Motion compensation for MRI-guided radiotherapy

Abstract

Radiotherapy aims to deliver a lethal radiation dose to cancer cells immersed in the body using a high energetic photon beam. Due to physiologic motion of the human anatomy (e.g. caused by filling of internal organs or breathing), the target volume is under permanent motion during irradiation, diluting the applied dose into regions around the target volume. Traditionally, the target volume is expanded about margins encompassing the geometric uncertainty in order to retain the target dose. As a result, however, unwanted dose to nearby organs at risk is increased significantly, which effectively limits the use of clinically promising treatment techniques such as hypo-fractionated therapies to treat cancer more effectively and economically. This thesis develops motion compensation techniques for radiotherapy using MR-imaging integrated into a novel treatment device, the MR-linac, which allows to resolve the current anatomy state during ongoing radiation with diagnostic MR-image quality. As a result the geometric uncertainty of the target positions is reduced which enables new, more effective radiotherapy paradigms. In the first chapters of this thesis, integrated MRI is employed to automatically register volumetric changes in the human anatomy. It is described how significant acceleration of this registration process can be achieved by spatially undersampling volumetric MRI-acquisitions. The calculations showed, that the MR-imaging time speed could theoretically be increased by 300% with only minor losses in registration quality. The on-line volumetric imaging feature was subsequently used to observe the anatomic motion during simulated delivery of a novel treatment scheme for the therapy of renal cell carcinoma. Significant motion could be revealed, eventually causing target dose variations of more than 10% and significant migration of the target dose into the adjacent organs at risk. Used on-line, new intra-fraction options, such as replanning or emergency beam stop could be implemented using the constantly refreshed calculation of dose delivered to the patient. The beam shaping unit (multileaf collimator, MLC) integrated in the MR-linac features real-time control of the geometry of the treatment beam. Potentially, optimal target conformity can be achieved by steering the beam in synchrony with the moving target. In order to assess the expected performance of such target tracking on an MR-linac, the Elekta Agility 160 MLC, was assessed using an experimental imaging pipeline. Low mechanic latencies of under 20ms could be shown, which enables real-time MLC tracking with minimal geometric errors. In order to compensate for residual tracking errors caused by latencies of the mechanical part and software processing, a novel tracking margin generator was designed, which aimed to retain target dose coverage. In experiments, the margin generator could show significant reductions of underdosages caused by MLC tracking errors. Considering the remarkable scalability of the integrated imaging and beam shaping options offered by the MR-linac, a multitude of novel treatment techniques become available. As described in this thesis, rapid target tracking, dose reconstruction and combinations thereof can be implemented using theon-board MR-imaging. In order to apply the described techniques to patients, stringent quality assessment and autonomous feedback processes have to be implemented for the clinical practice.