• Marada Amrutha Reddy
• Dr. Anuradha Padala

Digital photography provides prominent diagnostic information used to identify and treat diseases. Fundus imaging is performed to acquire the two-dimensional representation of an eye's optic nerve, retinal tissue, and retinal vessels. The imaging captures retinal vessels, which are used to diagnose ophthalmic diseases. Retinal vessel segmentation helps visualize and identify the abnormalities in vessels. Proper segmentation helps make an accurate diagnosis and early surgical decisions and avoids the risks of blindness. The tiny and curved lines of vessels make the process of segmentation challenging. Although different state-of-art segmentation models are built using a sequence of encoder-decoder deep learning techniques, they fuse semantically dissimilar feature maps. The skip connections in UNet provide feature maps. The thinner vessels show a low contrast against the background. This research aims to improve the contrast between retinal vessels for better segmentation. A modified architecture using UNet is used to precisely localize the vessels. Dense blocks replace the convolutional blocks. The encoder-decoder ladder is bridged using a residual block. It is observed that the proposed architecture improves sensitivity, reaching a value of 95.7 percent.

Image segmentation; Deep Learning; Retinal vessel segmentation; Unet; Dense layer; Residual Block

"Dense Residual Bridge UNet for Retinal Vessel Extraction", International Journal of Emerging Technologies and Innovative Research (www.jetir.org), ISSN:2349-5162, Vol.9, Issue 5, page no.b615-b623, May-2022, Available :http://www.jetir.org/papers/JETIR2205186.pdf

"Dense Residual Bridge UNet for Retinal Vessel Extraction", International Journal of Emerging Technologies and Innovative Research (www.jetir.org | UGC and issn Approved), ISSN:2349-5162, Vol.9, Issue 5, page no. ppb615-b623, May-2022, Available at : http://www.jetir.org/papers/JETIR2205186.pdf