Abstract
The study used spectral domain optical coherence tomography (SD-OCT) and full width at half maximum image segmentation to investigate the morphological changes of retinal blood vessels in patients with diabetic retinopathy (DR).Seventy-five patients with type 2 diabetes mellitus (DM) without DR and 65 patients with DR were studied. The vascular images of superior temporal region B of the retina were obtained by SD-OCT. The edges of retinal vessels were identified by the full-width-at-half-maximum image segmentation method. The lumen diameter, wall thickness (WT), wall cross-sectional area (WCSA), and wall-to-lumen ratio (WLR) were investigated.We found that compared with no-diabetic-retinopathy (NDR) group, patients in DR group had increased retinal arteriolar lumen diameter (RALD), retinal arteriolar outer diameter (RAOD), and WT(128.80 µm vs. 104.88 µm; 147.01 µm vs. 135.60 µm; 18.29 µm vs. 15.26 µm; P < 0.05 respectively). And, the retinal venular lumen diameter, retinal venular outer diameter, and venular WT in the DR group also increased (146.17 µm vs. 133.66 µm; 180.20 µm vs. 156.43 µm; 17.01 µm vs. 11.38 µm; P < 0.05 respectively). The morphological changes of retinal vessels were significantly correlated with DR stage.In conclusion,in diabetic patients with DR, both retinal arteries and veins are widened with increased vascular thickness.
Active Learning with Bayesian UNet for Efficient Semantic Image Segmentation
