Kernel-based Reconstruction of Cardiac PET Images Using MR Information

Abstract

Positron emission tomography (PET) is used to observe processes within the human body using radioactive tracers. Quality of PET images is compromised by statistical noise, especially in the heart where cardiac and respiratory motion occur. Image prior information is generally useful for improving PET image quality. Sources of prior anatomic information include computed tomography (CT) or magnetic resonance imaging (MRI). In this work, we used MR information in the kernel framework to help reconstruct cardiac PET images and compared it with the kernel reconstruction from PET data only. The kernel-based reconstruction method [1], incorporates prior information in the reconstruction algorithm with the use of kernels. Our results show kernel-based image reconstruction using MR prior anatomic information gives numerically equivalent results to the original kernel method that uses composite frames to reconstruct dynamic PET images.