Accurate vascular reconstruction with MAP-EM method from few projections

The aim of this study is to reconstruct three-dimensional images of intracerebral vessels from few two-dimensional projection views. The reconstruction of intracerebral vessels from few projections by means of Maximum A Posteriori (MAP) - Expectation Maximization (EM) gives very slow convergence. The proposed algorithm reconstructs images in two stages. In the first stage a rough estimated image is reconstructed by the MAP-EM method with a Gibbs prior. Then the last image in the first iteration sequence (ex. after 15 iterations) is used for the second penalty function in the second reconstruction sequence. This penalty function can accelerate the convergence of image reconstruction for the MAP-EM method. In this penalty function we considered the connectivity of vessels between adjacent slices. The idea is based on the assumption that a blood vessel is a line consisting of at least three voxels that have almost the same voxel values. We evaluated the proposed method by using a simulation phantom and an experiment phantom made with a curved wire. The number of projection views used was five (RAO 60, 30, LAO 60, 30 and 0 deg.). The quality of the image was evaluated by the profiles of reconstructed three-dimensional images. These profiles showed the improvement in voxel values at low voxel-value areas with the proposed method. The results showed that the proposed method was useful for the reconstruction of vessels with few projections.