Fast MC Modeling And Inverse Method For Determination Of Optical Parameters Of Turbid Medium

The interactions of light and turbid media mainly can be divided into two types ofprocesses, absorption and scattering. Many materials in nature are optically turbidwhere light scattering dominates the interaction, such as ceramic, milk, cellsuspension, most of human tissues, etc. Determination of the optical parameters ofturbid medium samples has wide range of applications in material research andclinical studies. Therefore, the development of efficient tools for accurate modeling ofthe interaction between light and turbid medium is critically important todetermination of optical properties of turbid materials and establishment of thefoundation for design of new instrument. As a result, study of new methods todetermine the optical parameters of turbid media has attracted much attention inrecent decades.Our research consists of the following parts to develop a fast method for inversedetermination of optical parameters of a homogeneous turbid sample from themeasured signals:1. The goal of this thesis research is to develop an inverse algorithm that can becombined with a new multi-parameters spectrophotometer employing a simple opticaldesign for implantation in comparison to the integrating-sphere based complex design.In the new spectrophotometer design the solid angle range afforded by a light detectorfor measuring the diffuse transmittance or diffuse reflectance signals is generally threemagnitudes smaller than the one measured with an integrating sphere. Consequently,it is very theoretically challenging to develop a method for rapid and accuratecalculation of the light signals measured within small solid angle ranges for inversedetermination of the optical parameters of a turbid medium sample using an iterationmethod. Furthermore, successful development of such method can lead to wideranged applications.2. Computational cost is always one of factors which limit the application andaccuracy of Monte Carlo simulation methods despite its advantages of algorithmsimplicity, strong adaptability to solve various boundary-value problems and ease incoding. For calculation of light signals measured within small solid angle range theMonte Carlo method takes very long times to obtain enough detected photons by a detector because of low collection efficiency. Therefore, we need to first develop afast Monte Carlo method. For this purpose, we have adopted and developed aspeed-up Monte Carlo code based on the GPU (Graphic Processing Unit) parallelcomputing approach.3. In order to further improve the speed of inverse solution, we have analyzedvarious existing methods on fast calculations of optical signals which include WhiteMonte Carlo、Forced Detection Monte Carlo、Perturbation Monte Carlo. On the basisof these analyses, we have proposed and developed a new perturbation algorithmwhich can be applied for inverse solution of the three optical parameters definedwithin the framework of radiative transfer theory though rapid iteration. Combiningwith the GPU parallel computing, the new inverse method has been named as the iMCwhich can provide300-fold or higher speed-up for calculation of light signals andinverse determination of optical parameters in comparison to the existing methods.In the end of this thesis, we summarize the results of our algorithm research andcompare the speed–up between the convention Monte Carlo algorithms and iMC fastalgorithm. Experimental data on determination of the optical parameters frommeasured light signals from intralipid samples are included for validation of the iMCmethod. We further discuss the directions of future research.