Application Of Serum Raman Spectroscopy Combined With Multi-classification Algorithm In The Detection Of Ductal Carcinoma In Situ And Breast Cancer

Ductal carcinoma in situ(DCIS)and breast cancer are common breast diseases in women and pose a serious threat to women’s life and health.Early screening of DCIS and breast cancer is helpful to develop targeted treatment plans for patients in time.Due to the long time cycle and high cost of traditional medical detection methods,it is of great significance to find a rapid and low-cost method to detect DCIS and breast cancer.Since Raman spectroscopy can provide molecular structure information of biological samples and has the advantages of simple operation,high sensitivity and high speed,it has been widely used in biomedical and other fields.In this study,we collected the serum Raman spectra from 241 healthy individuals,463 breast cancer cases and 100 DCIS cases,and established three classification models to classify the spectra for the detection of DCIS and breast cancer.The main content of this paper is as follows:1.Raman spectroscopy combined with KNN model was used for the study of early screening of DCIS and breast cancer.Firstly,the acquired raw spectra were first preprocessed by Savitzky-Golay algorithm,airPLS algorithm,and min-max normalization,and the molecular assignments of the main feature peaks were given by Raman spectroscopy analysis.Then,the SPXY algorithm was used to divide all the data into a training set and a test set,and a KNN diagnostic model was built.Finally,the classification performance of the model was tested via the test set,and the final classification accuracy of the model was obtained as 88.02%,and the AUC value of the ROC curve was 0.91.This indicates that the classification performance of the KNN model is good and can be used in the early screening of DCIS and breast cancer.2.Raman spectroscopy combined with SVM model was used for the study of early screening of DCIS and breast cancer.In order to compare the classification effect of SVM with different kernel functions,we selected the linear,polynomial and Gaussian radial basis kernel functions to build SVM diagnostic models.And the final classification accuracies of the three models on the test set were 94.63%,90.5%and 95.46%,and the AUC values of ROC curves were 0.993,0.983 and 0.994,respectively.The results showed that the SVM model had the best classification performance when the Gaussian radial basis kernel function was chosen.3.Raman spectroscopy combined with 1D-CNN model was used for the study of early screening of DCIS and breast cancer.In general,the common CNN is a two-dimensional CNN used for image recognition,which uses the two-dimensional matrix representing the grayscale and color of the image as input data.Since Raman spectrum is a one-dimensional signal,a 1D-CNN classification model suitable for Raman spectrum classification is established in this study.On the test set,the model finally achieved a classification accuracy of 98.76%,with an AUC value of 0.999 for its ROC curve.From the results,it can be seen that the classification performance of the 1D-CNN model is better than that of KNN and SVM models,and it has the best effect on the screening of DCIS and breast cancer.The above research work shows that the combination of serum Raman spectroscopy and multi-classification algorithm has great potential in the early detection of DCIS and breast cancer,and can be an effective adjunctive diagnostic tool.