A component-based parallel PET image reconstruction visual system

Abstract

By quantifying the emission counts of positron-electron annihilation, PET provides a unique way of studying physiological and metabolic functions of the human body. The PET images can be reconstructed by iterative methods, e.g., the expectation maximization (EM) algorithm. The computation-bound problem is expected to be exacerbated due to the ever-increasing demand for faster image reconstruction and higher resolution PET images. As Java technology matures, a number of parallel/distributed applications are implemented in Java because of its rich sets of packages. We propose the design and implementation of a component-based parallel software system to facilitate the PET image reconstruction process on a clustered environment. Performance issues are addressed in the design of the system, which are essential to the success of any parallel computing environment. The system introduces the Java Bean component model and class libraries which are written in Java and that run on any machine that implements the Java Virtual Machine. It is user friendly and its operation requires little software/hardware knowledge from the user.