Research On Key Techniques For Human Pathological Tissue Detection Based On Terahertz Time-domain Spectroscopy

Terahertz time-domain spectroscopy(THz-TDS)is a rapidly developing non-destructive testing technology,which has sensitive responses for water and biomolecules.Thus it has bright prospect in medical diagnosis,biological medicine,food safety fields and so on,and draws widespread attentions in domestic and overseas.With the support of the National Science Foundation of China,this dissertation takes gastric cancer as investigated object and carries out a series of work on pathological tissue detection with THz-TDS.The work of this paper mainly focuses on the identification of pathological tissue through data-driven and spectroscopy mechanism analysis approaches.The main contributions and innovations of this dissertation are as follows:(1)The comparison and verification of parameter responses of pathological tissue in terahertz field.Fresh and paraffin-embedded tissues are detected with THz-TDS,and the response differences between pathological and normal tissues are analyzed.The absorption coefficient spectra,refractive index spectra and complex permittivity spectra of different human tissues are compared.The Double Debye model is used to study the relaxation processes of pathological and normal tissues,and static dielectric parameter,median dielectric parameter,infinite dielectric parameter and relaxation times are calculated with the grid method.Based on the research of relaxation effect,Finite Difference Time Domain(FDTD)is employed to simulate the propagation processes of THz wave in human tissue as a dispersive medium.The phase and peak of transmitted THz wave are analyzed to explain the optical property differences between pathological and normal tissues.(2)The quick identification of pathological tissue with TDS-TDS.A full wavelength discrimination model based on optical and dielectric parameters is built.The correlation between model parameter and pathological tissue is analyzed to improve the sensibility of full wavelength model.To deal with the individual variation,differential correction pretreatments problem,including the collinearity between variables of full wavelength spectra,data dimension reduction and the generalization ability of model with small sample volume,are mainly studied.As a result,the proposed tissue identification model presents great accuracy and generalization ability.(3)The tumor biomarker information extraction from THz spectra data of pathological tissue.Human tissue is a kind of multicomponent mixture,and the mixture spectra have overlap and peak shift phenomenon.This dissertation proposes a resolving method based on Self-Modeling Mixture Analysis(SMMA)to work out the deconvolution problem of multicomponent mixture system to extract tumor biomarker information.In addition,the THz absorption spectra of many materials have base lines,leading to determination difficulty of pure variables with the contribution from one material.To deal with the situation,we employ the inverted second order derivative to conduct spectral pretreatment to improve resolution performance.The result of the unmixing method is compared with Constrained Non-negative Matrix Factorization(CNMF).It is demonstrated that SMMA has unique and good resolution result in contrast to the pure mathematics method called CNMF.(4)The explaination and identification optimization of pathological tissue using THz-TDS.We propose an analytical method based on spectral resolution technique that the resolving results are associated with disease mechanism.Second order derivative pretreatment is employed to decrease the influence of materials with no characteristic absorption and reserve the absorption properties of human tissue as well.SMMA is used to decompose the preprocessing data,and the subspace spectra of characteristic materials with the same trend in the course of disease are extracted,and the correlation analysis between the resolving characteristic peaks and the absorptions of potential biomarkers are carried out with the help of pathologic knowledge,realizing more nuanced identification and analysis for human tumor.In a word,this dissertation studies several problems about pathological tissue detection using THz technique including optical and dielectric properties of pathological tissue analysis,mechanism study between THz wave and human tissue based on FDTD,spectral resolution investigation for multicomponent mixture system and pathological tissue identification technique.The study of this dissertation not only achieves quick and preliminary identification but also discriminates pathological tissue based on disease mechanism.The tissue samples and pathological analysis in this research are supplied by the Second Affiliated Hospital of Zhejiang University School of Medicine.