A normalization scheme for LOR-based motion correction in PET

Line of response (LOR) rebinning has been demonstrated to be an effective motion correction method for positron emission tomography (PET) imaging. In LOR rebinning, normalization of each motion-corrected event is needed before it is placed into its new sinogram bin. In general, due to data compression strategies the sinogram bincorresponding to the transformed LOR will receive contributionsfrom multiple LORs. This paper demonstrates that normalization of the corrected event must account for the relative change in its contribution to the corresponding sinogram bins before and after transformation. Failure to account for this factor may cause slice-to-slice count variations of transverse slices and visible horizontal stripe artifacts in the coronal and sagittal slices of the reconstructed images.