages. This prospective dataset of original (single-frame) enface scans and their corresponding averaged scans was divided into a training dataset and a validation dataset. In the training dataset, a DL algorithm (named pseudoaveraging) was trained using original scans as input and 4-frame averages as target. In the validation dataset, the pseudoaveraging algorithm was applied to single-frame scans to produce pseudoaveraged scans, and the single-frame and its corresponding averaged and pseudoaveraged scans were all qualitatively compared. In a separate retrospectively collected dataset of single-frame scans from eyes of diabetic subjects, the DL algorithm was applied, and the produced pseudoaveraged scan was qualitatively compared against its corresponding original.

RESULTS

This study included 39 eyes that comprised the prospective dataset (split into 5 eyes for training and 34 eyes for validating the DL algorithm), and 105 eyes that comprised the retrospective test dataset. Of the total 144 study eyes, 58% had any level of diabetic retinopathy (with and without diabetic macular edema), and the rest were from healthy eyes or eyes of diabetic subjects but without diabetic retinopathy and without macular edema. Grading results in the validation dataset showed that the pseudoaveraged enface scan ranked best in overall scan quality, background noise reduction, and visibility of microaneurysms (p < 0.05). Averaged scan ranked best for motion artifact reduction (p < 0.05). Grading results in the test dataset showed that pseudoaveraging resulted in enhanced small vessels, reduction of background noise, and motion artifact in 100%, 82%, and 98% of scans, respectively. Rates of false-positive/-negative perfusion were zero.

CONCLUSION

Pseudoaveraging is a feasible DL approach to more efficiently improve enface OCTA scan quality without introducing notable image artifacts.

Correlation of Diabetic Disease Severity to Degree of Quadrant Asymmetry in En Face OCTA Metrics

Purpose

To determine if diabetic retinopathy (DR) severity affects quadrant asymmetry (QA) of optical coherence tomography angiography (OCTA) metrics differentially.

Methods

Ninety eyes (60 patients) with no diabetes mellitus (DM) (n = 39) or varying levels of DR (n = 51) had OCTA images (3 × 3 mm, Cirrus5000) acquired five times and averaged. The vessel length density (VLD) and perfusion density (PD) of the superficial retinal layer (SRL) and deep retinal layer (DRL) were measured. QA was defined as the maximum minus minimum value among four parafoveal Early Treatment Diabetic Retinopathy quadrants, and compared with DR severity by linear regression including fixed effects for each individual and eye.

Results

The mean patient age was 55.5 years (range, 24–88 years) and 60% were male. Comparing severe nonproliferative DR or proliferative DR versus no DM/DR eyes, QA was significantly higher for SRL VLD, and PD (+0.67 ± 0.16 and +0.014 ± 0.003; P < 0.001) and DRL VLD, and PD (+1.25 ± 0.16 and +0.032 ± 0.003; P < 0.001). When comparing mild or moderate nonproliferative DR versus no DM/DR, the DRL VLD, and PD were significantly higher (+0.51 ± 0.13 and +0.015 ± 0.003; P < 0.001). For every step increase in DR severity, there was a +0.20 QA for SRL VLD, +0.004 SRL PD, +0.33 DRL VLD and +0.009 DRL PD (P < 0.001). Regression analysis comparing intraquadrant effect on DR severity demonstrated that the superior quadrant was most affected for all OCTA metrics.

Conclusions

DR severity affects VLD and PD more asymmetrically across Early Treatment Diabetic Retinopathy quadrants with a linear increase in QA for each worsening level of DR. Individual intraeye metrics such as QA can accurately quantify DR severity without concerns for intereye variabilities that could affect the reproducibility and reliability of OCTA quantification.

Effect of caffeine on superficial retinal vasculature of the macula in high myopes using optical coherence tomography angiography - A pilot study

PURPOSE

To monitor effect of caffeine on vasculature of the inner retina of high myopes METHODS: This was a crossover, self-control, randomized trial. Healthy young high myopes were recruited to take 200 mg of caffeine capsule and placebo capsule, randomly assigned in two visits separated by at least one week. Superficial retinal vasculature in terms of vessel length density (VD) and perfusion area density (PD) was captured and monitored using a spectral domain optical coherence tomography angiography (OCTA) machine. After baseline measurements, blood pressure (BP), intraocular pressure (IOP), and subfoveal choroidal thickness (ChT) were also monitored at 30-min intervals till 3 h.

