The impact of motion on onboard MRI-guided pencil beam scanned proton therapy treatments

Objective. Online magnetic resonance imaging (MRI) guidance could be especially beneficial for pencil beam scanned (PBS) proton therapy of tumours affected by respiratory motion. For the first time to our knowledge, we investigate the dosimetric impact of respiratory motion on MRI-guided proton therapy compared to the scenario without magnetic field. Approach. A previously developed analytical proton dose calculation algorithm accounting for perpendicular magnetic fields was extended to enable 4D dose calculations. For two geometrical phantoms and three liver and two lung patient cases, static treatment plans were optimised with and without magnetic field (0, 0.5 and 1.5 T). Furthermore, plans were optimised using gantry angle corrections (0.5 T +5° and 1.5 T +15°) to reproduce similar beam trajectories compared to the 0 T reference plans. The effect of motion was then considered using 4D dose calculations without any motion mitigation and simulating 8-times volumetric rescanning, with motion for the patient cases provided by 4DCT(MRI) data sets. Each 4D dose calculation was performed for different starting phases and the CTV dose coverage V 95% and homogeneity D 5%–D 95% were analysed. Main results. For the geometrical phantoms with rigid motion perpendicular to the beam and parallel to the magnetic field, a comparable dosimetric effect was observed independent of the magnetic field. Also for the five 4DCT(MRI) cases, the influence of motion was comparable for all magnetic field strengths with and without gantry angle correction. On average, the motion-induced decrease in CTV V 95% from the static plan was 17.0% and 18.9% for 1.5 T and 0.5 T, respectively, and 19.9% without magnetic field. Significance. For the first time, this study investigates the combined impact of magnetic fields and respiratory motion on MR-guided proton therapy. The comparable dosimetric effects irrespective of magnetic field strength indicate that the effects of motion for future MR-guided proton therapy may not be worse than for conventional PBS proton therapy.


Introduction
The use of online magnetic resonance imaging (MRI) has emerged as a groundbreaking approach in the field of radiotherapy-based cancer treatment, enabling the real-time visualization of the tumour and surrounding anatomy during the treatment delivery.MRI stands out as a non-ionizing imaging technique that offers exceptional soft-tissue contrast.For conventional photon radiotherapy, onboard MRI guidance is already in clinical use with two commercial systems (Mutic andDempsey 2014, Raaymakers et al 2017).
Due to the dose deposition in sharp Bragg peaks (Bragg and Kleeman 1905), pencil beam scanned (PBS) proton therapy has the potential to deliver highly conformal dose distributions to the tumour while reducing the dose to nearby organs at risk (OARs) (Lomax et al 1999a) compared to conventional photon radiotherapy.On the other hand, protons are very sensitive to any variations in patient positioning and internal anatomical changes.As a result, accurate real-time image guidance becomes more essential and demanding to exploit the benefits of proton therapy fully.Consequently, online MRI guidance for PBS proton therapy has been an active field of research and first prototype systems for MRI-guided proton therapy are being developed (Oborn et al 2017, Hoffmann et al 2020, Pham et al 2022).
As the particles are charged, however, the effect of the magnetic field on proton beam dose deposition has to be considered during the dose calculation and treatment planning.This has been investigated in various simulation studies using Monte Carlo (MC) simulations (Raaymakers et al 2008, Moteabbed et al 2014, Hartman et al 2015, Oborn et al 2015, Fuchs et al 2017, Kurz et al 2017, Burigo and Oborn 2019, Lühr et al 2019, Santos et al 2019, Burigo and Oborn 2021), while analytical or numerical approaches to account for the deflection of proton beams in a perpendicular magnetic field have also been proposed (Wolf and Bortfeld 2012, Hartman et al 2015, Fuchs et al 2017, Schellhammer and Hoffmann 2017) and integrated into analytical dose calculation algorithms by Padilla-Cabal et al (2018, 2020), Duetschler et al (2023b).Furthermore, results from the operation of experimental systems have been reported (Schellhammer et al 2018a, 2018b, Lühr et al 2019, Padilla-Cabal et al 2019, Gantz et al 2020, 2021, Fuchs et al 2022).
To date, however, only a limited number of treatment planning studies have investigated the clinical benefits associated with onboard MRI guidance for proton therapy.Moreover, these studies, which encompassed pediatric (Acharya et al 2020), liver (Moteabbed et al 2021a), prostate (Moteabbed et al 2021b), and lung cancer patients (Rabe et al 2023), did not consider the effect of the magnetic field on the proton dose distributions.In another study, Kurz et al (2017) used MC-based treatment planning to investigate the robustness against setup errors and anatomical changes for prostate treatment plans optimised with and without a magnetic field.None of these studies did, however, account for the deteriorating effects of motion on the delivered dose distributions, although lung and liver treatments are particularly susceptible to respiratory motion during the treatment delivery.On the other hand, these treatment sites are affected by respiratory motion and could eventually benefit from MRI guidance the most, as the real-time MR images could be used to gate the treatment or for tumour tracking.
Thus, for the future implementation of online MRI-guided proton therapy for moving tumours, the dose calculation should not only consider the impact of the magnetic field on the dose distributions, but also the effect of motion on the dose.Therefore, in this paper, a fast MR-compatible analytical dose calculation, as developed by Duetschler et al (2023b), has been extended to 4D dose calculations and used to perform, to our knowledge, first dosimetric simulations that consider the joint impact of respiratory motion and magnetic fields on PBS proton therapy.As such, and as a 'worst case' scenario, we investigate here the potential effect of unmitigated motions on MR-guided PBS plans, as well as the potential effectiveness of rescanning (arguably the simplest motion management technique) on such treatments.As motion-induced density changes in the beam path could influence not only the proton range but also the proton beam trajectory in the presence of a magnetic field, these results are also compared to reference cases without magnetic fields.

