Label-free Detection Of Echinococcus Based On Serum Raman Spectroscopy Combined With Machine Learning

As a parasitic disease,echinococcosis not only brings great suffering to patients,but also has a poor prognosis.The current detection methods of echinococcosis are expensive and time-consuming.Therefore,finding a cheap and rapid screening method of echinococcosis is of great significance for the prevention and treatment of echinococcosis.Raman spectroscopy,as a non-destructive,inelastic scattering light scattering technology,can reflect the information of the biochemical components contained in the detection object,so it is widely used in the intersection of optics and biomedicine.Machine learning,as a hot technology at present,shines brilliantly in various fields.In this study,we used a portable Raman spectrometer to collect 400 serum Raman spectra of healthy volunteers and 385serum Raman spectra of patients with echinococcosis,and the labelfree echinococcosis detection was realized from traditional machine learning and deep learning.The two main research areas of this paper are as follows:1.Classification and screening of echinococcosis was carried out based on feature extraction combined with traditional machine learning algorithms.Spectra are preprocessed by Vancouver Raman algorithm and norm normalization,and all spectra are divided into training set and test set by Kennard-Stone algorithm.Serum Raman spectroscopy contains a large amount of information,using principal component analysis to reduce the dimensionality of high-dimensional spectral data and make feature selection based on contribution rate.Then,based on traditional machine learning,three classification and recognition models including linear discriminant analysis,logistic regression,and support vector machine were constructed.The optimal hyperparameters of the model were determined by the method of five-fold cross-validation and grid search.Finally,on the test set,the classification accuracy of the three models was 81.83%,91.72%,and 89.17%,respectively.The results demonstrate the feasibility of feature extraction combined with traditional machine learning in echinococcosis screening.2.Deep learning was used to classify and screen echinococcosis.Based on deep learning,two deep learning classification models,deep neural network and improved convolutional neural network,are constructed to classify and recognize the same set of data.The optimal hyperparameters of the two models were determined by five-fold cross verification and grid search,and the optimal model was trained by back propagation combined with gradient descent method.Finally,the classification recognition accuracy of the two models in the test set is 88.54%and 94.90%.The above results demonstrate the great potential of Raman spectroscopy combined with deep learning in echinococcosis screening.The results of this study prove that feature extraction combined with traditional machine learning and deep learning can effectively distinguish echinococcosis patients from healthy volunteers.Compared with feature extraction combined with traditional machine learning,deep learning is more sensitive to the screening of echinococcosis,and convolutional neural network is superior to traditional machine learning in all aspects.The above studies show that serum Raman spectroscopy combined with machine learning for label-free screening for echinococcosis is faster and more accurate,which is expected to provide a new idea for the identification and diagnosis of echinococcosis.