RESULTS

Eighteen subjects (6 male, 24.3 ± 3.1 years) completed the study. After taking the caffeine capsule, there was a significant increase in BP (p < 0.01), and reduction in ChT (p < 0.01), with no change in IOP (p = 0.36). VD demonstrated a trend of reduction at the central 1-mm circle, and 1-3 mm annulus (p < 0.01) following the ETDRS grid. Reduction trend of PD appeared at the central 1-mm circle, 1-3 mm annulus, and the entire 3-mm circle (p < 0.01). Compared with baseline, VD and PD reductions were significant 180 min after taking the caffeine capsule at the central 1-mm circle, but the reduction was small (VD: by 1mm^-1; PD: by 1%). Changes in other regions were not significant.

CONCLUSIONS

The current study found significant reduction in VD and PD after taking 200 mg of caffeine capsule. Such a small amount of alteration may be clinically irrelevant.

Effects of Induced Astigmatism on Spectral Domain-OCT Angiography Quantitative Metrics

PURPOSE

To analyze the effect of induced astigmatism on en-face spectral-domain optical coherence tomography angiography quantitative metrics.

DESIGN

Prospective crossover study.

METHODS

Normal eyes without astigmatism and with 0.75, 1.75, and 2.75 diopters (D) of with-the-rule (WTR) astigmatism were imaged using a 3 × 3-mm scan pattern SD-OCTA CIRRUS 5000 HD-OCT with AngioPlex (Carl Zeiss Meditec, Dublin, CA, USA). Quantitative parameters, including foveal avascular zone metrics, parafoveal vessel length density (VD), and perfusion density (PD) were corrected for magnification secondary to axial length and analyzed. Univariate linear regressions were performed within each eye to correlate quantitative metrics to the level of an induced astigmatic cylinder.

RESULTS

Fifteen eyes from 15 patients were imaged. Every 1-D increase in induced WTR astigmatism was associated with a statistically significant decrease in VD and PD within all Early Treatment Diabetic Retinopathy Study inner ring quadrants; however, especially more so nasally (VD: 0.63; P < .001; PD: 0.0089; P = .001). For every 1-D increase in induced astigmatism, the resulting decrease in the inner ring superior quadrant was 12% greater for VD and 16% greater for PD versus that in the inferior quadrant. The resulting decrease in the inner ring nasal quadrant was 40% greater for VD and 48% greater for PD versus that in the temporal quadrant.

CONCLUSIONS

Increasing levels of induced WTR astigmatism correlated with globally diminishing VD and PD, was more symmetrical for vertical than horizontal quadrants, and was most pronounced nasally. This may be due to a high prevalence of horizontally oriented vessels nasally and the horizontal optical defocus induced by WTR astigmatism.

Correlation of Quantitative Measurements with Diabetic Disease Severity Using Multiple En Face OCT Angiography Image Averaging

PURPOSE

To evaluate the effect of averaging en face OCT angiography (OCTA) images on quantitative measurements of the retinal microvasculature and their correlation to diabetic retinopathy (DR) disease severity.

DESIGN

Cross-sectional cohort study.

PARTICIPANTS

One hundred five eyes (65 patients) comprising 28 eyes from 19 healthy, aged-matched control participants, 14 eyes from 9 diabetics without DR, and 63 eyes from 37 diabetics with varying levels of DR.

METHODS

Spectral-domain CIRRUS 5000 (Carl Zeiss Meditec, Dublin, CA) OCTA images with no macular edema or significant motion artifact were acquired 5 times with the 3 × 3-mm scan pattern. En face images of the superficial retinal layer (SRL) and deep retinal layer were registered and averaged. Vessel length density (VLD), perfusion density (PD), and foveal avascular zone (FAZ) parameters were measured on averaged versus single OCTA images.

MAIN OUTCOME MEASURES

Univariate and multivariate linear regression correlated quantitative metrics to DR severity and best-corrected visual acuity (BCVA).

RESULTS

Eighty-four eyes (55 patients) met the inclusion criteria. Almost uniformly, lower VLD and PD parameters were associated significantly with worse DR severity and BCVA. Multivariate linear regression for DR severity resulted in an R² value of 0.82 and 0.77 for single and averaged groups, respectively. No variables remained associated significantly with DR severity in multivariate analysis with single images, but in averaged images, increased superior SRL PD significantly predicted worse DR severity (coefficient, 52.7; P = 0.026). Multivariate linear regression for BCVA had an R² value of 0.42 and 0.47 for single and averaged groups, respectively. Foveal avascular zone size was not associated with DR severity when single OCTA images (P = 0.98) were considered, but was highly associated when using averaged images (coefficient, 6.18; P < 0.001). Foveal avascular zone size was predictive for logarithm of the minimum angle of resolution BCVA with averaged images (0.21; P = 0.004), but not with single images (P = 0.31).

CONCLUSIONS

Averaging of en face OCTA images improves the visualization of capillaries, particularly increasing the clarity of the FAZ borders, and therefore improves the correlation of vessel density and FAZ-specific parameters to DR severity and BCVA.

