A filtered backprojection algorithm for axial head motion correction in fan-beam SPECT.

In this study we present an approximate, but practical, three-dimensional filtered backprojection (FBP) reconstruction algorithm in fan-beam SPECT to correct for axial motion (both translation and rotation). A one-dimensional filter kernel was applied to the projections. It is assumed that the object is rigid and that its axial motion can be characterized by three components: one-dimensional translation and yaw and pitch rotations. It is further assumed that the motions that have occurred during the SPECT acquisition have been determined separately. The determined angular-view-dependent translation/rotation parameters were incorporated into the proposed FBP algorithm to correct for multiple axial head motions. The proposed axial head motion correction algorithm was evaluated using simulated three-dimensional Hoffman brain phantom data. Projections both with axial translation and with axial rotation, and with their combinations were generated. Images of a Hoffman brain phantom reconstructed using the proposed FBP algorithm and the conventional FBP algorithm were compared. Artefacts were observed in images without motion correction, but the artefacts were greatly reduced using the proposed reconstruction algorithm.