基于多角度分析统计的眼底图像微动脉瘤自动检测

糖尿病视网膜病变(DR)是新增失明的主要原因，而视网膜微动脉瘤(MA)是糖尿病视网膜病变的早期临床表现之一。然而，由于微动脉瘤体积小，与周围背景的对比度差，很难用肉眼直接观察出来。本文提出了一种基于多角度分析统计的特征提取方法，每种预处理方法作为一个角度提取一组特征，利用多层感知机分类器(MLP)进行分类，最后对所有的分类统计求均值。该方法在e-ophtha-MA数据库上进行了评估，并获得了较高的灵敏度。使用FROC指标，最终平均得分为0.510,灵敏度为0.91,AUC值为0.98,与现有最先进的方法相当。实验证明，本文提出的基于多角度分析统计提取微动脉瘤的方法增强了MA的识别能力，有利于分类，提高了检测性能。

Automatic Detection of Microaneurysms in Fundus Images Based on Multi-angle Analysis Statistics

Diabetic retinopathy (DR) is the main cause of new blindness, and retinal microaneurysm (MA) is one of the early clinical manifestations of DR. However, due to their small size and poor contrast with the surrounding background, MAs are difficult to directly observe with the naked eye. This paper proposes a feature extraction method based on multi-angle analysis statistics. Each preprocessing method extracts a set of features as an angle, uses multi-layer perceptron classifier (MLP) for classification, and finally calculates the mean value of all classification statistics. The proposed method was evaluated on e-ophtha-MA database and high sensitivity was obtained. Using the FROC index, the final average score was 0.510, the sensitivity was 0.91, and the AUC was 0.98, which was comparable to the state of the art methods. The experimental results show that the method of extracting microaneurysms based on multi-angle analysis statistics proposed in this paper enhances the recognition ability of MA, is conducive to classification, and improves the detection performance.