Abstract
Purpose :
Differential artery-vein analysis, which promises better disease detection and classification, is not available with existing clinical optical coherence tomography angiography (OCTA) instruments. This study is to test the feasibility of using En-face optical coherence tomography (OCT) intensity and profile features to guide artery-vein differentiation in OCTA.
Methods :
Figure 1 illustrates key procedures of the proposed En-face OCT intensity profile analysis guided artery-vein classification in OCTA. Localized intensity normalization, bottom hat filtering and edge enhancing techniques were implemented to segment vessels in OCT (Fig. 1f) and OCTA (Fig. 1h). Vessel source nodes along the boundary in the segmented OCT vessel map were selected, and multi-feature analysis was employed to classify them into arteries and veins. For each vessel segment in OCT, we obtained intensity profiles and measured five features: ratio of width to central reflex, brightness, distribution of intensity, mean and median of the profile. After the feature extraction, we used a clustering algorithm to classify each vessel as an artery or vein. Based on the classified source nodes (marked blue and red in Fig 1f), a blood vessel tracking algorithm was used to classify the whole vessel map (Fig. 1g). The OCT artery-vein map was overlaid with OCTA vessel map to guide the artery-vein classification in OCTA (Fig. 1i). Additional capillary branches were further classified using our recently demonstrated vessel tracking and morphological operations to obtain the final OCTA artery-vein map (Fig. 1j). All classification results were validated using manual labeling from two independent observers.
Results :
En-face OCT feature analysis was used to guide artery-vein classification in OCTA. The performance of artery-vein classification is summarized in Table 1. The algorithm was tested on 100 OCT/OCTA images from 50 patients at the UIC Retina clinic. As shown in Table 1, 96.81% and 96.33% accuracies were achieved in identifying artery and vein in En-face OCT images. The accuracies were 96.77% and 96.25% respectively, for identifying blood vessels as artery and vein in the OCTA images.
Conclusions :
En-face OCT image guided artery-vein classification provides a feasible method to differentiate arteries and veins in OCTA.
This abstract was presented at the 2019 ARVO Annual Meeting, held in Vancouver, Canada, April 28 - May 2, 2019.