Foundation of graph-based biomarkers from OCTA B-scan

This paper introduces what we believe to be novel quantifiers for monitoring macular vascularity (MV) using optical coherence tomography angiography (OCTA). The study refines a recently developed image-processing approach to convert the OCTA cube into graphical representations, termed OCTA-graph. It then investigates the characteristics of the OCTA-graphs in small and large fields-of-view (FOV) across healthy and MV conditions. A set of five inventive OCTA-graph-based biomarkers is calculated: the area under the graphical representation of MV, the area under the graphical representation of retinal and choroidal blood vessels, and the OCTA-graph’s peaks. A total of 393 OCTA B-scans were encoded, including FOV of 3 mm (n = 189) and 6 mm (n = 204). The dataset included healthy controls (n = 251) and five MV conditions (n = 142). In healthy eyes, the resulting OCTA-graphs displayed two distinct MV graph patterns, representing the retinal and choroidal vessels. In pathological cases, alterations in the shape and amplitude of OCTA-graph patterns were apparent. The proposed OCTA-graph-based biomarkers showed statistically significant differences between normal and abnormal conditions (p-value<0.01). Violin plots showed susceptibility to magnitude drifts in OCTA-graph patterns related to disease status, particularly in diabetic retinopathy. Taken together, OCTA-graph and its derived biomarkers offer innovative perspectives for monitoring macular vasculature density in OCTA.