Deeply-Supervised Multi-Dose Prior Learning For Low-Dose Pet Imaging

Abstract

Positron emission tomography (PET) is an advanced imaging modality for tumor staging and therapy response. However, PET radiation exposure has raised public concerns and it is in need to develop low-dose PET imaging techniques. This paper proposes to explore prior information inherited in different levels of low-dose PET images with deep learning and then utilize them to estimate high-quality PET images from the image with the lowest dose. The proposed method is evaluated on the in vivo dataset with encouraging performance.