结合代价敏感半监督集成学习的糖尿病视网膜病变分级

针对传统糖尿病视网膜病变（糖网）分级诊断系统中，由于数据集中缺少病灶区域的标记和类别分布的不平衡性导致无法有效地进行监督性分类的问题，提出基于代价敏感的半监督Bagging（CS-SemiBagging）的糖网分级方法。首先，从眼底图像上删除视网膜血管，并在此图像上检测疑似的红色病灶（微动脉瘤（MAs）与出血斑（HEMs））；然后，从颜色、形状和纹理方面提取22维的特征用于描述每个病灶区域；其次，构建一个CS-SemiBagging模型对MAs与HEMs进行分类；最后，依据不同病灶的数量将糖网划分为4级。通过对国际公共数据集MESSIDOR进行糖网分级评估实验，所提方法获得平均准确率为90.2%，与经典的半监督学习的Co-training方法相比提高了4.9个百分点。实验结果表明，CS-SemiBagging方法在无需提供病灶标注的情况下，能够高效自动地对糖网进行分级，从而既能免除医学图像中标注病灶的费时费力，又可以避免样本类别分布不平衡对分类算法的性能影响，获得较好的效果。     

An automated decision-support system for non-proliferative diabetic retinopathy disease based on MAs and HAs detection

Diabetic retinopathy (DR) has become a serious threat in our society, which causes 45% of the legal blindness in diabetes patients. Early detection as well as the periodic screening of DR helps in reducing the progress of this disease and in preventing the subsequent loss of visual capability. This paper provides an automated diagnosis system for DR integrated with a user-friendly interface. The grading of the severity level of DR is based on detecting and analyzing the early clinical signs associated with the disease, such as microaneurysms (MAs) and hemorrhages (HAs). The system extracts some retinal features, such as optic disc, fovea, and retinal tissue for easier segmentation of dark spot lesions in the fundus images. That is followed by the classification of the correctly segmented spots into MAs and HAs. Based on the number and location of MAs and HAs, the system quantifies the severity level of DR. A database of 98 color images is used in order to evaluate the performance of the developed system. From the experimental results, it is found that the proposed system achieves 84.31% and 87.53% values in terms of sensitivity for the detection of MAs and HAs respectively. In terms of specificity, the system achieves 93.63% and 95.08% values for the detection of MAs and HAs respectively. Also, the proposed system achieves 68.98% and 74.91% values in terms of kappa coefficient for the detection of MAs and HAs respectively. Moreover, the system yields sensitivity and specificity values of 89.47% and 95.65% for the classification of DR versus normal.     

Automatic diabetic retinopathy diagnosis using adjustable ophthalmoscope and multi-scale line operator

Diabetic Retinopathy (DR), the most common one of diabetic eye diseases that cause loss of vision and blindness, has become one of major health problems today. However, DR can be eased through timely treatment and periodical screening. In this paper, we proposes an automatic diabetic retinopathy diagnostic system to help patients know about their retinal conditions. We design a portable ophthalmoscope, which is composed of a retinal lens, a smartphone and a frame between them to help patients take fundus images anywhere and anytime. Then the images are transmitted to be analyzed, including localization of optic disk and macular, vessel segmentation, detection of lesions, and grading of DR. We use a multi-scale line operator to improve accuracy in segmenting small-scale vessels, a binary mask and image restoration to reduce the effect of the existence of the vessels on optic disk localization. After the analysis, the fundus image are then graded as normal, mild Non-Proliferative Diabetic Retinopathy (NPDR), moderate NPDR or severe NPDR. The grading process uses region segmentation to improve the efficiency. The final grading results are tested based on the fundus images provided by the hospitals. We evaluate our system through comparing our grading result with those graded by experts, which comes out with an overall accuracy of up to 85%.     

