PET Image Reconstruction Based on Graph Learning

Positron emission tomography (PET) is a powerful functional imaging modality, but it has low resolution at low dose tracers. In order to improve the quality of reconstructed images, this paper proposes a novel PET image reconstruction method based on graph learning. Firstly, the graph Laplacian matrix is obtained from the low-quality PET images through graph learning. Then, the PET image is iterative reconstructed using the weight matrix. Our method utilizes graph learning to better leverage the structural information of PET images. The results from simulations and in-vivo data experiments demonstrate that the proposed method achieves better performance over traditional methods and KEM method.