Preparation and optimization of the Walk-Through PET for clinical applications.

Recent advancements in PET imaging have led to the development of Total-Body PET systems, which offer significantly higher sensitivity, enabling reduced radiotracer doses and shorter scan times compared to conventional PET systems. However, the high costs of current Total-Body PET scanners limit their integration into clinical practice. To address this, the MEDISIP research group at Ghent University is developing a Walk-Through PET system. This innovative design features two flat detector panels that scan patients in a standing position and uses cost-effective monolithic detectors. While this design reduces costs and enables high patient throughput, it also introduces several challenges. This project focuses on overcoming some of these challenges by developing advanced spatial and timing calibration methods for monolithic detectors, ensuring optimal performance. Additionally, CT-less attenuation and scatter correction techniques are designed to lower radiation doses and scan times by eliminating the need for CT scans. Furthermore, artifacts due to limited angles in the flat-panel geometry are corrected to improve image accuracy. These advancements provide a basis for the evaluation of the system's performance (being spatial and temporal resolution), and its potential use in population screening applications. Beyond the Walk-Through PET system, the results of this work are also applicable to other PET geometries.

Keywords:PET image reconstruction, PET imaging, Deep learning

Disciplines:Biomedical instrumentation, Biomedical image processing