Real-time 2D/3D image registration in radiation oncology

Based on earlier research from our groups in the determination of 3D patient position from projection images, we propose to develop a real-time 2D/3D registration method for online tumour motion detection in image-guided radiation therapy (IGRT) at an update rate of 5 Hz. The result of our research effort will be a software suite that tracks the tumour target volume on a real-time x-ray during irradiation in 3D and intercepts the treatment beam if motion beyond a given safety margin is detected. IGRT using on-line x-ray imaging is the latest development in radiation oncology and currently among the hottest topics in the field of medical physics. On-board imaging units allow for acquisition of planar images as well as cone-beam CT volume data. Patient position relative to the treatment plan can be determined on a daily basis, and tumour motion due to weight loss, movement of internal organs, or inaccurate patient positioning can be taken into account. However, short - term movement like breathing leads to considerable problems in precision irradiation of the targets in the lung or the abdomen and cannot be compensated by this method. Several researchers have therefore suggested the use of implanted gold markers or position tracking electronics in the tissue surrounding the tumour. In our proposed project, we plan to achieve the same goal by speeding up 2D/3D registration to real time extent. Then, any implantation of markers or sensors is unnecessary. This will be accomplished by optimizing some of our already proven algorithmic methods for ultra-fast generation of digitally reconstructed radiographs and optimized merit functions such as stochastic rank correlation for standard high-performance computing hardware. State-of the art IGRT equipment as well as clinical expertise in radiation oncology is available for the project. Extensive pre-clinical and clinical studies as well as dosimetric examinations in close collaboration between our departments form an integral part of the research effort. Furthermore, we plan to integrate the University of Applied Sciences Technikum Wien into the project by providing several master theses in the field of biomedical informatics. The project is intended for a three year duration. Within that period, we will deliver a well-tested implementation of our algorithms. A registration update rate of 5 Hz is intended for efficient tumour motion control. Dissemination of results is planned by means of publications in peer-reviewed journals, patents, and collaborations with major vendors of IGRT equipment.

The purpose of this project is to derive an organ motion pattern for a tumor during irradiation by means of x-rays taken during the treatment. For instance, breathing motion causes considerable uncertainty in the localization of lung tumors. Therefore it is necessary to irradiate a larger area, which endangers nearby organs and healthy tissue. By simulating x- ray images from three-dimensional computed tomography data, one can follow the motion of the tumor if the software system for simulation and comparison of the real x-ray data is fast enough. The development of such a software system was the purpose of this project. By algorithmic research and utilization of novel computer hardware, we were able to achieve this goal. Our results show that it is possible to follow the so-called clinical target volume the area to be irradiated at a pace of 0.2 seconds with an accuracy of approximately 2 mm. Freely available research article: Gendrin C, Furtado H, Weber C, Bloch C, Figl M, Pawiro SA, Bergmann H, Stock M, Fichtinger G, Georg D, Birkfellner W. Monitoring tumor motion by real time 2D/3D registration during radiotherapy. Radiother Oncol 102(2):274-80, 2012 http://www.ncbi.nlm.nih.gov/pmc/articles/PMC3276833/