PLS-ANN判别分析自体荧光光谱识别胃癌

PLS-ANN Discriminant Analysis on Autofluorescence Spectra to Identify Gastric Cancer

本文对58例胃癌病人离体标本的癌浆膜和正常浆膜进行以308nm为激发光的自体荧光光谱检测,采用多因素分析法进行光谱信息提取,以识别胃癌。研究表明偏最小二乘法结合神经网络法(简称PLS-ANN)进行判别分析,诊断胃癌的灵敏度为86%,特异度为100%,准确率为93%,有望成为手术中快速识别胃癌在胃壁的浸润范围的有效方法。

Measurement of fluorescence spectra was performed at excitation wavelength of 308 nm and emission wavelength in the range of 328 nm -596 nm. The partial Least-squares and artificial neural network （PLS-ANN） method was used to analyze autofluorescenee spectra of gastric cancer. The 58 cancer samples and normal samples were taken from stomach serosa. The normalized and centerized spectra of two kinds of samples showed similar hut divergent patterns. PLS-ANN classification algorithm could differentiate cancer tissues from normal tissues with a sensitivity of 86%, a specificity of 100% and a total success rate of 93%. We concluded therefore that the PLS-ANN method was a fast, more effective choose for identification of gastric cancer.