Retinal vessel segmentation based on neural network application

Automated segmentation of retinal vessels plays an important role in diagnosing diabetic diseases. In this paper, I propose the convolutional neural network (U-Net) to yield more precise segmentations of retinal vessels from various fundus images. The performance of this neural network was first tes...

Full description

Saved in:
Bibliographic Details
Main Author: Tan, Chin Guan
Other Authors: Jiang Xudong
Format: Final Year Project
Language:English
Published: Nanyang Technological University 2020
Subjects:
Online Access:https://hdl.handle.net/10356/138852
Tags: Add Tag
No Tags, Be the first to tag this record!
Institution: Nanyang Technological University
Language: English
Description
Summary:Automated segmentation of retinal vessels plays an important role in diagnosing diabetic diseases. In this paper, I propose the convolutional neural network (U-Net) to yield more precise segmentations of retinal vessels from various fundus images. The performance of this neural network was first tested on the DRIVE database, and it achieved a relatively high score in terms of area under the Receiver Operating Characteristic (ROC) curve, with an astounding result of 0.9790, in comparison to the other existing methods published. On the STARE database, this method also yield satisfying results. This shows that the U-net architecture is a very effective and efficient model to aid in the early diagnosis of diseases.