Segmentation of corneal endothelium images using a U-Net-based convolutional neural network
作者：
Highlights：
• A fully automatic method for segmentation of the corneal endothelium images is proposed.
• The core of the method is a U-Net-based convolutional neural network.
• Tests on a dataset of 30 corneal endothelium specular microscopy images were performed.
• The obtained performance of cell edge discrimination is at the level of 0.92 AUROC.
• The errors of the cell number and the cell size are both below 6.5%.
摘要
•A fully automatic method for segmentation of the corneal endothelium images is proposed.•The core of the method is a U-Net-based convolutional neural network.•Tests on a dataset of 30 corneal endothelium specular microscopy images were performed.•The obtained performance of cell edge discrimination is at the level of 0.92 AUROC.•The errors of the cell number and the cell size are both below 6.5%.
论文关键词：Corneal endothelial cells,Image segmentation,Convolutional neural network,U-Net
论文评审过程：Received 23 January 2018, Revised 5 April 2018, Accepted 10 April 2018, Available online 19 April 2018, Version of Record 7 June 2018.
论文官网地址：https://doi.org/10.1016/j.artmed.2018.04.004
