PLS-ANN分析血浆自体荧光光谱的二阶导数光谱识别胃癌

PLS-ANN Discriminant Analysis on the Second Derivative Spectra of Autofluorescence Spectra of Plasm to Identify Gastric Cancer

探讨偏最小二乘法结合神经网络法(简称PLS-ANN)分析血浆自体荧光光谱的二阶导数光谱识别胃癌的优势,对20例胃癌病人和23例健康人血浆进行以405nm为激发光的自体荧光光谱检测,采用PLS-ANN法分别对血浆自体荧光光谱和二阶导数光谱进行判别分析。PLS-ANN分析血浆的自体荧光光谱法诊断胃癌的灵敏度为75%,特异度为83%,准确率为79%,PLS-ANN分析血浆的二阶导数光谱法诊断胃癌的灵敏度为90%,特异度为96%,准确率为93%。结果表明PLS-ANN分析血浆的二阶导数光谱法识别胃癌,优于PLS-ANN分析血浆的自体荧光光谱法,有望成为快速识别胃癌的较理想方法。

To study the advantage of using partial Least-squares and artificial neural network （PLS-ANN） m analyze the second derivative spectra of plasm of gastric cancer autofluorescence spectra, we measured 20 pieces of plasm of gastric cancer and 23 pieces of health control autofluorescence spectra at the excitation wavelength of 405nm and emission wavelength in the range of 425 - 700nm. The PLS-ANN method was used to analyze the autofluorescence spactra and the second derivative spectra of plasm of gastric cancer. The classification algorithm using PLS-ANN to analyze the second derivative spectra could differentiate cancer plasm from normal plasm with a sensitivity of 90 96, a specificity of 96 96 and a total success rate of 93 96. The result is superior to that from PLS-ANN analyzing the autofluorescence spectra, which gave 75 %, 83 96 and 79 96, respectively. We concluded therefore that the method was faster and more effective for the identification of gastric cancer.