AbstractID: 9637 Title: Implementation of an algorithm for fast DRR calculation and
estimation of a beam source from a radiograph
We present an implementation of an algorithm for fast calculation of DRR from a CT
based volume. We also present a theoretical and computational method to utilize our fast
DRR computation algorithm to calculate the beam source orientation from a given
radiograph. The fast DRR algorithm involves decomposing the 3D volume data into a set
of basis functionals that are invariant in one dimension. To generate a DRR when the
beam source rotates around an axis, the appropriate functional set is cylindrical
harmonics. Each cylindrical harmonic functional is projected into the projection plane
from the source. The projection of the volume into the projection plane involves an
appropriate summation of all projections. Since the cylindrical harmonic functionals are
invariant of beam source axial rotation, new projection of the volume, as a result of the
beam rotation, can be recomputed by an appropriate phase shifting and summing the
basis functionals projection sets. Using this method in recalculating DRR of a volume,
after the initial decomposition, offers a significant advantage in computational speed with
respect to other methods, such as ray casting or shear-warp factorization algorithms.
However, the main advantage of this algorithm is for source localization and patient
orientation and motion compensation. One can apply an inverse process (back-projection)
to evaluate the orientation of the beam source from a projected image. This information is
used in computing the patient re-positioning for treatment delivery, using a neural
network estimator algorithm.