Investigations On Near-field Photon Transport Theory And Nonlinear Image Reconstructions For LOT

Imagings for cervical and skin tissue require are generally characterized by high sptial-resolution and small detecting depth.However,present mesoscopic imaging techniques,e.g.,the optical coherence tomography(OCT),fail to provide functional information,which might make them lose the ability of(early-stage)cancer detection.Laminar optical tomography(LOT)is a new mesoscopic functional imaging technique,which utilizes a confocal microscopy-based setup to allow noncontact imaging with 100~200?m spatial-resolution over depths of 0~2.5mm,and thus is emerging as an extremely powerful tool to resolve melanin,hemoglobin,and fluorochromes in superficial tissues.However,the inverse problem in typical LOT image reconstruction is assumed to be linear,which seriously decrease its availability and reliability.For the crucial step of reconstruction—forward problem,three models for describering the near-filed photon migration are proposed in this paper and demonstrated to be applied in LOT image reconstruction;For the inverse problem,a nonlinear LOT absorption image reconstruction is proposed and validated through phantom and animal experiments.Innovations in this paper are summarized as follow:1.Forward model based on the TT&TVR scheme is proposed to improve the efficiency of the traditional MC-based forward modeling for LOT.In this scheme,the predictions at all detection sites can be achieved through only one single MC,which takes only 1/10 of the original computational cost and 1/6.9?10~4 of original memory(1?10~7 emitted photons).The good performances of the reconstructions are demonstrated numerically.2.A semi-analytical forward model based on VS-DA is proposed to address on the high computational complexity of most analytical near-field modelings in inhomogeneous medium.In this method,a series of isotropic point sources along the incident direction are set to replace the original collimated light source.Their intensities and positions are solved with the VS equation,which is established according to the uniqueness and superposition principles.The VS-DA model is proved to inherit the mathematical simplicity of the DA approximation while considerably extending its validity in modeling the near-field photon migration in low-albedo medium.The superiorities of the VS-DA are validated through comparisions with several existed near-field models.3.Coupling between the transport theory and its diffusion approximation in the subdomain-based hybrid models for enhanced description of the near-field photon-migration can be computationally complex even physically inaccurate.A physically-consistent novel coupling method that links the transport and diffusion physics of the photons according to TPK scheme is proposed,where distribution of the fully diffusive photons at a transition time is provided by a computationally-saving auxiliary time-domain diffusion solution to serve as a complementary or complete isotropic source of the temporally-integrated transport equation over the early-stage and the diffusion equation over the late-stage,respectively,from which the early and late photo-densities can be acquired independently and summed up to achieve a steady-state modeling of the whole transport process.The proposed scheme is validated with numerical simulations for a cubic geometry.4.Since mesoscopic tomography depends on dense samplings and small source-detector seperation,the transport-based nonlinear reconstruction is normally computationally expensive.By taking advantage of SVD for efficient inverse calculation,we herein propose a practical nonlinear absorption reconstruction for LOT,where the time-consuming SVD inversion of the sensitivity matrix is accelerated by explicitly recursive computations with enhanced scaling scheme.The overall computational complexity of solving the SVD inversion is sharply reduced.The proposed method is applied in LOT with phantom and in-vivo experiments.The near-field models of photon migration and the nonlinear image reconstruction proposed in this paper can be utilized not only for LOT,but also for other mesoscopic imagings,as well as high-density and/or bulk imagings.