| Optical molecular tomography(OMT),as a promise imaging modality,has been continuously developing and contributing to biomedical research and applications with high detection sensitivity,high specificity,high resolution,low cost,and simple equipment.However,due to the strong scattering property of biological tissues,the limited boundary measurements and the complexity in biological systems,the inverse source reconstruction in optical molecular tomography is severely ill-posed.And the accuracy,stability,resolution and the analysis process of the reconstruction results are still facing great challenges.This paper focuses on the inverse algorithms,strategies to reduce the ill-posedness,multiple-source resolution and clustering analysis and some in vivo experiments.The contents and innovations of this work are as follows:1.Inverse reconstruction algorithms based on compressive sensing.The ill-posed problem of OMT was commonly solved by some greedy methods and convex sparse regularization techniques.In this study,the compressive sampling matching pursuit(CoSaMP)algorithm and the LO.5-norm regularization method based on homotopy were also developed to improve the accuracy of OMT.To evaluate these algorithms,a series of numerical simulation studies on a 3D digital mouse atlas had been carried out,including single source reconstruction,double source reconstruction and stability analysis experiments.The comparison results demonstrated that the CoSaMP had a better performance than other three algorithms based on greedy strategy and the L0.5-norm method outperformed the comparative methods in terms of location accuracy,source quantitation and the stability and sparsity of the reconstructed results.2.Reconstruction strategies for reducing the ill-posedness.The multispectral information-based algorithm,permissible source region method and optimize finite element mesh method were three commonly utilized strategies to reduce the ill-posedness.In this study,based on the light propagation theory and permissible source region method,an adaptive hp finite element method(hp-FEM)combined with the simplified spherical harmonics equations(SP3)model was reported.Beside that,in Cerenkov luminescence tomography,a modified weight multispectral CLT(wmCLT)reconstruction strategy was developed,by considering the tissue optical properties and the multispectral weight.In experiments parts,we made some comparison of the SP3 model and some other modes,h-FEM and hp-FEM,traditional multispectral,hybrid-spectral reconstruction strategies and wmCLT and got the conclusion that these two strategies could significantly improve the accuracy of reconstructed source location and intensity.3.Multiple-source resolutionand clustering analysis.Because of the ill-posedness of the inverse problem and the diversity of tumor biological characteristics,the multiple-source resolution andanalysis are meaningful but challenge problems.In this study,based on the compression perception theory,a non-convex sparse regularization algorithm(nCSRA)framework was proposed to improve the capacity of multiple-source resolving in aspect of spatial resolution,intensity resolution,and size resolution.Moreover,the multiple-source analysis was treated as an unsupervised pattern recognition problem based on the reconstruction results.Further more,a novel hybrid clustering algorithm combining the advantages of affinity propagation(AP)and K-means was developed to identify multiple sources automatically.The clustering results indicated that the proposed method can overcome the limitation of human vision and automatically determine the source numbers and source regions objectively.4.The study of in vivo small animal experiments.In vivo small animal experiment as a major criteria and an essential step to testify the effectivity of the reconstructed methods in OMT.In this study,the common imaging process of the in vivo experiments and the Imaging system have been introduced in detail.According to the previously mentioned methods,some in vivo experiments have been conducted to verified the effectivity and reliability of all proposed reconstructed methods in our study,including the homogeneous cubic phantom experiment,source reconstruction based on the implantation mouse model,bladder reconstruction,in vivo imaging of xenograft tumors and multiple-source resolutionand analysis experiments.Furthermore,the in vivo experiments also would provide a new way in improvement and transformation of OMT reconstruction methods. |
