Research On Key Techniques Of Biosensors Based On Upconversion Luminescence

An upconversion luminescence biosensor that utilizes upconversion nanoparticles（UCNPs）as labels,consists of upconversion nanoprobes,an immunochromatographic strip,and an upconversion analyzer.Due to the advantages of easy operation,rapid detection,high sensitivity,and high stability,the upconversion luminescence biosensor has broad application prospects in various fields of disease diagnosis,food safety,environmental protection,and national security.With the aim of upconversion luminescence biosensor for the detection of macromolecule biomarker cardiac troponin I（c Tn I）and small-molecule biomarker methamphetamine（MET）,this dissertation focuses on the method for optimizing the performance of upconversion nanoprobes,the preparation of immunochromatographic strip,and the design of upconversion analyzer.However,the upconversion luminescence biosensor cannot accurately distinguish weakly positive and negative samples.This dissertation also focuses on the techniques to improve the sensitivity of qualitative detection and the accuracy of quantitative detection for distinguishing weakly positive and negative samples.The main contributions of this dissertation are as follows:1.Research on methods to optimize the performance of upconversion nanoprobes.Upconversion nanoprobes consist of UCNPs and biorecognition molecules.Initially,the incomplete crystal form and size distribution difference of UCNPs affect the detection results.The key factors of material ratio and high temperature reaction time for the preparation of core-shell UCNPs by the thermal decomposition method were improved.The core-shell UCNPs with goodβcrystal form,dispersibility,and uniform size distribution were prepared to achieve stable upconversion luminescence under the excitation of 980 nm near-infrared light.Additionally,there is no hydrophilic group on the surface of core-shell UCNPs to couple biorecognition molecules.The silica coating method was improved for hydrophilic modification of core-shell UCNPs through surface silicidation,amination and carboxylation to prepare UCNP@Si O₂-COOH with a dispersion coefficient of less than 0.15.Finally,the above optimization methods were verified to lay a foundation for the design of upconversion luminescence biosensor.2.Design of upconversion luminescence biosensor for the detection of c Tn I and MET.For the detection of c Tn I,upconversion nanoprobes UCNP–c Tn I and sandwich-based immunochromatographic strips were prepared.For the detection of MET,upconversion nanoprobes UCNP–MET and competitive-based immunochromatographic strips were prepared.Then,the wavelet transform denoising method was applied to suppress the noise signal,and its denoising effect was analyzed.Compared with median filtering and bilateral filtering denoising methods,the wavelet transform denoising method had the highest peak signal-to-noise ratio and the best denoising effect.Finally,the performance of upconversion luminescence biosensor was verified by testing c Tn I and MET in terms of response characteristic curve,reproducibility,specificity,and reliability.The results showed that the upconversion luminescence biosensor had good reproducibility,specificity,and reliability for the detection of c Tn I and MET.Additionally,the linear range for c Tn I was 0.1-50 ng/m L with the correlation coefficientR² of 0.9891.The linear range for MET was 0.1-100 ng/m L with the correlation coefficientR² of0.9869.3.Research on the support vector machine（SVM）algorithm to improve the sensitivity of qualitative detection.Although the upconversion luminescence biosensor showed good performance for MET detection.Using only the T/C ratio of the upconversion luminescence intensities T to C of the test and control lines to distinguish weakly positive and negative samples renders the results inaccurate because of the noise interference.Therefore,a SVM algorithm was proposed for two-classification task to qualitatively distinguish weakly positive and negative samples,further improving the sensitivity of qualitative detection.The histogram of oriented gradient algorithm was used to extract the features of image samples,and a two-classification model based on SVM was constructed.By testing 60 image samples,the accuracy,the precision,and the recall of SVM algorithm for two-classification of weakly positive（0.05,0.1,and 0.5 ng/m L）and negative（0 and 0.01 ng/m L）image samples were 98.3%,97.3%,and 100%,respectively,which were higher than the accuracy（78.3%）,the precision（79.5%）,and the recall（86.1%）of the T/C ratio method.The results showed that the SVM algorithm improved the sensitivity of qualitative detection.4.Research on the convolutional neural network（CNN）algorithm to improve the accuracy of quantitative detection.The SVM algorithm realized qualitative analysis for two-classification of weakly positive and negative image samples,further improving the sensitivity of qualitative detection.However,it cannot achieve quantitative analysis of weakly positive and negative samples.Therefore,a CNN algorithm was proposed for multi-classification of weakly positive and negative image samples,further improving the accuracy of quantitative detection.A multi-classification model based on CNN that consisted of one input layer,two convolutional layers,two pooling layers,two fully connected layers,and one output layer was constructed.By testing 200 images samples,the precisions of the CNN algorithm for multi-classification of 0,0.01,0.05,0.1,and 0.5ng/m L image samples were 90.5%,87.5%,89.2%,92.9%,and 100%,respectively,which were higher than those of the SVM algorithm（78.6%,66.7%,74.4%,76.7%,and 94.6%）.The recalls of the CNN algorithm for multi-classification of 0,0.01,0.05,0.1,and 0.5ng/m L image samples were 95.0%,87.5%,82.5%,97.5%,and 97.5%,respectively,which were higher than those of the SVM algorithm（82.5%,65.0%,72.5%,82.5%,and87.5%）.The accuracy of the CNN algorithm was 92%,which was higher than that of the SVM algorithm（78%）.The results showed that the CNN algorithm improved the accuracy of quantitative detection.