Accelerated computation of digitally reconstructed radiographs

We describe a method to accelerate the computation of digitally reconstructed radiographs (DRRs) from CT data. Our experiments have shown up to a factor of 2.9 speedup in DRR rendering times. With this method, an operator specifies classes of tissues that are to be ignored in the DRR computation, effectively creating empty voxels in the CT volume. The method segments the volume into empty and non-empty regions, and computes the DRR using just the non-empty regions, using the hardware capabilities of readily available commodity graphics cards. The method is useful for problems such as intensity-based 2D-to-3D registration, where it is acceptable to ignore the soft tissues. (c) 2005 Published by Elsevier B.V.