| It is crucial to detect, diagnose, and treat tumours at their early stage for improving the cure rate of cancer and the survival rate of patients. Laser induced fluorescence spectroscopy has the potential to detect early staged cancer in vivo. Exogenous 5-Aminolevulinic acid (5-ALA) is hope of enhancing fluorescence intensity of tissue and improving the diagnostic rate. So the studies of the mechanism of exogenous 5-ALA-induced protoporphyrin IX (5-ALA-Ppâ…¨) selectively accumulated in tumours and the optimization of 5-ALA dose and detecting time are significant in this field. In this paper, on the models of multifocal colonic carcinomas at all stages of SD rats, the above problems were studied in a systematic way. The main contributions were summarized as follows:1. Colonic carcinoma animal models were setup. Multifocal colonic carcinomas at all stages were induced by weekly peritoneal injections of 1,2-dimethylhydrazine-dihydrochloride (1,2-sym DMH, dose of 25mg/kg body weight (BW)) over 20 weeks in 50 male SD rats and 50 female ones. These models included 205 advanced carcinomas (AC), 121 early carcinomas (EC), 88 dysplasias and 228 normal tissues.2. The content of 5-ALA-Ppâ…¨ in blood and tissue of SD rats were quantitatively measured and the mechanism of 5-ALA-PpK selective accumulation in tumours were analyzed. In order to analysis the distribution of 5-ALA-Pp IX in blood and tissue, two different methods were employed. One was the combination of chemical quantitative extraction assay with fluorescence spectrophotometer. The other was direct spectra measurement of laser-induced fluorescence in vivo. The influences of the dose of 5-ALA intravenous injection (15, 25, 30 and 40 mg/kg BW) and the interval time after injection (1, 2, 2.5, 3 and 4 hours) on 5-ALA-PpK were compared and discussed. Especially after 2. 5 hours of intravenous injection of 5-ALA in 25mg/kg BW dose, the selective accumulation of 5-ALA-Pp â…¨ in various tissues, such as AC, EC, dysplasia and normal colontissue, were studied in detail. The results indicated that ?the optimal 5-ALA dose was 25mg/kg BW intravenous injection and the optimal interval time was 2. 5h. ?The distribution of 5-ALA-PpK was different in various pathological staged colon tissues. The rate of mean 5-ALA-PpK concentration of AC and EC to normal tissue was 1. 35 and 1. 22 respectively in colon tissue, and the fluorescence intensity rate was 4. 56 and 2. 04 respectively. ?The concentration of 5-ALA-Pp IX in small intestine was maximum among small intestine, colon, liver and kidney. It verified the guess proposed by Peng et al about the distribution and metabolize of 5-ALA-PpIX in tissues.3. The fluorescence spectra process mode "preprocess - features extraction - model classification" was employed. And the corresponding methods were developed to each procedure.4. In the preprocess procedure, each fluorescence spectrum was normalized to the mean of its spectrum surface. Thus the weak fluorescence signals were amplified properly and could compare with each other, which were sampled from different pathological staged colon tissues of various of SD rats.5. In the features extraction procedure, two different methods were developed: principal component analysis (PCA) and wavelet multiresolution analysis. Compared with covariance PCA, the adaptive neural network PCA could extract fluorescence spectra information more effectively. Through wavelet multiresolution analysis, fluorescence spectrum signal was divided into 5 level signals, of which 3 level were reserved. Each level signal was separated into 4 feature bands.6. In the model classification procedure, two classification methods were also developed: multivariate Logistic regression and BP neural network.a) Using PCA-Logistic method, the influence of 5-ALA dose and detecting time on classification were compared and analyzed. The results showed the classification accuracy of fluorescence spectra with the dose of 25mg/kg BW and 2. 5h were the highest. It further verified that it was very important to optimize the5-ALA dose and detecting time.b) With fluorescence spectra features and BP neural network (BPNN) , all fluorescence spectra were classified into two categories: normal and lesion which included dysplasia, early cancer, and advanced cancer. The results showed that the predict properties of 6 PCs-BPNN, 12 wavelet features-BPNN, and 6 PCs + 3 wavelet features BPNN were similar. The best predict sensitivity and specificity were 97. 33% and 94. 44%. And the sensitivity to early cancer was 100%.c) With fluorescence spectra features and BPNN, all fluorescence spectra were classified into four categories: normal tissue, dysplasia, early cancer and advanced cancer. The results showed that the predict property of 6 PCs-BPNN, 12 wavelet features-BPNN, and 6 PCs + 3 wavelet features-BPNN were very different. Among the 3 neural network structures, the last one had the best predict accuracy rate. And the predict accuracy rates to normal tissue, dysplasia, early cancer and advanced cancer were 87.04%, 61.54%, 88.10% and 86.59 respectively.
|
PDF Full Text Request