Optimization-based cone-beam CT reconstruction from a virtual isocenter treatment scan

Abstract

Linac-mounted, cone-beam computed tomography (CBCT) imaging systems help with patient setup and treatment volume verification in image guided radiation therapy (IGRT). However, analytic-based reconstruction algorithms, such as FDK, make assumptions about the scanning trajectory that restrict the image isocenter to the mechanical isocenter of the treatment gantry. This limitation prevents tomographic imaging during virtual isocenter treatments. This form of treatment utilizes of a treatment isocenter that is not at the mechanical isocenter of the linac. To implement such a treatment, the patient must be moved as the gantry rotates to ensure the correct distance between the source and virtual isocenter is maintained. Though a trajectory such as this violates the circular trajectory needed for FDK, new optimization-based reconstruction algorithms impose no assumptions of the acquisition trajectory. As such, these algorithms have to potential to reconstruct a volumetric image from the kV projections acquired during a virtual isocenter treatment. We demonstrated this by acquiring a CBCT scan of such a virtual isocenter trajectory in Varian's TrueBeam Developer Mode and subsequently reconstructing it.