4D dose calculation considering magnetic field
The analytical dose calculation algorithm accounting for the deflection of the proton beams in a perpendicular homogenous magnetic field is described in detail in Duetschler et al (2023b) and we only provide a short description here.The dose calculation is based on ray casting (Schaffner et al 1999), which has been adjusted to account for the deflected beam trajectories.The deflected beam trajectories are determined through the utilization of look-up tables (LUTs) that consist of incremental rotation angles as a function of water-equivalent depth.Specifically, the trajectory is reconstructed iteratively, and in each step the direction is adjusted based on the rotation angle obtained from the LUT.These LUTs were generated from TOPAS MC (Perl et al 2012, Faddegon et al 2020) calculations of single pencil beams (70-229 MeV, 115 energies) in water in orthogonal magnetic fields (0.5/1.5 T).A validation of the algorithm against MC in different media and for patient cases is also presented in Duetschler et al (2023b).
In this work, this algorithm has been further extended to account for deformable motion and motioninduced density changes.The information about the motion and density changes is provided by a 4DCT, from which the motion is extracted in the form of deformation vector fields (DVFs) using deformable image registration (DIR).The first step, for 4D dose calculation, which simulates the interplay effect between the patient's motion and the dynamical beam delivery, is the calculation of the delivery time of each pencil beam.Based on the delivery time, the pencil beams are then assigned to the closest phase of the 4DCT.Considering the impact of the magnetic field as described above, the dose contributions to all phases are then calculated.The doses from all phases are finally warped according to the DVFs and accumulated on a reference phase.In this work, DIRs were performed using a 3D multi-resolution B-spline registration with the open-source software Plastimatch5 were employed.

Investigated cases 2.2.1. Geometrical phantoms
As a first check on the correctness of the developed 4D dose calculation algorithm, it was first applied to the two digital geometrical phantoms depicted in figure 1.Each phantom comprises a stationary high-density 45°wedge (1000 HU) and a spherical target with a diameter of 70 mm (0 HU).The sphere is positioned within either water (0 HU) or lung tissue (−700 HU), referred to as 'Phantom A' and 'Phantom B', respectively.In order to simulate respiratory motion, the spherical target and the surrounding material undergo rigid translation, while the wedge remains stationary.Specifically, a one-dimensional sin 4 motion is applied in the superior-inferior (SI) direction, with an amplitude of 15 mm and a period of 5 s.

