|
PLS-ANN Discriminant Analysis on Autofluorescence Spectra to Identify Gastric Cancer
MA Jun1 , SHI Xiao-feng1 , ZHENG Rong-er1 , ZHU Yu-ping1 , LI Ying1 , MAO Wei-zheng2 , MENG Ji-wu1
(1. Department of Physics, Ocean university of China, Qingdao 266071, Shandong, China；
2. Depatment of General Surgery, Medical College Hospital of Qingdao University, Qingdao 266003, Shandong, China)
Abstract: Measurement of fluorescence spectra was performed at excitation wavelength of 308nm and emission wavelength in the range of 328-596nm. The partial Least-squares and artificial neural network (PLS-ANN) method was used to analyze autofluorescence 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 but 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.
|
