This study investigated anatomical changes and prostate motion that occurred during a time frame corresponding to reported daily adaptive replanning time of 20 – 40 min in MR-linac systems [20,21,22,23], as well as the dosimetric consequences for one fraction, following these changes. This is to our knowledge the first study to relate the dosimetric effects of anatomical change and prostate displacements in localized prostate cancer patients, to the prolonged adaptive process time in an MR-linac workflow.
From our results it is evident that there is a variety in bladder volume increase, which indicate a different rate of bladder filling for different patients. The largest posterior prostate shifts were caused by large bladder volume increase, often in combination with muscle relaxation. Regarding the rectum, there was both an increase and decrease in volume. The largest increase in rectal volume was due to gas cavities, which either increased in size from MR1 to MR2 or appeared in MR2. These gas cavities caused the largest prostate shifts in the anterior direction. Our observations show that bladder and rectum volume changes are probable and individual and are therefore important processes to monitor during the adaptive replanning in the MR-linac. A possible approach for controlling the bladder volume could be to use a bladder filling protocol. The bladder volume would then initially be approximately the same at the start of every fraction. However, during the adaptive replanning process, there will still be a volume increase.
Prostate motion observed in this study, agrees well with results from previous studies [14, 15, 17]. The probability of prostate motion increases over time [13,14,15,16,17,18]. For example, Nederveen et al.  identified prostate displacements larger than 3 mm within only 3 min and Langen et al.  found that with time, the prostate tends to drift. The average prostate tracking time studied was 10 min and during this time prostate drifts larger than 10 mm occurred. Langen et al. further reported that with this motion pattern the prostate does not revert to its original position. Ballhausen et al.  also found a similar drifting motion of the prostate during 15 min and concluded that it is beneficial to reduce the treatment session duration. Cramer et al.  suggested a repositioning of the patient if the treatment session exceeded 4–6 min. These studies all observed prostate motion during a period frame substantially less than the time spent on adaptive replanning in an MR-linac.
Nejad-Davarani et al.  studied the dosimetric effect of anatomical changes in prostate cancer patients for an MRI-only workflow. MR images of volunteers with both full and empty bladder were acquired, with 45 min apart. A reference plan of the full bladder case was recalculated onto the empty bladder images. PTV D95% was on average reduced by 11.5% when comparing the empty bladder treatment plans to the full bladder plans. The decrease in PTV D95% was caused by prostate displacements in the anterior direction. Although Nejad-Davarani et al. used a bladder filling protocol with the full bladder condition as reference, the study indicates that differences in bladder filling can cause prostate shifts large enough to significantly decrease the dose coverage to the target, which is consistent with our findings. Nejad-Davarani et al. also found a larger reduction of dose to the target than we did in this study, which may be due to their longer separation between image acquisition.
Accordingly, with previous results and new data presented in this study, one can conclude that there is a considerable risk of prostate displacement during the adaptive process in the MR-linac. Before beam on, it will be crucial to verify the patient and prostate position thoroughly. This becomes especially important if imaging during treatment delivery is not available. According to our results, when using a PTV margin of 3 mm, there is a 20.0% risk of having to repeat adaptive replanning because of prostate motion larger than 3 mm. This indicates that a PTV margin of 3 mm might be too small to use while the adaptive replanning takes 20–40 min in the MR-linac workflow, as it currently does.
Furthermore, the treatment delivery time when using an MR-linac system for ultra-hypofractionation of the prostate can take up to 15 min . This means that from that the irradiation begins the prostate has probably started to drift and this drifting motion will continue throughout the extended treatment delivery. Even if imaging throughout treatment delivery is used and translational couch shifts are applied based on prostate position, there could be negative dosimetric impacts on the OARs because of further deformation of these.
There are of course many possible benefits of using an MR-linac instead of a conventional linac for prostate cancer patients. The main advantages are the soft tissue contrast of the MR images, the ability of imaging during irradiation without additional dose and the possibility of adaptive treatment [28, 29]. However, the prolonged time for adaptive replanning and treatment delivery of the MR-linac is a drawback, which increases the risk of a target underdosage unless the anatomical changes are taken into account. On the MR-linac there is the possibility of acquiring images during the daily replanning and/or right before beam on. Adjusting the plan or patient position based on these images would likely reduce underdosage to the target. However, position verification imaging before treatment start must be thorough. As shown in this study, prostate motion can occur in all directions because of various anatomical changes and a two dimensional (2D) image in one or two planes might not be sufficient to detect motion causing target underdosage, especially when using a 3 mm PTV margin.
The dose to the rectum was in this study optimised according to the local dose protocol which is valid for a PTV margin of 7 mm. However, it is not entirely accurate for the treatment plans with 5 and 3 mm PTV margins. If the locally used PTV margin of 7 mm is changed to 5 or 3 mm, it is highly likely that the dose criterion for the rectum also would be changed. The comparison between D1 and D2 using the local DVH criteria was performed in order to clearly show that with a smaller PTV margin the rectum dose could be reduced. The dose to the rectum was nonetheless decreased for D2, regardless of PTV margin. Rectum is one of the most important OARs to consider when planning the treatment for prostate cancer patients. Therefore, a decrease in rectal dose can appear positive, since the side effects might be reduced. However, the observed results in this study demonstrate that this would be at the expense of an impaired dose coverage of the prostate.
All treatment plans in this study were planned using VMAT. VMAT is not yet available on the MR-linac unit, instead intensity modulated radiotherapy (IMRT) is the available delivery technique. Tetar et al.  reported a use of 15 field-IMRT for their prostate cancer patients, which generated a similar dose distribution as for a VMAT treatment plans, which justifies our method using VMAT plans in this evaluation. Since VMAT treatment plans provides a more conform dose distribution and a faster delivery time , VMAT will likely be clinically available for MR-linacs in the future.
A limitation of this study is that it is not performed on an MR-linac. However, the purpose of this study was to quantify anatomical changes that can occur during the duration of adaptive replanning. This was achieved by acquiring MR images on a conventional MR camera and thereafter calculate the dose on sCTs, without the need of an MR-linac system. Anatomical changes and dosimetric effects were evaluated for one fraction only in this study. This approach required the least amount of assumptions of the patient’s motion pattern, since only one pair of images was available for each patient. It is not likely that a patient moves in the exact same pattern with the same amount of deformation in rectum and bladder throughout all seven fractions. One option could have been to add the dosimetric impact from one fraction to the remaining six, with the assumption that during these six fractions, treatment is delivered exactly as planned. This scenario was also considered unlikely. A future extension to this study will be to acquire MR images of volunteers, during multiple days. More images with less time in between could be acquired during the same imaging session. Another possibility could be to use a 4D MR sequence. With more images acquired during multiple imaging sessions it could be possible to assess if a patient’s motion pattern is similar or differs from day to day. It would also be possible to evaluate whether the bladder fills at the same rate and how much the rectal activity can differ between fractions. This would enable investigation of how beneficial a bladder filling protocol or dietary protocol could be for individual patients. In addition, the motion of the prostate can be studied to examine if displacements mainly occur early, late or evenly throughout the treatment session. The distribution between sudden shifts and slow drifting motion could also be evaluated. Such information could be useful when choosing between conventional radiotherapy and treatment on an MR-linac.
The MR-linac systems are constantly evolving and the MR-linac workflow will most likely become faster as development continues. However, until a faster daily adaptive replanning process is possible, this study could help raise awareness about the possible limitations of treating prostate cancer patients with an MR-linac and underline that there are potential risks and disadvantages for this specific category of patients.