《A computer-aided healthcare system for cataract classification and grading based on fundus image analysis》学习笔记

Abstract

       This paper presents a fundus image analysis based computer aided system for automatic classification and grading of cataract, which provides great potentials to reduce the burden of well-experienced ophthalmologists (the scarce resources) and help cataract patients in under-developed areas to know timely their cataract conditions and obtain treatment suggestions from doctors. The system is composed of fundus image pre-processing, image feature extraction, and automatic cataract classification and grading. The wavelet transform and the sketch based methods are investigated to extract from fundus image the features suitable for cataract classification and grading. After feature extraction, a multiclass discriminant analysis algorithm is used for cataract classification, including two-class (cataract or non-cataract) classification and cataract grading in mild, moderate, and severe. A real-world dataset, including fundus image samples with mild, moderate, and severe cataract, is used for training and testing. The preliminary results show that, for the wavelet transform based method, the correct classification rates of two-class classification and cataract grading are 90.9% and 77.1%, respectively. The correct classification rates of two-class classification and cataract grading are 86.1% and 74.0% for the sketch based method, which is comparable to the wavelet transform based method. The pilot study demonstrates that our research on fundus image analysis for cataract classification and grading is very helpful for improving the efficiency of fundus image review and ophthalmic healthcare quality. We believe that this work can serve as an important reference for the development of similar health information system to solve other medical diagnosis problems.

    本文提出的眼底图像分析是基于计算机帮助系统的，这种系统是对于白内障的自动分级和分类系统，这种方法对于降低经验丰富眼科医师（稀缺资源）的负担具有很大的潜力，同时能帮助白内障患者在欠发达地区及时了解他们的白内障的患病请款以及获得医生的治疗建议。该系统是由眼底图像预处理，图像特征提取和自动白内障分类和分级组成。使用小波变换和基于轮辐方法从眼底图像中提取特征的方法适用于对白内障的分类分级。特征提取后，多类判别分析算法用于白内障的分类，包括两类（白内障或不是白内障）分类和轻微，中度和严重三个分级。一个真实的数据集显示，使用基于小波变换的方法，对于两类分类和分级的正确率分别是90.9%和77.1%，而使用基于轮辐的方法，对于两类分类和分级的正确率分别是86.1%和74.0%，这是与基于小波变换方法的对比。试点研究表明，我们研究眼底图像从而对白内障分类分级这样做对分析改善眼底图像审查和眼科医疗服务质量的效率非常有帮助。我们相信，这项工作可以作为类似的卫生信息系统的发展，以解决其他医疗诊断问题的重要参考。

1. Introduction

Its goal is to reduce the burden of scarce resources and improve the effectiveness and efficiency of fundus image review, through which to enable active and enhanced healthcare services.

本系统的目标就是要降低眼科医生的负担，并且提高对眼底图片检查的有效性和效率，通过它来活跃和增强医疗服务。

segmentation and location of retinal structures  视网膜结构的分割和位置

retinal lesions  视网膜病变

vessels  血管

optic disc 视神经盘

fovea  视网膜的中心凹

diagnose systems for specific retina-related diseases  特定的视网膜相关疾病的诊断系统

microaneurysms  为动脉瘤

diabetic retinopathy 糖尿病视网膜病变

age-related macular degeneration  老年性黄斑变性

glaucoma  青光眼

cardiovascular diseases  心血管疾病

nuclear cataract  老年性核硬化性白内障 

cortical cataract  皮质性白内障

posterior sub-capsular cataract  后囊下白内障

2. The framework of computer-aided cataract classification and grading system

介绍了系统的基本框架

3. Cataract feature extraction and classification

3.1Feature extraction based on wavelet transform

哈尔小波变换（Haar wavelet transform）

3.2. Feature extraction based on the sketch method with discrete cosine transform

4. Initial result evaluation

训练和测试数据集中有445张眼底图片，对应的没有，轻度，中度，重度分别有199，148，71，27张，以下第3节中所描述的特征抽取方法中，多级的判别分析是用于分类和白内障的分级。多费希尔分类算法是由从数据组中随机选择70％的样品，然后通过使用其他30％的样品进行测试训练。通过重复训练和测试过程100次，获得的上述分类分级方法的整体性能。

5. The real-world application

三个部分两个阶段两点优势

6. Conclusions