Retinal blood vessels segmentation in fundus images based on modified U-Net
Paper ID : 1107-MVIP2020
Masoome Qolami *, Ali Aghagolzadeh, Mahdi Ezoji
The retinal vessels Segmentation in fundus images is crucial in the early diagnosis of ocular diseases such as diabetic retinopathy and hypertension. In this paper, an automated approach based on convolutional neural networks is proposed which is a modified version of the U-Net architecture. we achieved 96/70% and 96/09% accuracy on the DRIVE and STARE dataset, respectively. Moreover, the proposed modified U-Net has led to a significant improvement in AUC.
Retinal blood vessels, Segmentation, deep learning, U-Net
Status : Paper Accepted (Oral Presentation)