The Applications Of Discrete Shmaliy Moments In The Spectra Analysis

With the rapid development of modern analytical instruments and methods,researchers can obtain a wealth of information from a large number of samples in a relatively short period of time.However,it is difficult to directly analyze multi-target components in complex systems by conventional methods alone.Therefore,how to use other means to extract the most important information from a large amount of data has become a problem that researchers need to solve urgently.The emergence of chemometrics provides convenience for the analysis of complex systems.Discrete Shmaliy moments（DSMs）have the ability to describe global image/curve features,unique multi-resolution,and inherent invariance of rotation,translation,and scaling.Based on the chemical 2D/3D spectra,the discrete Shmaliy curve/image moments can be used to extract the feature information related to the target component in the spectra,thereby achieving the simultaneous quantitative analysis of multiple target components in complex samples,even if there are uncorrected interference signals.In addition,we also applied the image moment method in the feature extraction of medical images for the first time to meet the needs of rapid and accurate diagnosis of diseases.This master’s degree thesis mainly focuses on the practical application of the DSM method,which based on conventional chemistry（2D/3D）spectra and medical images of brain tumors,respectively.These research works are divided into the following three chapters（Chapter 2 to Chapter 4）,the first chapter is the introduction,and then it will be carried out in turn.Chapter 1 Introduction.It mainly introduces the development of analytical chemometrics and its application in conventional chemical spectra analysis,the development of moment methods,and theirs analytical researches and applications in complex systems.Finally,the content of this paper is briefly summarized.Chapter 2 Applications of Discrete Shmaliy Moments on the Quantitative Analysis of Multi-target Compounds Based on the UV-vis and HPLC-PAD SpectraTo extract the features in first-order or second-order signals,the two kinds of Discrete Shmaliy Moment（DSM）methods were proposed and applied to the quantitative analysis of multi-target compounds in complex based on the UV-Vis and HPLC-PAD spectra of samples for the first time.The statistical parameters including correlation coefficient for calibration（R）,correlation coefficient for leave-one-out cross-validation(R_loo-cv),and prediction（R_p）,root mean square errors for calibration（RMSE）,root mean square error for leave-one-out cross-validation(RMSE_cv),and prediction（RMSE_p）to evaluate the robustness and reliability of the established model,as well as the prediction ability and accuracy in practical applications.In addition,we also compared with Tchebichef moment（TM）and other classical methods such as multivariate curve resolution-alternating least square（MCR-ALS）,partial least squares（PLS）regression and N-way partial least squares（N-PLS）,the proposed methods are more convenient and efficient,which not only provides another suitable tool for the quantitative analysis of multi-target components in complex samples,but also extends the application of moment invariants in chemical signal analyses.Chapter 3 An Effective Approach to the Feature Fusion of IR and GC Spectra for the Determination of Nine Components in Red WineIn the research work of the previous chapter,it was a quantitative analysis based on a single spectra/data.However,in some cases,the spectra/data measured by a single instrument contains insufficient information,and it is difficult to make satisfactory quantitative analysis results based on such spectra/data at this time.Therefore,if the information obtained by different instruments is effectively fused to expand the amount of information contained in the samples,it may help to obtain more accurate and reliable quantitative analysis results.In the study of this chapter,we used the discrete Shmaliy curve moment method as the extraction method of the feature information of IR and GC spectra,and combined with two data fusion methods（signal fusion and feature fusion）.Afterwards,the linear quantitative models of target components were established to verify the feasibility of ideas.Compared with the partial least squares（PLS）and Tchebichef moments（TM）methods,the DSM method possesses the more powerful capability of the feature extraction and resolution from chemical spectra.This study indicates that the feature fusion with DSMs is not only more suitable for the analysis of complex samples but also extends the application for conventional analytical instruments.Chapter 4 Magnetic resonance imaging（MRI）combined with discrete Shmaliy moment approach to detect the brain tumorsBased on the magnetic resonance imaging（MRI）of the subject’s brain,combined with the discrete Shmaliy image moment method（DSM）,we propose a new approach to distinguish patients with tumor from healthy people.Brain tumors are one of the common diseases in clinic,and the chance of survival can be increased if the tumor is detected correctly at its early stage.Therefore,accurate detection of tumors is the key to treatment.In this chapter,it is proposed for the first time that the DSM method were employed to extract the feature information of MRI,and support vector machine（SVM）method was used to establish the classification model to detect the brain tumors.The10-fold cross-validation and independent test sets were used to evaluate the model,and four common prediction methods were selected for comparison.The results show that our proposed DSM-SVM approach can accurately discriminated between brain tumor and healthy groups,and it can provide a feasible tool for tumor detection.Chapter 5 Including main conclusions and outlook.