4DCT(MRI) patient data sets
For three liver and two lung cancer patients, anatomy and density information is provided by their respective clinical 3DCTs.The 3DCTs for the liver patients were acquired during end exhalation (EE), whereas the lung 3DCTs were obtained during deep-inspiration breath-hold (DIBH) (Josipovic et al 2016).We also refer to these 3DCTs as reference phases.For investigating realistic irregular respiratory motion, the 3DCTs were animated with multiple breathing cycle motion extracted from 4DMRIs of healthy volunteers (von Siebenthal et al 2007, Jud et al 2018) to generate so-called 4DCT(MRI)s as described in Boye et al (2013a) for liver and Duetschler et al (2022) for lung.The used liver and lung CTs have a resolution of 1.96 × 1.96 × 2.5 mm 3 and 1.96 × 1.96 × 2 mm 3 , respectively.
One volunteer 4DMRI consisting of 10 distinct breathing cycles, with a temporal resolution of 2.77 Hz, was used to generate the three liver 4DCT(MRI)s.In contrast, for the two lung 4DCT(MRI)s, a second 4DMRI comprising 15 breathing cycles, with a temporal resolution of 2.25 Hz, was utilized.The liver 4DMRI has a spatial resolution of 1.8 × 1.8 × 3-4 mm 3 (von Siebenthal et al 2007), while the lung 4DMRI has a uniform voxel size of 3.125 × 3.125 × 3.125 mm 3 (Duetschler et al 2022).
For every case and breathing cycle, we computed the median and 95th percentile motion amplitude across all voxels within the CTV.The resulting motion variability for the studied 4DCT(MRI)s is summarized in figure 2. For both indications, the differences in amplitude between the CT anatomies can be attributed to different tumour locations.The liver and lung 4DCT(MRI)s have an average period of 6.6 ± 0.8 s and 4.8 ± 0.4 s, respectively.
The same data was used to investigate uncertainties associated with PBS proton 4D dose calculation in Duetschler et al (2023a), where the data is described in more detail.Furthermore, the analytical dose calculation algorithm considering the impact of the magnetic field was validated against MC for the Liver 1 and Lung 1 cases in Duetschler et al (2023b).

Planning CT and target volume definition
For treatment plan optimisation, the clinical target volumes (CTV) of all phases of the first breathing cycle were combined to form the internal target volume (ITV).The ITV was further extended by a 2 mm margin to form the planning target volume (PTV) and ensure target coverage in the static case.For the planning CT, the mean intensity throughout the first breathing cycle was calculated for all voxels.However, for voxels in the ITV, the maximum intensity value was utilized, following the definition by Botas et al (2018).

Treatment planning
For simulating a patient lying in an MRI scanner, we assumed a uniform magnetic field within a cylindrical region surrounding the isocenter for all dose calculations and treatment plan optimisations.Like both commercially available photon MR-linac systems, a radius of 35 cm was assumed.Uniform field strengths of 0.5 T and 1.5 T were simulated and compared to the scenario without any magnetic field (0 T).
On the static planning CT, treatment plans were optimised as described in Duetschler et al (2023b) assuming magnetic field strengths of 0, 0.5 and 1.5 T. A single field (see figure 1) was selected for the geometrical phantoms, while simplified treatment plans with two fields were optimised using a single-field uniform-dose (SFUD) approach for the 4DCT(MRI) liver and lung cases.Furthermore, to correct for the different beam entrance and beam paths within the patient introduced by the magnetic field, additional treatment plans were optimised with gantry rotations of +5°and +15°for 0.5 T and 1.5 T, respectively.This resulted in similar beam paths within the patient as compared to the scenario without any magnetic field.Plans were optimized with a 2.5 mm distal and 4 mm lateral spot spacing.