Construction and verification of retinal vessel segmentation algorithm for color fundus image under BP neural network model

To improve the accuracy of retinal vessel segmentation, a retinal vessel segmentation algorithm for color fundus images based on back-propagation (BP) neural network is proposed according to the characteristics of retinal blood vessels. Four kinds of green channel image enhancement results of adaptive histogram equalization, morphological processing, Gaussian matched filtering, and Hessian matrix filtering are used to form feature vectors. The BP neural network is input to segment blood vessels. Experiments on the color fundus image libraries DRIVE and STARE show that this algorithm can obtain complete retinal blood vessel segmentation as well as connected vessel stems and terminals. When segmenting most small blood vessels, the average accuracy on the DRIVE library reaches 0.9477, and the average accuracy on the STARE library reaches 0.9498, which has a good segmentation effect. Through verification, the algorithm is feasible and effective for blood vessel segmentation of color fundus images and can detect more capillaries.

     

Automatic segmentation of optic disc in retinal fundus images using semi-supervised deep learning

Diseases of the eye require manual segmentation and examination of the optic disc by ophthalmologists. Though, image segmentation using deep learning techniques is achieving remarkable results, it leverages on large-scale labeled datasets. But, in the field of medical imaging, it is challenging to acquire large labeled datasets. Hence, this article proposes a novel deep learning model to automatically segment the optic disc in retinal fundus images by using the concepts of semi-supervised learning and transfer learning. Initially, a convolutional autoencoder (CAE) is trained to automatically learn features from a large number of unlabeled fundus images available from the Kaggle’s diabetic retinopathy (DR) dataset. The autoencoder (AE) learns the features from the unlabeled images by reconstructing the input images and becomes a pre-trained network (model). After this, the pre-trained autoencoder network is converted into a segmentation network. Later, using transfer learning, the segmentation network is trained with retinal fundus images along with their corresponding optic disc ground truth images from the DRISHTI GS1 and RIM-ONE datasets. The trained segmentation network is then tested on retinal fundus images from the test set of DRISHTI GS1 and RIM-ONE datasets. The experimental results show that the proposed method performs on par with the state-of-the-art methods achieving a 0.967 and 0.902 dice score coefficient on the test set of the DRISHTI GS1 and RIM-ONE datasets respectively. The proposed method also shows that transfer learning and semi-supervised learning overcomes the barrier imposed by the large labeled dataset. The proposed segmentation model can be used in automatic retinal image processing systems for diagnosing diseases of the eye.

     

Exudates and optic disk detection in retinal images of diabetic patients

Diabetic retinopathy is the progressive pathological alterations in the retinal microvasculature that very often causes blindness. Because of its clinical significance, it will be helpful to have regular cost‐effective eye screening for diabetic patients by developing algorithms to perform retinal image analysis, fundus image enhancement, and monitoring. The two cost‐effective algorithms are proposed for exudates detection and optic disk extraction aimed for retinal images classification and diagnosis assistance. They represent the effort made to offer a cost‐effective algorithm for optic disk identification, which will enable easier exudates extraction, exudates detection and retinal images classification aimed to assist ophthalmologists while making diagnoses. The proposed algorithms apply mathematical modeling, which enables light intensity levels emphasis, easier optic disk and exudates detection, efficient and correct classification of retinal images. The algorithm is robust to various appearance changes of retinal fundus images and shows very promising results. Fundus images are classified into those that are healthy and those affected by diabetes, based on the detected optic disk and exudates. The obtained results indicate that the proposed algorithm successfully and correctly classifies more than 98% of the observed retinal images because of the changes in the appearance of retinal fundus images typically encountered in clinical environments. Copyright © 2014 John Wiley & Sons, Ltd.     

