An Optimization Algorithm to Inverse Problem in 2-D Optical Computed Tomography by BP Neural Network

An optimization approach is introduced to inverse model of optical CT, which can reconstruct the optical properties namely scattering coefficients of thick tissue such as brain and breast tissues. A modeling and simulation tool named FEMLAB and based on finite element method has been tested wherein the forward models based on the diffusion equation. Solve the inverse model as an optimization approach, including the optimization methods of optical properties. An improved BP neural network which based on Levenberg-Marquardt optimization algorithm was employed to solve this inverse problem. The simulation results show that the algorithm can reconstruct the optical image quickly and accurately.