4D dose calculation parameters
For all dose calculations, the beam model of PSI-Gantry2 was employed (Pedroni et al 2004, Zenklusen et al 2010, Safai et al 2012) and a uniform dose grid spacing of 2.5 mm 3 was used.Similarly, 4D dose calculations are based on the delivery dynamics of PSI-Gantry2, which are used to calculate the estimated spot delivery sequence.Spot delivery times are calculated assuming 3.5 ms and 4.0 ms for both lateral scanning directions and an energy switching time of 80 ms.
To assess the full influence of motion, we conducted 4D dose calculations without any motion mitigation.Further, we simulated the scenario of volumetric rescanning (VS) performed eight times.These two motion scenarios are denoted as 'VS1' and 'VS8', respectively.Furthermore, as the start of the treatment delivery is usually not synchronized with the respiratory breathing state, which has been shown to influence substantially the dosimetric plan quality (Duetschler et al 2023a), 10 equally distributed starting phases in the first breathing cycle were simulated for each 4D dose calculation scenario.
The 4D dose calculations for the three liver and two lung patient cases all included irregular free breathing motion, as provided by the multiple-cycle 4DCT(MRI)s.

Plan evaluation
To assess the effects of motion, we compared the 4D dose distributions, which were accumulated on the reference phase, with the static 3D dose distribution as optimized on the planning CT.Additionally, we evaluated dose-volume histograms (DVH) of the reference phase CTV.Further, V 95% and D 5% -D 95% within the CTV as measures of target dose coverage and homogeneity were evaluated for each magnetic field strength.Furthermore, the mean dose to the liver or lung surrounding the CTV was studied.

Geometrical phantoms
Static and 4D dose distributions for Phantom A and B are displayed in figures 3 and 4.These dose distributions illustrate the impact of different magnetic field strengths and gantry angle adjustments.As discussed in Duetschler et al (2023b), a similar static plan quality is achieved independent of the magnetic field strength.Moreover, a similar plan quality is also achieved when altering the gantry angle to emulate the beam trajectory without magnetic field influence.For all five plans of both phantoms, motion and its induced interplay with the dynamic beam delivery leads to substantial hot and cold spots in the CTV for no rescanning (VS1).This is particularly evident in the sagittal view, while the axial plane highlights the impact of the magnetic field on the treatment plan.For Phantom A 8-times rescanning (VS8), leads to an improvement in the dose coverage and homogeneity for plans with no or low magnetic field, while a further dose degradation can be observed for 1.5 T in figure 3.For phantom B (figure 4), however, both plans with 1.5 T have the best target coverage and homogeneity for simulations with VS8.This is also visible in figure 5, which further shows a comparable impact on the plan quality due to motion (VS1) for 0 T (black), 0.5 T (red) and 1.5 T(blue).
The dose distributions and DVHs in figures 3-5 were calculated assuming the start of the treatment delivery at the start of the first breathing cycle.On the other hand, as the start of the treatment delivery is typically not synchronized with the patient's breathing, the mean values from 10 different starting phases for V 95% and D 5% -D 95% for the CTV are listed in tables 1 and 2 for Phantom A and B, respectively.For both geometrical phantoms, the respiratory motion has a similar impact on the CTV dose coverage and homogeneity independent of the magnetic field strength.Furthermore, it should be remarked that even 8-times VS was relatively ineffective as a motion mitigation technique for both studied phantoms and could not restore the CTV dose coverage V 95% , which remained well below 95% for all scenarios.

