Research On Detection Method Of In Vivo Tissue Based On Multispectral Imaging

The spectral imaging technology,which can acquire the structure and spectral domain information of in vivo tissue simultaneously and has the advantage of high resolution,large imaging range and abundant information,is widely used in biomedical field.For the detection and imaging of in vivo tissue,a compact imaging system with fast imaging speed and low cost is the key to expand the application of spectral imaging technology.In this thesis,a multispectral imaging system is built based on the narrow-band filters and spectral reconstruction algorithm,and their feasibility and effectiveness are investigated by Monte Carlo simulation,phantom experiments and in vivo experiments.Firstly,the principle of multispectral imaging technology is introduced,and specific narrow-band filters are selected as the beam splitting component to bulid the multispectral imaging system.The optical path and control software of the imaging system are optimized to achieve the matching of different spectral channels.Then,the precision and the stability of the measurement system are evaluated to determine the optimal conditions.In addition,a multispectral images reconstruction method based on Wiener estimation is proposed to reconstruct full-spectrum high-dimensional spectrum data from the discrete,finite-wavelengths data with narrow-band.Secondly,based on the Monte Carlo simulation model,the look-up-table forward model combined with the inverse method of nonlinear optimization is proposed to estimate the tissue optical parameters.Three kinds of phantoms with different concentrations of Evens Blue are prepared,and multispectral images of the phantom are acquired,then the diffuse reflectance of the full spectrum is obtained by Wiener estimation.A forward model is established based on the optical properties of the phantom,and the multivariate nonlinear optimization method is used to reconstruct the optical parameters of the phantom.The results show that,the average root mean square error of absorption coefficient is 5.56%,and the root mean square error of reduced scattering coefficient is 2.55% at 500-700 nm.The accuracy of the system and the inverse algorithm are verified.Finally,the detection methods of in vivo blood oxygenation based on multispectral imaging are studied.The multispectral images of the mouse ear region are collected by the imaging system,combined with Wiener estimation and second derivative method,the average oxygen saturation of blood vessel region is90%±2%.The oxygen saturation of tissue is 64%±4%.Additionally,considering the needs of in vivo blood oxygenation,a tissue forward model with blood oxygenation changes is established based on the human skin structure,composition and optical properties.The blood oxygen saturation is calculated by nonlinear optimization and second derivative method,respectively.The results show that,the changes of blood oxygen saturation are 16.58% and 10.41%,respectively,which is consistent with the results of the fingertip oximetry sensor,thus verifying the accuracy of the system and method in tissue characteristics detection.