Penalized PET/CT Reconstruction Algorithms With Automatic Realignment for Anatomical Priors

Two algorithms for solving misalignment issues in penalized PET/CT reconstruction using anatomical priors are proposed. Both approaches based on a recently published joint motion estimation and image method. The first approach deforms the to align it with functional one while second both images them measured data. Our current implementation alternates between alignment reconstruction. We have chosen parallel level sets (PLSs) as representative penalty, incorporating spatially variant penalty strength. performance was evaluated simulated nontime-of-flight data generated an XCAT phantom thorax region. used attenuation map prior. results demonstrated that methods can estimate deform accordingly. However, of depends highly workflow alternating process. shows faster convergence rate correct is less sensitive workflow. presence information improves but slow down approach. show improved noise decreases.