基于神经网络的肺癌细胞识别方法的研究  

APPROACH TO THE RECOGNITION OF LUNG CANCER CELLS ON THE BASIS OF NEURAL NETWORK

作　　者：陆新泉[1] 李宁[1] 陈世福[1] 叶玉坤[2] 

机构地区：[1]南京大学计算机软件新技术国家重点实验室,南京210093 [2]中国人民解放军八一医院,南京210002

出　　处：《计算机应用与软件》2001年第10期1-3,6,共4页Computer Applications and Software

基　　金：国家自然科学基金(69875006;类别A);国家九五攻关课题(96-906-01-18)的资助

摘　　要：细胞识别是图像处理和模式识别领域的一个研究热点,有着十分广泛的应用前景。本文提出了基于神经网络算法FTART2的肺癌细胞识别方法,讨论了FTART2的网络结构、输入矢量的标准化及分类算法。用513个样本对网络进行训练,再用716个样本组成测试集进行测试,实验结果表明:本文提出的基于FTART2的肺癌细胞分类器与基于标准BP的分类器相比,具有学习速度快、分类精度高的特点。Recognizing cell is a studying focus in the field of image process and pattern recognition, with a very broad application prospect. This paper presents a recognition method of lung cancer cells on the basis of the neural network algorithm FTART2, and discusses the network structure of FTART2, standardization of input vector, and classification algorithm. We train the network with 513 samples and test it with a test set of 716 samples. The result shows the lung cancer classifier based on FTART2, compared with the classifier based on standard BP, has the features of faster learning speed and higher classification accuracy.

关 键 词：神经网络 模式识别 图像处理 肺癌 细胞识别 计算机 

分 类 号：R734.2[医药卫生—肿瘤] TP391.41[医药卫生—临床医学]