high performance computing on graphics processing units: hgpu.org

There is no one size fits all solution when it comes to CT reconstruction. Many different CT reconstruction algorithms and implementations have been devised in an attempt to solve the problem of producing an image under a specific set of constraints. One optimal CT reconstruction implementation can look very different from another optimal implementation; depending on the data, quality, and time constraints. In this paper, we present a framework that is able to dynamically create and compile new implementations that optimize the multiple objectives contained in CT reconstruction. We then show the results of this framework when applied to a GPU accelerated version of the FDK back-projection algorithm.