Quasi-exact region-of-interest reconstruction from helical cone-beam data

This paper presents a new algorithm for reconstruction from truncated cone-beam projections acquired with a helical motion of the X-ray source. Both approximate and exact algorithms have been proposed for this problem. Approximate methods are fast and require few interpolations; they are appropriate for the first generations of multi-row medical CT scanners for which the ratio between the pitch of the helix and the spatial resolution is sufficiently small. Exact algorithms are important for industrial scanners and for future generations of medical scanners as they can be applied with arbitrarily large values of the pitch at any resolution. We describe the zero-boundary method and demonstrate its exactness. Results are then provided demonstrating accurate reconstruction from simulated data for large pitches.