| Fourier transform infrared spectroscopy is a fast spectral scan,which has a wide range of samples,gas,liquid,solid,suspension,elastomer,etc.can be measured by infrared,and enjoys the "fingerprint spectrum of organic compounds",the technology has been widely used in biomedicine.In this study,the Fourier transform infrared spectroscopy signals of serum samples of breast cancer patients and control groups were taken as the research object,and combined with the pattern recognition algorithm,screening models of invasive ductal carcinoma and non-invasive ductal carcinoma in breast cancer patients were established.The main work in this study is as follows:1.The use of Fourier transform infrared spectroscopy combined with pattern recognition algorithms is the first time to screen for invasive ductal carcinoma and noninvasive ductal carcinoma in breast cancer.After processing the original serum Fourier transform infrared spectroscopy data,the principal component analysis(PCA)is used to reduce the dimensionality of the high-dimensional spectral data,and finally the classification model of the support vector machine(SVM),decision tree(DT)and K nearest neighbor(KNN)are established respectively.The highest accuracy rates obtained by these three models are 95.7%,78.6% and 95.7%,respectively.The experimental results show that the use of Fourier transform infrared spectroscopy combined with pattern recognition algorithms is feasible in the screening of invasive ductal carcinoma and non-invasive ductal carcinoma.2.This paper further uses Kernel Principal Component Analysis(KPCA)feature extraction algorithm for dimensionality reduction analysis,combined with DT,KNN and SVM classification models for verification analysis.The ID3 and C4.5 algorithms are used in the DT algorithm,and the highest accuracy rates obtained by the two algorithms are 77.1% and 85.7%,respectively.The accuracy of different k values is verified in the KNN algorithm.K is from 1 to 20.When k = 4,the highest accuracy rate is 98.6%.In the SVM algorithm,different kernel functions are used for analysis,including linear kernel functions,polynomial kernel functions,and RBF kernel functions.The highest accuracy rates of the three kernel functions are 88.6%,100%,and 97.1%,respectively.It shows that the KPCA feature extraction algorithm combined with the pattern recognition classification algorithm has a better screening effect for breast cancer serum Fourier transform infrared spectrum signal.The results of this study show that Fourier transform infrared spectroscopy signals combined with PCA-SVM,PCA-DT,and PCA-KNN models can effectively distinguish between invasive ductal carcinoma and non-invasive ductal carcinoma in breast cancer.It provides a new method for clinical classification of breast cancer.At the same time,it also shows that the use of KPCA feature extraction algorithm has a better dimensionality reduction effect on the experimental data of this study than the PCA algorithm,and improves the classification accuracy and efficiency of the model.And it provides a new idea for the development of a screening method for invasive ductal carcinoma and noninvasive ductal carcinoma based on Fourier transform infrared spectroscopy. |
