Statistical Pattern Recognition in Small Sample Size Situations and its Application to Detection of Hepatocellular Carcinoma

本研究的目的是建立一种新的血液检测系统，用于检测与丙型肝炎病毒（HCV）感染相关的肝细胞癌（HCC）。为了开发新的测试系统，我们必须建立一个特征选择方法，以找到有效的生物标记在小训练样本量的情况下，相比非常高的维度。我们发现了2个甲基化基因（SPINT 2和SRD 5A2）。使用2个甲基化基因，AFP。PIVKA II。我们设计了一套检测肝癌的血液检测系统，该系统具有较高的检测能力（73，2%的灵敏度和87。7%特异性）。这表明我们的血液检测系统可能会导致临床应用。

The purpose of this study is to develop a novel blood test system for detecting hepatocellular carcinoma(HCC) associated with hepatitics C virus(HCV) infection. In order to develop the novel test system, we must establish a feature selection method to find effective bio-makers in small training sample size situations, compared with very high dimensions. We found 2 methylated genes(SPINT2 and SRD5A2). With use of 2 methylated genes, AFP. and PIVKA II. we designed the blood test system for detecting HCC, which achieved high abilities(73. 2% sensitivity and 87. 7% specificity) for 258 test samples provided from 4 institutes. This suggests that our blood test system might lead to clinical applications.