医学图像识别中多分类器融合方法的研究进展  

Advances in fusion of multiple classifiers for medical image recognition

作　　者：谢华[1] 夏顺仁[1] 张赞超[1] 

机构地区：[1]浙江大学生物医学工程教育部重点实验室

出　　处：《国际生物医学工程杂志》2006年第3期152-157,共6页International Journal of Biomedical Engineering

基　　金：国家自然科学基金资助项目(60272029);浙江省自然科学基金资助项目(M603227)

摘　　要：计算机辅助医学图像分析识别对多种疾病的临床诊断有着重要的意义。由于医学图像自身的复杂性,单一分类器的识别性能常常难以满足临床上的要求,因此近年来,作为一种能有效改进单一分类器识别性能的方法,多分类器融合技术被逐步应用到包括乳腺X光片识别、肿瘤细胞识别以及内窥镜图像分析等领域,并取得了更为满意的识别结果。在参阅大量文献的基础上,对多分类器融合识别技术的理论分析及其在医学领域的研究及应用现状进行了综述,进而对其存在的问题进行了分析以及前景展望。The computer-aided medical image analysis and recognition technique is of great importance for the clinical diagnosis of many crucial diseases, but the inherent complexity of medical image always requires more reliable and accurate classification methods. In order to improve the performance of single classifier, different multi-classifier fusion methods have been widely used to identify different medical images in recent years, such as mammograms, endoscopic images, cell smears, etc. This paper gives an overview of different classifier fusion methods, theoretical foundations and application of the technique in medical areas. In the end, the problems existed in current fusion techniques are presented and foreground that may dominate this area of research in the future are expected.

关 键 词：医学图像 多分类器 集成学习 模式识别 计算机辅助诊断 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]