利用人工神经网络识别小圆蓝细胞瘤亚型  

Identification of small,round blue-cell tumors subtypes by artificial neural network

作　　者：郭松[1] 刘心平[1] 史呈伟[1] 李立群[1] 孟群[2] 

机构地区：[1]哈尔滨医科大学科研处 [2]哈尔滨医科大学公共卫生学院,黑龙江哈尔滨150081

出　　处：《哈尔滨医科大学学报》2010年第3期230-232,共3页Journal of Harbin Medical University

摘　　要：目的研究人工神经网络方法在识别肿瘤亚型中的有效性。方法利用小圆蓝细胞瘤的基因芯片数据,采用人工神经网络方法从分子层面对其亚型进行分类。在计算出每个基因信噪比指标的基础上,采用加权投票方法确定出特征基因,并利用线性神经网络方法构建分类器。结果该分类器在内部验证中获得了97.82%的正确率,在独立检验样本集的预测过程中获得了95%的预测精度。结论人工神经网络模型在对癌症的亚型识别以及治疗候选靶点的确定等领域具有很大的应用潜力。Objective To investigate the effectiveness of artificial neural network in identification of tumor subtypes.Methods Based on the microarray data of the small,round blue-cell tumors,artificial neural network was employed to classify the tumor subtypes in molecular level.Firstly,the odds ratio was calculated for each gene.Secondly,a weighted vote rule was used for identification of feature genes.Finally,the classifications were constructed by linear neural network.Results The average accuracy of the classification in the inner-validation was 97.82%,and 95% was obtained by using the independent test sample set.Conclusion There is a potential application of the artificial neural network for tumor diagnosis and therapy by identification of tumor subtypes and candidate targets.

关 键 词：人工神经网络 肿瘤亚型 基因芯片 

分 类 号：R730.4[医药卫生—肿瘤]