br Fig Contouring for MRgRT CTV consisting of prostate
Fig. 1. Contouring for MRgRT: CTV consisting of prostate and MG-132 of vesicles (green contour), PTV (CTV + 3 mm; red contour) visualized in an axial, sagittal and coronal plane. The urethral contour (cyan contour) and urethral PRV (urethra + 2 mm) can be best seen in the sagittal plane. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)
day. The deformable registration algorithm implemented on the MRIdian and employed for online adaptive minimizes a cost function that measures the similarity between the images. It also uses a reg-ularization term to obtain smoother deformation fields and prevents sharp discontinuities. The optimization method relies on a simple gra-dient descend performing the registration firstly on a down sampled version of the image serving the results as initial guesses for each upper level.
The electron density map generated from the CT for dose calculation underwent the same deformation applied to the OAR contours. The newly generated electron density map for that particular fraction was briefly checked by the radiation technologist and physicist for the presence of missing tissue densities and mismatch for air pockets in rectum. In case of mismatch, structure densities were overridden and corrected online before dose calculation and plan adaptation.
2.5. Treatment planning
Treatment planning and delivery was performed with static field intensity modulated radiotherapy (IMRT). A relatively high number of beams were used (15) which provided enough degrees of freedom and flexibility to re-adapt the plan and account for anatomical changes. Typically around 45 segments were generated which in combination with the different beam angle incidence produced treatment plans achieving the modulation needed for selective urethral sparing. The MRIdian Linac version uses a double focused, double-stacked multileaf collimation (MLC) in combination with 6MV FFF photons, allowing for highly conformal dose distributions and steep dose gradients at the borders with adjacent OARs. The obtained dosimetry for treatment planning in MRgRT is comparable to VMAT techniques . Our treatment planning approach for MRgRT has been developed with daily plan adaptation in mind , and will be presented below.
Dose calculation was performed with a Monte-Carlo algorithm im-plemented in the MRIdian system based on VMC and EGSnrc codes [25,26]. The algorithm can complete an IMRT plan calculation subject to a magnetic field in 2 min. For clinical plans, a statistical uncertainty of 1% was used with a dose grid resolution of 0.2 cm × 0.2 cm × 0.2 cm.
2.6. MR-guided online adaptive workflow
A summary of the treatment workflow for MRgRT with daily plan adaptation implemented at our center is visualized in Fig. 2. After MR acquisition and patient registration, CTV and OAR contours always needed to be online adjusted by the attending radiation oncologist to
correct for variations in the position of the upper part of the prostate and base of the seminal vesicles. For the daily plan adaptation whilst the patient is in treatment position, only OARs in the first 2 cm outside the PTV were corrected to allow for a fast online workflow. At each fraction a new electron density map for dose calculation was generated after applying deformable image registration. Subsequently, two plans were generated: the baseline plan re-calculated on the anatomy-of-the-day (predicted plan) and the re-optimized plan. The re-optimized plan was generated by OAR partitioning within the first 2 cm from the PTV surface and updating all necessary structures for treat-ment planning by means of an automated script. Both plans were re-viewed by the radiation oncologist and physicist whether they met the preset plan objectives. An example of the potential of plan adaptation in one patient undergoing an MRgRT treatment can be seen in Fig. 3, where the baseline plan, predicted plan and re-optimized plan for a particular fraction can be observed. Initial anatomy at baseline showed some distance between the prostate and rectum allowing for adequate coverage of CTV and sparing of rectum. However, rectum distension brought forth a pitch and deformation on the prostate at one particular fraction, resulting in suboptimal CTV coverage and an increased dose to the rectum. Online plan re-optimization following proposed strategy resulted in adequate rectum sparing and recovery of CTV coverage.