Imaging plays an important role in the delivery of external beam radiation therapy. It is used to confirm the setup of the patient and to ensure accurate targeting and delivery of the therapeutic radiation dose. Most modern linear accelerators come equipped with a flat panel detector opposite the MV source as well as an independent kV imaging system, typically mounted perpendicular to the MV beam. kV imaging provides superior soft tissue contrast and is typically lower dose to the patient than MV imaging, however it can suffer from artifacts caused by metallic objects such as implants and immobilization devices. In addition to being less artifact prone, MV imaging also provides a direct measure of the attenuation for the MV beam which is useful for computing the therapeutic dose distributions. Furthermore either system requires a large angular coverage, which is slow for large linear accelerators. We present a method for reconstructing tomographic images from data acquired at multiple x-ray beam energies using a statistical model of inherent physical properties of the imaged object. This approach can produce image quality superior to traditional techniques in the case of limited measurement data (angular sampling range, sampling density or truncated projections) and/or conditions in which the lower energy image would typically suffer from corrupting artifacts such as the presence of metals in the object. Both simulation and real data results are shown.

Date Published: 18 March 2015
PDF: 8 pages
Proc. SPIE 9412, Medical Imaging 2015: Physics of Medical Imaging, 94125M (18 March 2015); doi: 10.1117/12.2082068

Published in SPIE Proceedings Vol. 9412:
Medical Imaging 2015: Physics of Medical Imaging
Christoph Hoeschen; Despina Kontos, Editor(s)