Towards online MRI-guided radiotherapy

First, we present two offline position verification methods which can be used in radiotherapy for detecting the position of the bony anatomy of a patient automatically with portal imaging, even if every single portal image of each segment of an (IMRT) treatment beam contains insufficient matching information. Additional position verification fields will no longer be necessary, which speeds up the treatment and reduces the total dose to the patient. Second, a tool is described which enhances the way tumors can be delineated by using multiple imaging modalities. This tool is especially useful when multiple MRI sequences are available as well as the standard planning CT. We also developed a marker which is visible on MRI and on EPID. A separate CT for detecting the markers is no longer needed. The gold marker with steel core can be detected on various MRI sequences, reducing the overall systematic radiotherapy treatment error. The MRI linear accelerator (MRL) facilitates continuous patient anatomy updates regarding translations, rotations and deformations of targets and OAR during a course of radiotherapy. Accounting for this information demands high speed, online IMRT re-optimization. Therefore, we developed a fast IMRT optimization system which combines a GPU based Monte-Carlo dose calculation engine for online beamlet generation (GPUMCD) and a fast inverse dose optimization algorithm (FIDO). We show that for the presented cases the beamlets generation and optimization routines are fast enough for online IMRT planning. Furthermore, there is no influence of the magnetic field on plan quality and complexity, and equal optimization constraints at 0T and 1.5T lead to almost identical dose distributions. One of the most significant effects of the transverse magnetic field on the dose distribution occur around air cavities: the electron return effect (ERE). We investigated the effects of non-stationary spherical air cavities on IMRT dose delivery in 0.35T and 1.5T transverse magnetic fields by using Monte Carlo simulations. Our observations show the intrinsic ERE compensation by equidistant and opposing beam configurations for moving spherical air cavities within the target area. IMRT gives some additional compensation, but only in the case of correct positioning of the air cavity according to the IMRT compensation. For air cavities appearing or disappearing during a fraction this correct positioning is absent and gating or plan re-optimization should be used. Finally, we introduce an online 'virtual couch shift' (VCS): we translate and/or rotate the pre-treatment dose distribution to compensate for the changes in patient anatomy and generate a new plan which delivers the transformed dose distribution automatically. The VCS is the first step towards compensating all anatomical changes (translation, rotations, and deformations) by online re-optimization of the IMRT dose distribution.