Neural Network Based Motion Correction for Freely Moving Animal PET Imaging

During PET scanning, motion of animals can result in severe motion artifacts in the image. In this paper, we proposed a novel motion correction method for freely moving animals using PET histo-images and a deblurring neural network. The proposed method was tested on a moving resolution phantom and a cylindrical phantom, both of which obtained good results. The initial results show the promise of performing motion correction on the histo-images and using a neural network to generate high-quality images.