RBF神经网络和阈值分割实现视网膜硬性渗出自动检测

为自动检测出眼底图像中的硬性渗出,构建眼底图像的糖尿病视网膜病变自动筛查系统,提出一种基于RBF神经网络和阈值分割的硬性渗出自动检测方法.首先,利用基于最小类内离散度的改进Otsu分割方法对眼底图像绿色通道进行粗分割获取病灶候选区域;然后,利用logistic回归对候选区域的多个特征进行选择;最后,利用候选区域的优化特征集及相应判定结果建立RBF神经网络;此外,提出采用后处理以进一步提高检测精度.利用本文方法对50幅不同颜色、不同亮度的眼底图像进行硬性渗出自动检测,得到图像水平灵敏度100％,特异性90.9％,准确率96.0％;病灶区域水平灵敏度93.9％,阳性预测值95.5％;平均每幅图像处理时间13.6 s.结果表明本文方法稳定可靠,能快速有效地自动检测出眼底图像中的硬性渗出.     

Examining the variation of vascular structure in digital fundus images using textural pattern

One of the most significant retinal abnormality in which an individual loses the vision is diabetic retinopathy (DR). The appropriate way to treat this disease would be easier if it is detected at an earlier stage. The study on the vasculature extracted from illumination correction on the fundus image brings the presence of diabetic retinopathy. This preprocessing involves three steps. Initially illumination and reflectance estimation is done and then illumination correction is employed and finally the clipped histogram equalization is done to preserve the brightness of the image so that the information on the retinal image may not get saturated. Here, k-means segmentation process has been done and the local binary pattern (LBP) has been calculated. The selected feature vectors are then classified by using an echo state neural network (ESNN). The proposed method has been tested on publically available database DIARETDB1 that contained 89 DR fundus images in total. The result of detecting and classifying the pathology based on vasculature study on these images yielded sensitivity of 86.46%, specificity of 80.47%, and accuracy of 96.92%.     

结合局部熵和鲁棒主成分分析的眼底图像硬性渗出物检测方法

针对眼科医生诊断眼底图像工作耗时且易出错的问题，提出一种无监督的眼底图像硬性渗出物检测方法。首先，通过形态学的背景估计方法去除血管、暗病变区域和视盘；然后，以图像亮度通道为初始图像，利用硬性渗出物在眼底图像中的局部性和稀疏性，结合局部熵和鲁棒主成分分析方法分解得到低秩矩阵和稀疏矩阵；最后，归一化稀疏矩阵得到硬性渗出物区域。实验结果显示，在e-ophtha EX和DIARETDB1公开数据库上，所提方法在病灶水平上灵敏性为91.13%和特异性为90%，在图像水平上准确率为99.03%，平均运行时间0.5 s；与支持向量机（SVM）和K-means方法相比灵敏性高且耗时少。     

基于眼底图像的糖尿病视网膜病变分类系统设计

针对眼底图像,设计了一个糖尿病视网膜病变(Diabeticretinopathy,DR)分类系统,通过对视网膜血管图像进行定量分析来实现对DR病程的分类.采用Messidor数据集的眼底照片图像,这个数据集共包含100个研究项目,其中32张未患DR的眼底照片,24张患NPDR.根据数据集中DR患者和非DR人群的眼底图像以及眼科专家的分类结果,利用数字图像处理技术分析特征值的统计意义,判断该图像所反映的DR病程.预处理为提取特征值前的图像增强、主像素成分分析、匹配滤波以及Gabor滤波,对预处理后的图像进行直径、角度和分形维数等特征值提取.最终结果展示了直径、角度和分形维数的准确率达到了93％、96％、81.8％,提供有效的辅助诊断手段.糖尿病视网膜病变的特征值分析包括直径、角度和分形维数准确率较高.对于缺乏医疗条件的地区很有价值.     

