Optical tomographic reconstruction in a complex head model using a priori region boundary information.

In this paper we investigate the application of anatomical prior information to image reconstruction in optical tomography. We propose a two-stage reconstruction scheme. The first stage is a reconstruction into a low-dimensional region basis, obtained by segmentation of an image obtained by an independent imaging modality, into areas of distinct tissue types. The reconstruction into this basis recovers global averages of the optical tissue parameters of each region. The recovered distribution of region values provides the starting point for the second stage of the reconstruction into the spatially resolved final image basis. This second step recovers localized perturbations within the regions. The benefit of this method is the improved stability and faster convergence of the imaging process compared with a direct reconstruction into a spatially resolved basis. This is particularly important for the simultaneous reconstruction of absorption and scattering images, where ambiguities between the two parameters and the resulting problems of crosstalk require a good initial parameter distribution to ensure convergence of the reconstruction. We use a segmented brain model obtained from a magnetic resonance image as a test case to compare the performance of the two-stage reconstruction and the direct reconstruction from a flat prior, and show that the former achieves superior results in the recovery of localized absorption and scattering hot spots embedded in the background tissue.