Large vessel removal in quantification of vascular density in screening for diabetic retinopathy in optical coherence tomography angiography (OCTA) en face projections

Abstract

Purpose : In microvascular conditions such as diabetic retinopathy, the smaller vessels are lost early in the disease. Therefore, eliminating the larger vessels from the calculation of perfusion density may improve the sensitivity of vascular quantification to capillary loss. Moreover, if the caliber of the larger vessels is subject to significant individual variation, removing them from the quantification of perfusion density might result in a more accurate assessment of the severity of the disease.
This study is a comparison of the results from a vascular quantification algorithm applied to a set of 6x6mm OCTA superficial en face projections of normal eyes and eyes exhibiting diabetic retinopathy (DR), with and without masking the large vessels in each scan.

Methods : One scan from each of 19 eyes of patients with diabetic retinopathy and 20 normal eyes was acquired with a PLEX^® Elite 9000 SSOCT system (ZEISS, Dublin, CA). Superficial OCTA en face projections were generated for each scan. An automated algorithm that identifies and masks larger vessels was used to generate a large-vessel mask for each scan.
A vascular density algorithm was then used to calculate the average perfusion density (area density of perfused area, dimensionless) for a 6mm diameter circular area centered on the fovea, both with and without eliminating the large vessels from the area being averaged.
The resulting distributions for DR and healthy datasets were analyzed, and Welch’s t-statistic for heteroscedastic data was calculated.

Results : The plots show a comparison between the distributions of perfusion density with and without masking the large vessels. The statistical significance of the difference between the means of the perfusion density of the disease vs. normal was increased after masking the larger vessels.

Conclusions : Removing the larger vessels from en face OCTA projections may help in the automated identification and classification of microvascular ocular disease.

This is a 2020 ARVO Annual Meeting abstract.

 

Distribution of vessel densities in healthy and DR cases with and without masking the larger vessels

 

Left, right: A superficial OCTA slab, and the slab after masking the larger vessels.