Computerised approaches for the detection of diabetic retinopathy using retinal fundus images: a survey

Eye-related disease such as diabetic retinopathy (DR) is a medical ailment in which the retina of the human eye is smashed because of damage to the tiny retinal blood vessels in the retina. Ophthalmologists identify DR based on various features such as the blood vessels, textures and pathologies. With the rapid development of methods of analysis of biomedical images and advanced computing techniques, image processing-based software for the detection of eye disease has been widely used as an important tool by ophthalmologists. In particular, computer vision-based methods are growing rapidly in the field of medical images analysis and are appropriate to advance ophthalmology. These tools depend entirely on visual analysis to identify abnormalities in Retinal Fundus images. During the past two decades, exciting improvement in the development of DR detection computerised systems has been observed. This paper reviews the development of analysing retinal images for the detection of DR in three aspects: automatic algorithms (classification or pixel to pixel methods), detection methods of pathologies from retinal fundus images, and extraction of blood vessels of retinal fundus image algorithms for the detection of DR. The paper presents a detailed explanation of each problem with respect to retinal images. The current techniques that are used to analyse retinal images and DR detection issues are also discussed in detail and recommendations are made for some future directions.     

基于深度卷积神经网络的糖尿病视网膜病变分期及病灶检测

针对糖尿病视网膜病变（DR）图像分辨率过大、病灶特征过于分散难以获取以及正负难易样本不平衡而导致DR分期精确率一直无法得到有效提高的问题,提出了改进的基于快速区域的卷积神经网络（Faster R-CNN）和子图分割相结合的DR分期方法。首先,使用子图分割解决视盘区域对于病灶识别的干扰问题;其次,在特征提取阶段使用深度残差网络以解决病灶在高分辨率眼底图像中占比小而导致的特征难以获取的问题;最后,在感兴趣区域（ROI）生成时采用在线困难样本挖掘（OHEM）方法解决正负难易样本不平衡的问题。在国际公开数据集EyePACS进行DR分期实验,所提方法在DR病分期中精确率0期达到94.83%,1期达到86.84%,2期达到94.00%,3期达到87.21%,4期达到82.96%。实验结果表明,改进后的Faster R-CNN能对DR图像高效分期并自动标注出病灶。     

基于双注意力编码-解码器架构的视网膜血管分割

眼底视网膜血管的分割提取对于糖尿病、视网膜病、青光眼等眼科疾病的诊断具有重要的意义。针对视网膜血管图像中的血管难以提取、数据量较少等问题,文中提出了一种结合注意力模块和编码-解码器结构的视网膜血管分割方法。首先对编码-解码器卷积神经网络的每个卷积层添加空间和通道注意力模块,加强模型对图像特征的空间信息和通道信息(如血管的大小、形态和连通性等特点)的利用,从而改善视网膜血管的分割效果。其中,空间注意力模块关注于血管的拓扑结构特性,而通道注意力模块关注于血管像素点的正确分类。此外,在训练过程中采用Dice损失函数解决了视网膜血管图像正负样本不均衡的问题。在3个公开的眼底图像数据库DRIVE,STARE和CHASE_DB1上进行了实验,实验数据表明,所提算法的准确率、灵敏度、特异性和AUC值均优于已有的视网膜血管分割方法,其AUC值分别为0.9889,0.9812和0.9831。实验证明,所提算法能够有效提取健康视网膜图像和病变视网膜图像中的血管网络,能够较好地分割细小血管。     

基于多任务学习的眼底图像红色病变点分割

糖尿病性视网膜病变（糖网病）是导致成年人视觉损失的主要因素之一.早期的眼底筛查可以显著降低这种视觉损失的可能性.彩色眼底图像由于具有采集便利、对人体无伤害等特点,常被用于大规模的眼底筛查工作.对眼底图像中的红色病变点而言,微动脉瘤是轻度非增殖性糖网病的主要标志,出血点与中度及重度非增殖性糖网病的诊断有关,因此,眼底图像中出血点和微动脉瘤的准确分割对糖网病分级诊断具有重要参考价值.提出一种基于多任务学习的分割模型Red-Seg来对出血点和微动脉瘤进行分割.该网络包含两个分支,每个分支处理一种病变点.设计了一种两阶段训练算法,并且两个阶段使用不同的损失函数：第1阶段使用改进的Top-k带权交叉熵损失函数,将模型训练集中在难分样本上;第2阶段将最小化假阳性和假阴性作为Red-Seg模型训练的优化目标,进一步减少病变点误分.最后,在IDRiD数据集上进行模型验证,并与其他病变点分割方法进行对比.实验结果表明,在应用Red-Seg模型进行微动脉瘤和出血点红色病变点分割时,两阶段训练算法可以显著减少病变点误分情况,尤其是出血点分割的准确率和召回率都提高2.8%.同时,与HED、FCRN、DeepLabv3+和L-Seg等图像级分割模型相比,Red-Seg模型在微动脉瘤分割上获得了更好的AUC_PR.     

