多变量判别分析用于癌症诊断研究

用感应耦合等离子体原子发射光谱及石墨炉原子吸收光谱仪测定了正常人及癌症病人头发样品中15种元素的含量。所得数据用多元多项式扩展增维和逐步回归变量压缩技术以及PLS方法处理,得到了病人与正常人分类极为清晰的二维判别图。据此可将头发用作癌症临床诊断中的分析样品以取代血液样品。

Research on an Application of Multivariate Discriminant Analysis to Cancer Diagnosis

The concentrations of 15 trace elements in human hair samples including both normal people and cancer patient samples are determined by ICP-AES and GFAAS, These concentration data are then treated by use of a technique which involves polynomial expanding, model selection and PLS method. The good classification results between two classes of samples are obtained and showed in discrimination plane figure. It is believed that the hair sample can be used as a test sample instead of the blood sample in the diagnosis for the cancer.