Study On Mechanism Of Laser Induced Auto-fluorescence In Colorectal Cancer And Method Of Spectrum Classifying

Objective: To investigate mechanism of the auto-fluorescence spectrum in colorectal cancer detection, and Partial-least-square method was applied in analysis of auto-fluorescence spectrum.Methods: The auto-fluorescence spectra from tissues of colorectal cancer were taken by auto-fluorescence detection system. The quantities of riboflavin from normal and cancer tissues were measured by HPLC to evaluate relations between the auto-fluorescence spectrum and fluorescent material.Partial-least-square method was also used in the auto-fluorescence spectrum classifying, which belongs to the pattern recognition. By using partial-least-squares to analyze characters of the fluorescence spectrum, parameters of the spectrum were reduced to some main factors which were taken as the input of artificial neural network. Through the training and prediction, artificial neural network outputs classified spectrum. At the same time, the work of building mathematical model and optimizing the algorithm were completed.Results: Obtained the auto-fluorescence spectra and quantity of riboflavin in the corresponding tissue; the mean quantity of riboflavin in cancer and normal tissue are 609 ± 57 ng/g, 786±91 ng/g respectively. Auto-fluorescence spectrum classifying using Partial-least -squaremethod, together with artificial neural network, received specificity 92.3 %, sensitivity 96.2 % and reliability 94 %.Conclusions: Significant differences (p<0.001) in the quantity of riboflavin between cancer and normal tissue were found, which is important for better understanding mechanism of the auto-fluorescence spectrum. Partial-least-square method together with artificial neural network offered an efficient way to study the application in the auto-fluorescence classifying for diagnosis of colorectal cancer.