基于RBF神经网络的高效硬性渗出自动检测方法

为快速、可靠地自动检测出眼底图像中的早期糖尿病视网膜病变—硬性渗出,提出一种基于RBF神经网络的硬性渗出自动检测方法。首先,采用全局、局部相结合的基于图像直方图灰度统计分析的阈值分割法获取70幅眼底图像中的3590个病灶候选区域;然后,提取候选区域的24个特征,并利用logistic回归对其进行选择;最后,利用候选区域的10个特征及相应判定结果建立RBF神经网络;另外,提出采用后处理以进一步提高检测精度。利用该方法对50幅不同颜色、不同亮度的眼底图像进行处理并与眼科医师的判断结果相比较,得到图像水平灵敏度100%,特异性86.4%,准确率94.0%;病灶区域水平平均灵敏度93.9%,平均阳性预测值95.5%;平均每幅图像处理时间13.7s。该结果表明：基于RBF神经网络的自动检测方法能有效检测出眼底图像中的硬性渗出,且处理效率较高。     

基于IDA-RF眼底硬性渗出物的检测

为了准确检测眼底图像中的硬性渗出物,降低糖尿病性视网膜病变引起的失明,提出了一种基于IDA-RF的眼底硬性渗出物的检测方法。对眼底图像预处理,提取渗出物候选区域。利用[k]-means初始种群,与万有引力搜索算法相结合,改变步长更新公式,提出一种改进的蜻蜓算法（IDA）。IDA在寻优过程中对随机森林算法参数进行优化,并利用优化后的随机森林算法对渗出物候选区域分类,提取最终精确的硬性渗出物。该方法在公开的眼底图像数据库进行实验,与RF、DA-RF、GSA-RF相比,准确率达97.28%。实验表明,提出的方法能够准确检测硬性渗出物且鲁棒性能好。     

结合FCM聚类和边缘感知模型的眼底渗出物检测

眼底图像中渗出物是构成糖尿病视网膜病变(Diabetic Retinopathy,DR)的早期症状之一,提出一种结合模糊C-均值(Fuzzy C-Means,FCM)聚类和边缘感知模型的方法实现对渗出物的检测。为保证后期检测精度和效率,对眼底图像进行增强对比度和均衡亮度等预处理操作,用FCM聚类分割出渗出物候选区域,利用基于判断邻域灰度差异的边缘感知模型对候选区域进行筛选,通过移除视盘区域,从而得到真实的渗出物区域。在公开的数据集上进行实验,算法的灵敏度为86.65%,特异性为94.79%,阳性预测值为95.14%,准确度为92.09%。结果表明,该方法能够有效实现对眼底渗出物的自动检测。     

基于相位一致性的眼底图像视网膜血管分割

眼底图像的视网膜血管分割是眼底图像处理的重要组成部分,视网膜血管对于医学研究和临床诊断有着重要的作用。传统图像分割算法都有一定的缺陷,而相位一致性算法由于不受对亮度和对比度的影响,且有着较好的分割效果,可以用于图像特征的提取和分割。为此提出了将相位一致性算法应用于眼底图像的血管提取中,采用真实的眼底图像数据库进行实践,证明了可较好地用于眼底图像视网膜血管分割。