Research on Self-Adaptive Neural Network Identification Modeling of Early-Colorectal Cancer´s Fluorescence Spectra

A rapid, self-adaptive neural network modeling of fluorescence spectra for automatic diagnosis is built up in this paper. Against that the identification of fluorescence spectra can not be better solved by the neural network model featuring flat structure, a neural network integrated model with multiple-structure is proposed in this paper which The font-end of the model is the data preprocessing. Using the fluorescence spectrum produced by the latest developed third generation of laser-induced auto-fluorescence detection system as the experimental data which was compared with that of the first and second generation of equipment system, analysis is made which shows: the test result showed that the fluorescence spectrum´s accuracy of recognition to the early-colorectal cancer can reach 98% above, and the misdiagnosis rate was below 1% and provides a better foundation for the colorectal cancer auto-fluorescence spectra intelligent diagnostic system entering into the clinical application.