Expectation Maximization Reconstruction of PET Image with Non-rigid Motion Compensation

Motion artifact degrades PET imaging which often leads to inaccurate quantification of radioactivity concentration. Current motion correction algorithms generally lack the ability to compensate for non-rigid motions as well as perform poorly for frequent non-cyclical motions. In this paper, we derive from first principles a general list mode reconstruction algorithm that accounts for time-varying system responses. We further model the system response function in a way to incorporate the non-rigid motion. The resultant list mode reconstruction algorithm has the ability to compensate for frequent and non-rigid motions, with all detected events utilized. We also provide a histogram reconstruction algorithm that is tailored for gated PET acquisition with reduced usage of computational resource. This algorithm shows good motion compensation capability with computer simulated data and there is no obvious motion artifact visible on the reconstructed image