分类法定量评估频谱光学相干层析图像杯盘比  

Classification Based Quantitative Assessment of Cup-to-Disk Ratio in Spectral Domain Optical Coherence Tomography Images

作　　者：王玉萍[1] 陈强[1] 陆圣陶 牛四杰[1] 沈宏烈 

机构地区：[1]南京理工大学计算机科学与工程学院,南京210094

出　　处：《计算机辅助设计与图形学学报》2015年第8期1539-1547,共9页Journal of Computer-Aided Design & Computer Graphics

基　　金：中央高校基本科研业务费专项资金(30920140111004);江苏省"青蓝工程"(D040201);江苏省"六大人才高峰"(2014-SWYY-024)

摘　　要：在青光眼检测中,基于频谱光学相干层析图像较眼底图像具有更高的精确度和可靠性,为了克服眼底图像评估杯盘比精度不高的缺陷,提出一种基于频谱光学相干层析图像进行视网膜色素上皮层断点检测的杯盘比评估方法.首先利用主成分分析和支持向量机将频谱光学相干层析图像与其投影图像限定断点搜索区域,并对限定的图像区域进行断点识别;然后借助于类别标签矩阵对断点进行纠正;最后根据视杯视盘与断点的关系测量杯盘比.实验结果表明,该方法在准确检测视网膜色素上皮层断点、提高杯盘比测量精度方面效果显著.In the process of glaucoma detection, spectral domain optical coherence tomography（SD-OCT） images are better than fundus images in aspects of precision and reliability. To overcome the defects of low accuracy in cup-to-disk ratio assessment based on fundus images, this paper presents a novel method to detect retinal pigment epithelium（RPE） breakpoints in SD-OCT images and to effectively calculate the cup-to-disk ratio. Firstly, we train a classifier with principal component analysis and support vector machine, and then constrain an area where the breakpoints of RPE are searched using SD-OCT and project images. After recognition, the breakpoints are corrected with a label matrix. Finally, the cup-to-disk ratio is calculated based on the locations of the breakpoints. Experimental results demonstrate that the proposed method is effective to detect RPE breakpoints and calculate the cup-to-disk ratio.

关 键 词：青光眼 频谱光学相干层析图像 杯盘比 视网膜色素上皮层断点 主成分分析 支持向量机 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]