3D reconstruction from truncated rotational angiograms using linear prediction

For obtaining high-resolution reconstruction of the cerebral r vasculature, cone-beam projections in 3D computed rotational angiography (CRA) are acquired over a circular field of view (FOV) of 28 cm, resulting in a truncation of the data. This results in erroneous values of reconstruction within the region of interest that worsens laterally towards the periphery. In this paper, an application of linear prediction is explored for alleviating the effects of truncation in CRA, and its impact on image registration and also reprojection, an important tool in 3D visualization and image enhancement algorithms in CRA. New observations on the effects of taper in the extrapolated segment on filtered projections, and their implications on 3D reconstruction in CRA lead to windowed extrapolation. Results of the new algorithms on a mathematical phantom and real data are promising.