The fluorescence spectra overlap of normal serum and hypetriglyceridemia serum will result in low recognition rate. The paper reports a recognition method of hypertriglyceridemia serum using fluorescence spectroscopic analysis based on principal component analysis and probabilistic neural networks. Firstly, fluorescence spectra of normal serum and hypertriglyceridemia serum were measured with the excitation wavelengths of 260, 370, 580 nm. And the sample's initial feature vector was defined according to fluorescence intensities. Secondly, principal component analysis was used to extract the initial feature vector and establish new sample's feature vector according to the cumulate reliability (>95%). Finally, the four-layer probabilistic neural network was founded for serum samples' recognition. Study designed seven groups of experimental scheme by choosing different fluorescence spectrum for recognition effect comparison.The result showed thatrecognition rates of the normal serum and hypertriglyceridemia serum were 100% and 95% respectively using fluorescence spectra with the excitation wavelengths of 260 nm and 370 nm. The feasibility and effect of the recognition scheme were demonstrated. The study provides an important guidance for developing the recognition technique of hypertriglyceridemia serum using fluorescence spectroscopic analysis based on probabilistic neural networks.
展开▼