Opposing end-node and junction-node remodeling patterns in diabetic patients on optical coherence tomography angiography

This cross-sectional study aimed to investigate macular vascular network topology in diabetic patients. We analyzed 3 × 3 mm OCTA images from 235 eyes of 146 participants, divided into controls, diabetic patients without diabetic retinopathy (NoDR), and patients with nonproliferative diabetic retinopathy (NPDR). Skeletonized vascular maps were generated for the superficial (SCP) and deep capillary plexus (DCP). Vessel skeleton density (VSD), End Node (EN) density, Junction Node (JN) density, and Fractal Dimension (FD) were calculated in foveal, parafoveal, and total regions. Statistical comparisons were performed using linear mixed-effects models. EN density exhibited an inverted-U-shaped pattern across groups in both plexuses. VSD, JN density, and FD differed between groups, with lower values in NoDR and NPDR compared to controls. Discrimination analyses indicated that VSD performed better in differentiating NPDR from controls and NoDR, whereas JN density and FD were comparable for differentiating NoDR from controls. After adjusting for VSD across study groups, EN density in both layers and JN density in the SCP remained significant. Vascular metrics correlated with diabetes duration and glycemic level. EN and JN density emerge as promising biomarkers alongside VSD and FD for characterizing microvascular alterations in diabetic eyes, as topology-based parameters may provide information independent of vessel density.