Simulation and Analysis of Autofluorescence Spectra for Rectal Colon and Cervix Early Cancer Tissue Diagnosis

• Main Authors: Shou-Chia Chu, 朱朔嘉
• Other Authors: Huihua Kenny Chiang
• Format: Others
• Language: zh-TW
• Published: 2008
• Online Access: http://ndltd.ncl.edu.tw/handle/28423956590479471996

博士 === 國立陽明大學 === 醫學工程研究所 === 96 === In this research, we represented simulated spectra for cervical tissue at different excitation angles, and compared analytic results of three spectroscopic multivariate analysis methods-PCA, PLS, and MVLR for rectal colon cancer and oral cancer. In simulation, we built cervical autofluorescence simulation model. The model contains cervical consist of structure(epi,CIN,and stroma), tissue structure and optical characteristic with development of dyplasia stages, and the optimization of excitation angles and flouresent reception location. We discovered the best reception location is 200μm to excitation location and change of excitation angles is the important effect in differentiating variation of tissue. In the addition, fluorescence energy ratio of NADH/collagen can distinguish between CINⅢ and other kind of tissues. In analysis, the results obtained from use of the three widely used linear multivariate analysis methods showed great promise for fluorescence spectroscopic analysis in distinguishing between the different stages of colorectal and oral carcinogenic development. Both PLS and MVLR performed better in differentiating the normal and dyplasia group than did PCA. In addition, from comparisons of the calibration abilities of the PCA and PLS methods, we have concluded that PLS, which benefits from the cross-correlation of the input- and output-layer variables, could obtain a larger variance than could PCA, which benefits from the auto-correlation of the input-layer variables only. In this research, the results will provide better estimate bases for choice of excitation angles and fluorescent reception location or analytic ability of three multivariate analysis methods.