Reconstruction of Standard-Dose Pet From Low-Dose Pet Via Dual-Frequency Supervision and Global Aggregation Module

Abstract

Positron emission tomography (PET) imaging is a widely used technology in clinics. To meet the clinical diagnostic requirement, standard-dose tracers with radioactivity need to be injected into the body when acquiring PET images. To reduce imaging radiation hazard while maintain PET image quality, in this paper, we propose a novel and effective approach to reconstruct standard-dose PET (SPET) images from low-dose PET (LPET) images. Specifically, we first design a two-branch network to preserve richer high-frequency details by reconstructing low-frequency component and high-frequency component separately. Then, we design a global integration module to integrate global information for achieving better quality in those difficult-to-reconstruct regions. Validation on a real clinical dataset suggests that our approach outperforms previous methods both qualitatively and quantitatively.