基于ReSE-AG-Net的青光眼视盘视杯分割方法  

ReSE-AG-Net Based Optic Disk Optic Cup Segmentation for Glaucoma

作　　者：李豪杰 郭子洋 刘伟志 罗益 曹秒[1] LI Haojie;GUO Ziyang;LIU Weizhi;LUO Yi;CAO Miao(School of Life Science and Technology,Changchun University of Science and Technology,Changchun 130022;State Grid Jilin Eletric Power Co.,Ltd,Changchun 130022)

机构地区：[1]长春理工大学生命科学与技术学院,长春130022 [2]国网吉林电力有限公司超高压公司,长春130022

出　　处：《长春理工大学学报（自然科学版）》2024年第4期115-121,共7页Journal of Changchun University of Science and Technology(Natural Science Edition)

基　　金：吉林省科技厅项目(20240401073YY)。

摘　　要：分割视盘视杯是诊断青光眼的主要依据,因此提出了一种基于改进U-Net网络的青光眼视盘视杯分割方法,称为ReSE-AG-Net。并将ReSE-AG-Net、U-Net、M-Net、CE-Net和CDED-Net分别在Drishti-GS1数据集和RIM-ONE-R3数据集上进行测试,对得到分割后的视盘视杯的Dice系数和IoU进行对比。实验结果表明,在Drishti-GS1数据集上,分割视盘的Dice系数和IoU分别是0.989 1和0.959 0,分割视杯的Dice系数和IoU分别是0.931 0和0.871 0,在RIMONE-R3数据集上,分割视盘的Dice系数和IoU分别是0.978 9和0.958 7,分割视杯的Dice系数和IoU分别是0.863 7和0.760 1,对比U-Net模型和现有的分割模型均有一定的提升,证明了该方法的可行性。Segmentation of optic disk optic cup is the main basis for diagnosing glaucoma.Therefore a glaucoma optic disk optic cup segmentation method based on improved U-Net network called ReSE-AG-Net is proposed.ReSE-AG-Net,U-Net,M-Net,CE-Net and CDED-Net are tested on Drishti-GS1 dataset and RIM-ONE-R3 dataset respectively,and the segmented optic disc optic cups obtained are tested and compared.Dice coefficients and IoU of optic disk optic cups are compared.The experimental results show that the Dice coefficient and IoU of the segmented optic disk are 0.989 1 and0.959 0,and the Dice coefficient and IoU of the segmented videocup are 0.931 0 and 0.871 0 on Drishti-GS1 dataset,and the Dice coefficient and IoU of the segmented optic disk on RIM-ONE-R3 dataset are 0.978 9 and 0.958 7,and the Dice coefficient and IoU for segmenting the optic cup are 0.863 7 and 0.760 1,respectively,which compare the U-Net model and the existing segmentation models,proving the feasibility of the method.

关 键 词：青光眼 视盘视杯分割 U-Net ReSE-AG-Net 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]