摘要研究基于频域光学相干断层图像（Spectral-Domain Optical Coherence Tomography，SD-OCT）的地图状萎缩（Geographic Atrophy，GA）的定量分析，对于简化年龄相关性黄斑变性（age-related macular degeneration，AMD）疾病的诊断具有重要意义。近年来，使用频域光学相干断层扫描得到的眼部图像，对年龄相关性黄斑变性（AMD）视网膜病症进行诊断越来越趋于成熟，人们不断的追求新颖实用的方法来自动分割其视网膜色素上皮细胞层(retinal pigment epithelium ，RPE)和GA，这也是近些年医学图像分割的研究重点。目前，基于SD-OCT图像的GA定量分析主要有基于图像的梯度、灰度信息，活动轮廓和动态规划的方法及图论思想的分割算法。本次研究，主要是针对GA的特点，采用Matlab计算软件，使用局部灰度极值的方法对RPE进行分割，计算出RPE的具体位置，并相应位置进行标注分割。在得到RPE层的基础上，对所有图像进行投影。处理在投影图像中根据灰度的不同，分割出GA病变并再次投影，对源图像进行GA病变的标注。使用现有方法对SD-OCT的GA图像进行分割基础上，针对现有的方法的不足，提出相应的改进措施。63870

毕业论文关键词  地图状萎缩；黄斑变性；图像分割；灰度极值；投影

毕业设计说明书（论文）外文摘要

Title    Research on the quantitative analysis of GA     based on spectral-domain optical coherence tomography image   

Abstract Research on the quantitative analysis of Geographic Atrophy (GA) based on spectral-domain optical coherence tomography (SD-OCT) images is of much importance to simplify the diagnosis of diseases named age-related macular degeneration (AMD). Lately, it is becoming more and more mature to take use of SD-OCT to diagnose AMD. People are eager to find novel and practical methods to segment retinal pigment epithelium (RPE) and GA, which is the emphasis of medical image segmentation around these years. Currently, the automatic GA segmentation methods using SD-OCT are mainly based on the gradient and intensity information of images, active contour, dynamic programming method and graph theory to extract the target contours. In this research, image projections are done to form a new image using the RPE layers segmented with GA features by matlab. The intensity differences of the projection can help segment GA area. With another projection, the GA area in the source images can be spotted. In the meantime, some improvements or supplements are provided.

Keywords：Geographic Atrophy； Age-related macular degeneration； Image segmentation； gray extreme； projection

目   录

1 绪论 1

1.1 AMD的研究背景 1

1.2 RPE以及GA研究的现状 1

1.3 GA分割的意义 2

2 OCT眼科图像的详解 3

2.1光学相干断层扫描（Optical coherence tomography，OCT） 3

2.2 基于SD-OCT的眼睛图像 3

3 基于图像的灰度和梯度进行分割的相关基本知识 5

3.1灰度与灰度图 5

3.2二值图像与图像的二值化 5

3.3常用的几种阈值计算方法 6

4 使用灰度极值法对RPE进行分割 8

4.1由去噪后的灰度图得到二值图像，并分割出ILM层 8

4.2通过灰度极值，得到RPE层的实际位置