English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Zendy Plus

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Zendy Tools

Zendy Open

Accurate segmentation of retinal vessels is a basic step in Diabeticretinopathy(DR) detection. Most methods based on deep convolutional neuralnetwork (DCNN) have small receptive fields, and hence they are unable tocapture global context information of larger regions, with difficult toidentify lesions. The final segmented retina vessels contain more noise withlow classification accuracy. Therefore, in this paper, we propose a DCNNstructure named as D-Net. In the proposed D-Net, the dilation convolution isused in the backbone network to obtain a larger receptive field without losingspatial resolution, so as to reduce the loss of feature information and toreduce the difficulty of tiny thin vessels segmentation. The large receptivefield can better distinguished between the lesion area and the blood vesselarea. In the proposed Multi-Scale Information Fusion module (MSIF), parallelconvolution layers with different dilation rates are used, so that the modelcan obtain more dense feature information and better capture retinal vesselinformation of different sizes. In the decoding module, the skip layerconnection is used to propagate context information to higher resolutionlayers, so as to prevent low-level information from passing the entire networkstructure. Finally, our method was verified on DRIVE, STARE and CHASE dataset.The experimental results show that our network structure outperforms somestate-of-art method, such as N4-fields, U-Net, and DRIU in terms of accuracy,sensitivity, specificity, and AUCROC. Particularly, D-Net outperforms U-Net by1.04%, 1.23% and 2.79% in DRIVE, STARE, and CHASE three dataset, respectively.

Retinal Vessels Segmentation Based on Dilated Multi-Scale Convolutional Neural Network