肺癌血清近红外检测谱段分析

Analysis of NIR Spectral Bands for Lung Cancer Serum

运用近红外光谱技术采集肺癌阴性/阳性血清光谱,样本在4000~4500、4500~5200、6800~7200、11000~12000cm^(-1)谱段有吸收峰,包含N-H、C-H、C=O、O-H信息。结合支持向量机建立血清分类模型,筛选了血清分析谱段。采用等区间划分的模式进行了两次考察,结果表明8000~12000cm^(-1)谱段分析结果较好,而10800~11200cm^(-1)谱段的分析错误率最低。吸收峰谱段结果也表明6800~7200、11000~12000cm^(-1)谱段分析结果较好。综合以上各点,肺癌血清近红外光谱检测谱段可以设置为7000~12000cm^(-1)。

Collected near infrared spectra of lung cancer ＋/- serum samples.The spectra showed that serum has four absorption peaks in 4000-4500, 4500N5200, 6800-7200 and 11000-12000cm-l region, which reflected N-H, C-H, C=O and O-H.Optimal analysis region was studied with support vector machines.Separated whole spectra into little regions, the results showed that 8000N12000cm-1 has better classification results to lung cancer ＋/-.Furthermore, 1080ON11200 cm-1 has the best performance.Study the absorption peak, it can be seen that 6800-7200 and 11000-12000cm-1 can more accurate classification lung cancer ＋/-.Therefore, 7000-12000cm-1 can be set to analyze lung cancer serum samples.