支持向量机-微量元素法用于胃癌的模式识别

Support vector machines-trace element method for pattern recognition of gastric cancer

目的利用支持向量机-微量元素法建立胃癌的模式识别,以期作为胃癌诊断的辅助手段。方法以血浆作为研究对象,采用电感耦合等离子体发射光谱法测定健康者与胃癌患者血清中Zn、Fe、Cu、Ni、Co、Mo、Mn、Cr和Zn/Cu 9个指标,结合支持向量机分类模式,建立胃癌诊断的模型。结果多项式核函数的平均灵敏度为93.0%、平均特异度为91.4%、平均准确度为92.6%;径向基核函数的平均灵敏度为95.0%、平均特异度为88.6%、平均准确度为93.3%。结论采用的支持向量机-微量元素法为胃癌的预防、诊断和临床治疗提供了一种简便的辅助参考方法。

Objective To establish the identification model of gastric cancer by the support vector machine- trace element method as the supplementary means of the diagnosis of gastric cancer. Methods The inductive- ly coupled plasma atomic emission spectrometry and support vector machines for pattern classifi cation were adopted to make discrimination for 9 indicators of serum Zn, Fe, Cu, Ni, Co, Mo, Mn, Cr and Zn/Cu be- tween healthy volunteers and patients, and establish the diagnosis model of gastric cancer. Results The aver- age sensitivity, specificity and accuracy of polynomial kernel function were 93.0% ,91.4% and 92. 6% ;the average sensitivity, specificity and accuracy of RBF kernel function were 95.0% and 88.6% and 93.3%, respectively. Conclusions This method is suitable for aided diagnosis of gastric cancer, and also lays a foun- dation for the further study on the relationship between trace elements and various diseases.