CAJ | 학술논문

目的:总结基于眼底图像的人工神经网络分类器研究现状与进展,为基于眼底图像的疾病计算机辅助诊断提供基础.方法:本研究采用文献检索和综述的方法,对国内外关于眼底图像的人工神经网络(Artificial Neural Network, ANN)研究进行整理,主要对已发表的文献中ANN技术以及相关眼底图像的特征提取技术进行整理和分析.结果:本次研究检索到符合要求的研究共计27篇,最终纳入17篇研究进行资料提取.提取的指标为分类内容、特征提取、神经网络类型/方法、ANN输入层层数、隐藏层神经元数量、输出、分类指标等.针对不同样本的研究,提取的特征有所区别.目前研究眼底图像的主要ANN方法是后反馈神经网络(BP-ANN).结论:基于眼底图像的ANN能够为视网膜病变的辅助分析提供一定帮助.
목적:총결기우안저도상적인공신경망락분류기연구현상여진전,위기우안저도상적질병계산궤보조진단제공기출.방법:본연구채용문헌검색화종술적방법,대국내외관우안저도상적인공신경망락(Artificial Neural Network, ANN)연구진행정리,주요대이발표적문헌중ANN기술이급상관안저도상적특정제취기술진행정리화분석.결과:본차연구검색도부합요구적연구공계27편,최종납입17편연구진행자료제취.제취적지표위분류내용、특정제취、신경망락류형/방법、ANN수입층층수、은장층신경원수량、수출、분류지표등.침대불동양본적연구,제취적특정유소구별.목전연구안저도상적주요ANN방법시후반궤신경망락(BP-ANN).결론:기우안저도상적ANN능구위시망막병변적보조분석제공일정방조.
Objective: To summarize the artificial neural network classifier based on fundus image, which is fundamental to fundus image based computer aided diagnosis. Methods: In our study, literature search and review was utilized, and the artificial neural network studies on fundus images were obtained and which feature extraction technology were extracted and analyzed. Results: A total of 27 relevant studies were collected and eventually 17 of them were selected for information extraction. Of these studies, the content, feature extraction and neural network classification type and method, the number of artificial neural network (ANN) input layer, the number of hidden layer neurons, output, as well as classification were summarized. For different studies, the extracted features were different. Besides, the main ANN method for fundus images was back propagation ANN (BP-ANN). Conclusion: The ANN can provide certain help for retinopathy aided analysis on fundus images.