Iterative cone-beam reconstruction of Continuous Bed Motion transmission scans in PET

Continuous Bed Motion (CBM) acquisition has a number of advantages over the traditional Step-and-Shoot (SS) mode, including: Uniform axial signal-to-noise ratio, improved resolution due to subsampling in the axial direction, and reduction of noise due to detector normalization. In this study, a pair of Co-57 point sources are used to generate transmission and blank data on a Siemens Inveon small animal PET scanner. When exposed and rotated, the collimated sources project cones onto the PET field-of-view (FOV). CBM data were acquired in listmode format by translating the bed and rotating the sources at constant speeds. Unlike emission CBM, where parallel LORs are accumulated to form an emission sinogram, each transmission LOR is at a different angle from source to detector. Thus, the data is histogrammed into a series of projections sorted by projection angle [ø], azimuthal detector [r], and axial location[n]. An iterative cone-beam Order Subset Convex method was implemented to reconstruct this projection data. System geometry models and corrections are applied during the reconstruction to improve image quality and for resolution recovery. The goal of this study is to produce a high quality attenuation correction matrix for the emission data, and transmission images suitable for registration to the emission images.