面向青光眼筛查的眼底影像杯盘形态自动分析方法

Glaucoma is the first cause of irreversible blindness in China. Computer-aided screening has become the significant demand for the prevention of glaucoma. Examination of the optic nerve to look for changes in the cup-disc appearance from ophthalmoscopic images is the key issue for glaucoma screening. However, most of existing approaches for cup-disc analysis are focused on measure of the optic nerve, whereas there is lack of methods to recognize changes of cup-disc appearance. It is harmful to improve accuracy of glaucoma screening. To address this issue, this project will study the automatic morphological analysis method of cup-disc appearance, which consists of cup-disc region segmentation, cup-disc morphological character representation and cup-disc pattern recognition. Firstly, a 2D-3D image registration method will be studied to align the information of optic cup and disc for cup-disc analysis which obtained from both color fundus images and SD-OCT images. Secondly, the cup-disc segmentation approach will be proposed based on feature fusion with the help of multi-view learning. And the representation approach for appearance and morphological character of cup-disc is also proposed with geometry modeling and statistical analysis methods based on the segmented cup-disc region and contours. Finally, a cup-disc appearance recognition scheme will be designed by means of multi-classifier decision. The achievements of this project will establish an effective automatic morphological analysis scheme of cup-disc appearance, which can improve the sensitivity and specificity of computer-aided glaucoma screening, and recognize changes of cup-disc appearance caused by glaucoma at the early age.

青光眼是我国最常见不可逆致盲眼病，计算机辅助筛查已成为青光眼防盲的重大需求。发现和识别眼底影像中的视杯视盘形态改变是青光眼筛查的关键。然而，现有筛查方法多以参数测量为主，无法识别杯盘形态改变，严重影响了青光眼筛查的准确性。针对该问题，本项目拟在眼底影像配准的基础上，综合利用多模态影像信息，围绕杯盘区域分割、杯盘特征表示和形态识别等问题，研究杯盘形态分析方法。首先研究眼底影像的2D-3D配准方法，将彩色眼底图和SD-OCT影像携带的杯盘信息对齐到同一坐标空间；然后研究基于特征融合的杯盘区域分割方法，准确获得杯盘区域；基于几何建模和统计分析技术，研究杯盘形态的特征表示方法，有效描述杯盘形态改变；最后基于多分类器决策融合技术，研究青光眼杯盘形态识别方法，提出基于此的青光眼筛查模型。本项目将建立一种有效的眼底影像杯盘形态分析方法，提高计算机辅助青光眼筛查的敏感性和特异性，满足青光眼筛查的迫切需求。