大鼠胰腺及癌组织红外光谱连续小波特征提取及径向基人工神经网络识别  

Continuous Wavelet Feature Extraction of Pancreatic Normal and Cancerous Tissue's Infrared Spectra of Sprague-Dawley Rats and Identification of Radical Basis Function Neural Network

作　　者：程存归[1] 田玉梅[1] 张长江[2] 

机构地区：[1]浙江师范大学化学系,浙江省固体表面反应化学重点实验室,金华321004 [2]浙江师范大学电子信息工程系,金华321004

出　　处：《分析化学》2008年第8期1051-1055,共5页Chinese Journal of Analytical Chemistry

基　　金：浙江省自然科学基金资助课题(No.3014478)

摘　　要：采用水平衰减全反射(HATR)傅里叶变换红外光谱法(FTIR)测定了SD大鼠胰腺正常组织与非正常组织的谱图,提出了一种新的基于FTIR的连续小波特征提取与径向基人工神经网络分类方法以提高FTIR对早期SD大鼠胰腺癌的诊断准确率。利用连续小波多分辨率分析法提取FTIR特征量,对于提取的特征量采用径向基函数神经网络进行模式分类。对SD大鼠的胰腺正常组织、早期癌组织及进展期癌组织的FTIR,利用连续小波多分辨率分析法提取9个特征量,进行RBF神经网络分类判断。当目标误差为0.01,径向基函数的分布常数为5时,网络达到最优化,总的正确识别率为96.67%。并对影响分类结果的网络参数、目标误差和分布常数对分类样品的影响做了讨论。实验结果表明:此方法对早期胰腺癌具有较高的诊断率。Sprague-Dawley （SD） rat＇s normal and abnormal pancreatic tissues are determined directly by horizontal attenuated total reflectance Fourier transform infrared spectroscopic （HATR-FTIR） method. In order to diagnose rate earlier stage SD rat＇s pancreatic cancer with FTIR, a novel method of extraction of FTIR feature using continuous wavelet （CW） analysis and classification using the radial basis function neural network （RBFNN） was developed. FTIRs were collected from SD rats, and after preprocessing, 9 feature variants were extracted with continuous wavelet analysis. The differences between normal and abnormal samples were identified by radial basis function neural network based on the indices of 9 feature variants. When error goal was 0.01 and spread constant of radial basis function was 5, the network achieved the optimization, the total correct rate was 96.67%. It was practical to apply radial basis function neural network on the basis of HATR-FTIR to identify abnormal tissues. The effects of network parameters including error goal and spread constant, were investigated. The research result shows the feasibility of establishing the models with FTIR- CW-RBFNN method to identify normal, early carcinoma and advanced pancreatic cancer.

关 键 词：水平衰减全反射傅里叶变换红外光谱法 小波特征提取 径向基函数神经网络 胰腺癌 

分 类 号：R735.9[医药卫生—肿瘤]