Fig. 5

Representative comparison between four eyes within different DR severity classifications, where each column is an eye with different DR classification. First row: averaged OCTA images of the superficial vascular complex (SVC). Second row: averaged OCTA images of the deep vascular complex (DVC). Third row: pixel intensity coefficient of variance (PICoV) maps of the SVC. Fourth row: PICoV maps of the DVC. High-quality averaged images are produced via image registration of 10 serially acquired 6 × 6 mm foveal-centered OCTA images and pixel-wise averaging. PICoV maps are generated by masking a Deep Neural Network vessel segmentation over the registered image stack to reveal areas of perfusion heterogeneity from blue (no variation) to red (most variation)