基于粒子群算法的医学图像分类算法研究  

Research on medical image classification algorithm based on particle swarm optimization

作　　者：陈迪 李宁[2] CHEN Di;LI Ning(The First Affiliated Hospital of Air Force Military Medical University,Xi'an 710032,China;Shaanxi Yanchang Petroleum(Group)Co.,Ltd.,Xi'an 710075,China)

机构地区：[1]空军军医大学第一附属医院,陕西西安710032 [2]陕西延长石油(集团)有限责任公司,陕西西安710075

出　　处：《电子设计工程》2019年第2期189-193,共5页Electronic Design Engineering

摘　　要：针对医学图像具有较大的相似性和交叉性,易造成归属类别混乱的问题,提出了一种基于粒子群算法的医学图像分类方法。该方法使用形态学滤波和阈值法进行预处理;使用SIFT特征描述子来提取图像的局部特征,并使用聚类的方法得到SIFT特征的"视觉词汇";使用粒子群算法选出一些列多样性和精度更高的SVM、KNN和AdaBoost分类器对特征进行分类。对15种不同类型的医学图像进行分类的结果表明,所提出的方法取得了94.72%的分类精度,且相比于单个分类器的方法具有较大的性能提升。In view of the great similarity and cross-over of medical images and the confusion of attribution categories,a medical image classification method based on particle swarm optimization is proposed.The method uses morphological filtering and thresholding method to preprocess;SIFT feature descriptors are used to extract the local features of the image,and SIFT feature“visual vocabulary”is obtained by using clustering method;a number of columns are selected by using the particle swarm algorithm.The SVM,KNN,and AdaBoost classifiers,which are more accurate and more accurate,classify features.The results of the classification of15different types of medical images show that the proposed method achieves a classification accuracy of94.72%,and has a greater performance improvement compared to the method of a single classifier.

关 键 词：粒子群 医学图像 分类 SIFT 词袋模型 SVM 

分 类 号：TP394.1[自动化与计算机技术—计算机应用技术]