Two-stage deep learning framework for the restoration of incomplete-ring PET images.

Traditional Positron Emission Tomography systems rely on complete detector rings for full angular coverage and reliable data collection.

However, incomplete-ring Positron Emission Tomography scanners have emerged due to hardware failures, cost constraints, or specific clinical needs.

We present a two-stage deep-learning framework that, without incorporating any time-of-flight (TOF) information, restores high-quality images from data with about 50% missing coincidences-double the loss levels previously addressed by CNN-based methods.

This module starts the reverse diffusion process from the U-Net coarse prediction rather than pure Gaussian noise.