Generalized inter-frame and intra-frame motion correction in PET imaging - a simulation study

High resolution PET imaging is severely hampered by patient motion. Frame-acquired PET images suffer from inter-frame and intra-frame motion artifacts degrading image quality. The method of Multiple Acquisition Frames (MAF) [1] corrects inter-frame motion artifacts by removing average motion from all the independently reconstructed frames. The drawback is that a high motion threshold neglects considerable intra-frame motion and a low motion results in acquisition of low-statistic frames, thereby, degrading image quality. Increasing the number of frames proportionally increases reconstruction times [2]. Another approach for motion compensation in PET imaging is to employ (MLEM)-type deconvolution based on the Richardson-Lucy algorithm [3], [4]. The reconstructed image is used to estimate the original non-corrupted image using MLEM techniques [5], [6]. The drawback is the occurrence of attenuation artifacts that can arise from emission-transmission mismatches. Our approach picks up the striking features of these two themes to correct for inter-frame and intra-frame motion in dynamic PET acquisition. In the following sections, we characterize 3D rigid body motion and use the concept of averaging transformations inspired from Riemannian Geometry and Topology, present the proposed Generalized Motion Correction Algorithm followed by simulation results, discussion and conclusions.