Visualization of preretinal neovascularization (NV) using widefield swept source optical coherence tomography angiography (SS-OCTA)

Abstract

Purpose : NV located above the vitreoretinal interface indicates proliferative diabetic retinopathy. However, in OCTA enface images it is often difficult to distinguish whether NVs have penetrated this interface. We developed a widefield SS-OCTA system that highlights NVs over a field of view of up to 90 degrees.

Methods : Our widefield SS-OCTA system scans the retina at an A-scan rate of 1.7 MHz, enabling fields of view of up to 90 degrees while maintaining Nyquist sampling density. The high scan speed keeps the acquisition time at a comfortable 15s. A deep neural network (DNN) is used for 3D denoising of the acquired data. For automated detection of NV we implemented a robust ILM segmentation algorithm. After a graph-cut search and flattening of the retina, 3D smoothing filters and morphological operations on the structural OCT data yield the segmentation of the ILM layer. The standard deviation of voxel gray values in the area above the ILM provides the neovascularization segmentation.

Results : Our prototype SS-OCTA device produced excellent image quality far out into the periphery. With this device, we have so far imaged nearly 100 subjects with diabetic retinopathy at various disease stages. Figure 1 shows a montage of two 18 mm x 18 mm scans of a healthy subject to an ultra-widefield 23 mm x 18 mm SS-OCTA image. The zoom-in in the bottom left corner highlights that even the smallest capillaries around the foveal avascular zone can still be resolved despite the large field of view. An SS-OCTA image of a diabetic patient with proliferative NV, where the automatic NV segmentation result was overlaid in red can be seen in Figure 2. It allows the ophthalmologist to distinguish between proliferative and non-proliferative vascular abnormalities at first glance.

Conclusions : Widefield SS-OCTA in combination with DNN-based image quality improvement techniques and segmentation algorithms enhances and simplifies the diagnosis and disease monitoring of preretinal NV. The occurrence, the corresponding spatial location, and the specific extent of vascular anomalies are accompanying signs of many retinal diseases. In future studies we will therefore aim at demonstrating the benefit of widefield SS-OCTA imaging beyond diabetic retinopathy.

This abstract was presented at the 2022 ARVO Annual Meeting, held in Denver, CO, May 1-4, 2022, and virtually.

 

Figure 1. Ultra-widefield SS-OCTA image.

 

Figure 2. Neovascularization (red) in a diabetic retinopathy patient overlayed on top of a SS-OCTA scan.