基于RBF神经网络的消化道出血图像检测研究  

The Method of Bleeding Detection in WCE Images Based on RBF Function Network

作　　者：徐丽珍[1] 

机构地区：[1]杭州职业技术学院信息电子工程系,杭州310018

出　　处：《科技通报》2014年第3期48-52,共5页Bulletin of Science and Technology

基　　金：浙江省教育科学规划课题(NO:SCG178);杭州职业技术学院科研课题(NO:ky201338)

摘　　要：鉴于临床无痛无创消化道检测中无线胶囊内窥镜的广泛应用,提出使用计算机图像识别技术进行内窥图像出血区域的自动检测方法。该方法分析了出血像素与聚类中心的欧拉距离分布和RBF的应用特点,在CIEXYZ和CIEL*a*b*颜色空间设计了一种RBF分类器应用于图像出血模式的识别方法,并通过软件编程实现了出血区域自动检测新方法。实验表明该方法能正确识别出内窥消化道图像中的出血区域,从而将图像分类为出血模式与非出血模式。In view of the popular application of the Wireless Capsule Endoscopy（WCE） in the clinical painless and non-invasive gastrointestinal detection , Automatic detection method is proposed using computer image recognition technology for endoscopic bleeding region image. The Euler distance distributions of bleeding pixels to the clustering center are proposed and the application of Radial Basis Function （RBF） in image pattern recognition is analyzed. Based on these studies, a RBF classifier in CIEXYZ and CIERGB is built to detect the bleeding regions in WCE images. The bleeding detection software based on the RBF is programmed, and the experiments demonstrate that the bleeding regions in WCE images can be correctly detected. Therefore the images are classified into bleeding frames and non-bleeding frames.

关 键 词：RBF 消化道出血检测 图像识别 

分 类 号：R574[医药卫生—消化系统] TP391[医药卫生—内科学]