4DCT(MRI) patient data sets
Dose distributions for two example cases, Liver 1 and Lung 1, are shown in figures 6 and 7.For both cases, comparable static plans are achieved for all three magnetic field strengths and gantry angle corrections of +5 or 15°for 0.5 T and 1.5 T results in comparable beam paths compared to the 0 T plan.The 4D dose distributions without any motion mitigation show pronounced hot and cold spots with different interplay patterns for the different plan configurations.Especially for the Liver 1 case (figure 6), 8-times volumetric rescanning (VS8) restores a mostly homogeneous dose distribution for all five plans.DVHs for these two example cases can be found in figure 8.A plan degradation compared to the static plan for no rescanning can be observed for both cases with the worst CTV coverage for 0 T for Liver 1 and Lung 1. Especially for Liver 1, both plans within a 1.5 T magnetic field result in a better target coverage for VS1.   Figure 9 summarizes the results for all five 4DCT(MRI)s and using different starting phases for the delivery simulations.Apart from Liver 3, respiratory motion (VS1) results in a more pronounced degradation of the CTV dose coverage V 95% for plans without any magnetic field than for 1.5 T (see figure 9(a)).On average a decrease of V 95% of the CTV for VS1 compared to the static case of 19.9% was observed for the 0 T plan.A smaller decrease in the CTV coverage V 95% of 17.0% (17.7%) was observed for the 1.5 T (+15°) plan.Also 0.5 T (+5°) resulted in on average slightly smaller differences in V 95% due to motion of 18.9% (17.4%).The average increase in CTV D 5% -D 95% from the static plan due to motion is also slightly reduced for the 1.5 T plans (16.8% and 17.7%) compared to the 0 T plan (18.2%).However, larger average differences occur for 0.5 T (19.0% and 18.6%).In summary, across all five analysed cases, we observed a comparable effect on CTV dose coverage and homogeneity caused by motion, irrespective of the magnetic field strength.
8-times rescanning results in improved target dose coverage and homogeneity.For 0 T, an average difference to the static plan of −4.9% and 5.8% in V 95% and D 5% -D 95% in the CTV has been found.Calculations with a    magnetic field result in average differences compared to the static plan in the range of −2.6% (0.5 T) and −4.5% (1.5 T +15°) for the CTV coverage V 95% .Average differences in the CTV homogeneity D 5% -D 95% range between 3.3% (1.5 T) and 5.9% (0.5 T +5°).It should further be remarked that especially for the two lung cases with pronounced density heterogeneities, an acceptable target coverage (V 95% > 95%) could not be restored in our simulations, with VS8.For normal tissue doses, respiratory motion resulted in differences up to 5.9% in the mean dose to the surrounding liver/lung compared to the static plan.The impact of motion was comparable for all planning scenarios and independent of the magnetic field.The different beam trajectories introduced by the magnetic field also only resulted in small differences in the static mean liver/lung dose up to 1%.