Multiple scan averaging to yield accurate quantitative analysis of optical coherence tomography angiograms

Optical coherence tomography angiography (OCTA) is widely used in ophthalmic practice. Most OCTA studies based their findings on a single OCTA measurement. We conducted an observational study of 82 eyes from 82 healthy subjects to compare variations of OCTA parameters among five successive measurements. A 3 × 3 mm Early Treatment of Diabetic Retinopathy Study grid centred at fovea was used. An average from five successive OCTA measurements (both perfusion density and vessel density) was calculated to be used as the reference standard. There was no significant difference in perfusion and vessel densities among five successive OCTA measurements, and from different levels of averaging. Perfusion density was close to the reference standard when average from three measurements was used (discrepancy within 1.5%) as compared with using just one measurement (discrepancy from 3.2% to 4.5%). Vessel density was also close to reference standard when average from three measurements was used (within 0.8 mm⁻¹) as compared with using just one measurement (2 mm⁻¹). Software feature that allows OCTA devices to average quantitative parameters for analysis will be useful.

Impact of integrated multiple image averaging on OCT angiography image quality and quantitative parameters

PURPOSE

Multiple image averaging (MIA) is a new approach to improve OCT angiography (OCTA) imaging. The aim of this work was to analyze the impact of MIA on image quality and quantitative OCTA parameters.

METHODS

Twenty eyes from 20 healthy volunteers (55.65 ± 14.8 years) were prospectively enrolled. Imaging was performed using two commercially available OCTA devices (Canon OCT HS-100, Optovue AngioVue) using a uniform imaging protocol. Each participant had two single scans of the macula (3 × 3mm, Canon and Optovue) as well as five continuous single scan imaging procedures (3 × 3mm each) using the Canon device. Three out of five of these images with highest quality were manually chosen and then automatically processed by the Canon device using MIA. The superficial retinal plexus of the single scans and of MIA images was analyzed with regard to the device' own image quality scores (IQS), peak signal-to-noise ratio (PSNR), the size of the foveolar avascular zone (FAZ), and vessel density (VD). Image acquisition times were recorded. Parameters were compared between the devices and the different imaging protocols.

RESULTS

Average acquisition time was significantly higher for the MIA compared with the single measurements (29.09 ± 10.19 seconds (s) (MIA) vs. 5.56 ± 2.17 s (Canon single scan) vs. 20.28 ± 6.81 s (Optovue) (p < 0.001)). IQS showed no significant differences between the devices and between the recording protocols. PSNR was 12.38 ± 0.20 (Canon single scan), 13.01 ± 0.36 (Canon MIA), and 14.34 ± 0.60 (Optovue) (p < 0.001 between the groups). Mean FAZ area in Canon single scans was 0.29 ± 0.06 mm², 0.27 ± 0.07 mm² using MIA, and 0.27 ± 0.08 mm² using the Optovue device. There was no significant difference between mean FAZ measurements before and after averaging (Canon single scan vs. MIA, p = 0.168). VD of the parafoveal area using MIA was significantly lower compared with both single scans (p < 0.001).

CONCLUSIONS

MIA can improve PSNR, but it also reduces imaging speed and significantly affects VD measurements. Therefore, when comparing OCTA data, the use of uniform imaging protocols is required.

A Novel Automatic Method to Estimate Visual Acuity and Analyze the Retinal Vasculature in Retinal Vein Occlusion Using Swept Source Optical Coherence Tomography Angiography

The assessment of vascular biomarkers and their correlation with visual acuity is one of the most important issues in the diagnosis and follow-up of retinal vein occlusions (RVOs). The high workloads of clinical practice make it necessary to have a fast, objective, and automatic method to analyze image features and correlate them with visual function. The aim of this study is to propose a fully automatic system which is capable of estimating visual acuity (VA) in RVO eyes, based only on information obtained from macular optical coherence tomography angiography (OCTA) images. We also propose an automatic methodology to rapidly measure the foveal avascular zone (FAZ) area and the vascular density (VD) in the superficial and deep capillary plexuses in swept-source OCTA images centered on the fovea. The proposed methodology is validated using a representative sample of 133 visits of 50 RVO patients. Our methodology estimates VA with very high precision and is even more accurate when we integrate depth information, providing a high correlation index of 0.869 with the real VA, which outperforms the correlation index of 0.855 obtained when estimating VA from the data obtained by the semiautomatic existing method. In conclusion, the proposed method is the first computational system able to estimate VA in RVO, with the additional benefits of being automatic, less time-consuming, objective and more accurate. Furthermore, the proposed method is able to integrate depth information, a feature which is lacking in the existing method.