Adaptation of a Fast 3D PET Reconstruction Algorithm to an Inverse-Geometry CT System

Inverse-geometry CT (IGCT) is a new type of volumetric CT geometry. Volumetric coverage and high isotropic resolution demand a fast 3D reconstruction algorithm. The purpose of this work is to evaluate a very fast algorithm based on Defrise\´s Fourier Rebinning (FORE) algorithm for PET. IGCT employs a large array of focal spots opposite a detector array that shares the same axial extent as the source but is much smaller in the transverse direction. A single rotation is sufficient to acquire a 10 cm thick volume with isotropic resolution. A 3D reconstruction algorithm was developed based on FORE. For comparison, a Feldkamp-like algorithm was developed that generates an average of Feldkamp reconstructions from each source row. Projection data for a numerical torso phantom and a numerical "Defrise" phantom were simulated under the IGCT geometry. To evaluate the FORE algorithm on a real data set, a prototype IGCT system was assembled involving a 100times100 scanned-anode source array and a 48times48 photon-counting detector array mounted on a C-arm system (NovaRay, Inc., Palo Alto, CA). An anthropomorphic phantom containing a real porcine heart was scanned in a step-and-shoot mode with the C-arm in a horizontal orientation. The FORE algorithm averaged 30 sec/slice while Feldkamp averaged 30 min/slice. Like Feldkamp, FORE suffers from cone-beam blurring in z, however it does not suffer from cone-beam streaking artifacts. FORE proved to be both fast and reasonably accurate.