| Abstrakt: |
Diabetes mellitus is a metabolic disorder that causes a variety of vascular issues in the body. When the condition coexists with other general problems (high blood pressure, obesity, excessive cholesterol), the risk of ocular complications increases. Diabetes can damage the small blood vessels in the retina. This is referred to as diabetic retinopathy (DR). Therefore, the segmentation of the vascular network may assist in the automatic and early recognition and screening of DR. Since, manual vessel extraction is time-consuming, automation of this procedure is critical. In this paper, we focus on the implementation of different deep learning networks for accurate retinal vasculature semantic segmentation. We used three distinct models: SegNet, U-Net, and the convolutional neural network (CNN). These approaches were evaluated on three publicly available annotated databases: DRIVE, HRF, and CHASE-DB1. To compare the three proposed models with the previous methodologies, various metrics are calculated. The produced results were good and encouraging in terms of the quality of the visual diagnosis. [ABSTRACT FROM AUTHOR] |
