Comparison of two- and three-dimensional reconstruction methods in optical tomography.

We present a three-dimensional (3D) image reconstruction scheme for optical near-infrared imaging of highly scattering material. The algorithm reconstructs the spatial distribution of the optical parameters of a volume Omega from transillumination measurements on the boundary of Omega. We test the performance of the method for a cylindrical object with embedded absorbing perturbation for a number of different source and detector arrangements. Furthermore, we investigate the effect of a mismatched reconstruction, using a two-dimensional (2D) reconstruction model to image a single plane of the object from 3D tomographic data obtained in a single plane. The motivation for the application of 2D models is their advantage in speed and memory requirements. We found that the difference in the measurement data between 2D and 3D models depends greatly on the measurement type used, giving a much better agreement for mean time-of-flight data than for dc intensity data. Image artifacts that are due to data model mismatches can therefore be significantly reduced by use of mean time data.