Comparison of imaging geometries for diffuse optical tomography of tissue.

Images produced in six different geometries with diffuse optical tomography simulations of tissue have been compared using a finite element-based algorithm with iterative refinement provided by the Newton-Raphson approach. The source-detector arrangements studied include (i) fan-beam tomography, (ii) full reflectance and transmittance tomography, as well as (iii) sub-surface imaging, where each of these three were examined in a circular and a flat slab geometry. The algorithm can provide quantitatively accurate results for all of the tomographic geometries investigated under certain circumstances. For example, quantitatively accurate results occur with sub-surface imaging only when the object to be imaged is fully contained within the diffuse projections. In general the diffuse projections must sample all regions around the target to be characterized in order for the algorithm to recover quantitatively accurate results. Not only is it important to sample the whole space, but maximal angular sampling is required for optimal image reconstruction. Geometries which do not maximize the possible sampling angles cause more noise artifact in the reconstructed images. Preliminary simulations using a mesh of the human brain confirm that optimal images are produced from circularly symmetric source-detector distributions, but that quantitatively accurate images can be reconstructed even with a sub-surface imaging, although spatial resolution is modest.