Discussion
To the best of our knowledge, this paper represents the first study into the combined impact of magnetic fields and motion on PBS proton therapy dose distributions.Our analysis of the studied configurations revealed that the influence of motion on treatment plans was similar both with and without the presence of a magnetic field.Thus, our work provides valuable evidence that MRI-guided proton therapy, with adequate motion management, could be considered as an option for treating mobile tumours using similar motion mitigation methods as for normal PBS proton therapy, i.e. rescanning as studied here.Furthermore, for some of the studied cases, the plans optimised in the presence of a magnetic field were even slightly less affected by the respiratory motion.A possible explanation could be the planning strategy used in this study, where spot positions were shifted according to an energy-specific lateral scanning shift before the optimisation considering the magnetic field (Duetschler et al 2023b).Thus, while for plans without any magnetic field spots are placed on a regular rectilinear grid, this grid is distorted in the presence of a magnetic field due to the different materials in the beam path.It should, however, be further investigated, whether a slightly irregular spot placement could result in treatment plans which are more robust to respiratory motion and further cases should be investigated.
The effect of motion on the dosimetric plan quality can be greatly influenced by the proton beam direction (Knopf et al 2011, Chang et al 2017).Our treatment planning strategy results not only in a different deflected beam trajectory within the patient but also in an apparent difference in the beam entrance.As this could influence the impact of motion, additional treatment plans with +5°and +15°gantry rotations for 0.5 and 1.5T were generated and used to also study the impact of motion.The angles were manually selected for both magnetic field strength and could correct for the beam deflection and generate very similar treatment plans than without any magnetic field.Manually tuned gantry angle corrections were also used by Kurz et al (2017) for investigating the treatment plan robustness of MRI-guided proton therapy treatment plans.Instead of a manual gantry angle correction, Burigo and Oborn (2021) have also suggested automatic methods minimizing the mean path difference within the patient compared to the straight dose deposition without magnetic field.In this study, no substantial differences in the interplay effect with or without gantry angle correction could be observed.However, the different beam trajectories could result in significantly different dose distributions to critical nearby OARs.
In this work, the magnetic field was directed along the most pronounced SI direction of motion.The magnetic field perpendicular to the beam direction thus results in a deflection perpendicular to the main motion axis.This setup is motivated by the assumption that the patient would be treated in supine within the bore of an MRI.Recently, there has however been growing interest in upright patient positioning (Volz et al 2022).This could considerably reduce the costs associated with MRI-guided proton therapy, which could be implemented using a fixed horizontal beamline and an open-bore MRI with an upright patient positioning system to rotate the patient.Indeed, this is one of the motivations for investigating here the effects on plans with a 0.5 T magnetic field.For such a setup the magnetic field would be perpendicular to the SI motion direction.Different relative orientations of motion, proton beam and perpendicular magnetic field were investigated in a preliminary study, in which no significant differences could be observed.However, different configurations should be investigated separately before clinical use.
Most 4D dose calculation algorithms used for PBS proton therapy are based on warping and accumulating dose contributions to the different phases of a 4DCT.The temporal resolution of the 4D dose calculation is thus given by the temporal resolution of the 4DCT, which can be much coarser than the time between the delivery of two spots.A finer temporal resolution can be achieved using the deforming dose grid algorithm (Boye et al 2013a, Krieger et al 2018, Zhang et al 2019), which deforms the dose calculation grid according to the DVFs extracted from the 4DCT.The DVFs can then be interpolated in time for a finer (sub-)spot-wise dose calculation.Depending on the motion period, delivery dynamics and fractionation dose the temporal resolution can result in substantial dosimetric differences (Zhang et al 2019).Indeed, the 4DCT(MRI) cases used in this study have previously been used to investigate the limitations of phase-sorting 4D dose calculation (Duetschler et al 2023a).In that study, only a small dosimetric impact due to the temporal resolution of the 4D dose calculation was found.Consequently, a dose warping approach was chosen in this study, but a combination with the deforming dose grid algorithm is envisaged in future developments.In the same study, it was also shown that neglecting variations in respiratory motion can lead to a substantial underestimation of dosimetric motion effect.Thus, in this study, 4D dose calculations were performed for multiple-cycle 4DCT(MRI)s, which reflect free breathing motion.It should further be remarked, that one of the lung 4DCT(MRI)s used in Duetschler et al (2023a) was excluded for the present study.This is due to the current limitations of the analytical dose calculation algorithm in the presence of magnetic fields (Duetschler et al 2023b), which does not support the use of pre-absorbers, which would be necessary to achieve an acceptable target coverage for that case.
For better comparability, we chose to use the same planning target definition for treatment plans optimised with and without magnetic fields.Furthermore, positioning errors were not simulated, Fuchs et al (2017) however have previously shown similar robustness of treatment plans with and without magnetic field.On the other hand, the superior soft-tissue contrast of MRI-guided PBS proton therapy could result in more accurate patient positioning and the consequently reduced margins could result in better sparing of healthy tissue and OARs.Furthermore, MRI guidance could result in improved accuracy of other motion mitigation techniques, such as breath-hold, gating or tracking, which usually rely on internal or external surrogates.For instance, online cine-MRI could provide direct information about the tumour and surrounding anatomy in real-time.

Conclusion
We have presented the first study of the combined effects of magnetic fields and respiratory motion on the dose for PBS proton therapy.For the studied cases, with the magnetic field parallel to the most pronounced motion along the SI direction and perpendicular to the incident proton beams, a similar dosimetric effect due to motion was observed regardless of magnetic field strength.Depending on the planning strategy, the presence of a magnetic field could even slightly increase the robustness of the treatment plan to motion.

Figure 1 .
Figure 1.Phantom with high-density wedge (1000 HU) and spherical target (0 HU) surrounded by water (0 HU, Phantom A) or lung tissue (−700 HU, Phantom B).The CTV is delineated in white and the ITV in black.The direction of the proton beam is indicated by a red arrow.

Figure 3 .
Figure 3. 4D dose distributions for Phantom A (0 HU) for the different magnetic fields (0, 0.5 and 1.5 T) and with angle correction (0.5 T +5°and 1.5 T +15°).Dose distributions in the static planned case (left), without rescanning (VS1) (centre) and 8-times rescanning (VS8) (right) are shown.The CTV and PTV are delineated in white and for each scenario a sagittal and axial plane is shown.

Figure 4 .
Figure 4. 4D dose distributions for Phantom B (−700 HU) for the different magnetic fields (0, 0.5 and 1.5 T) and with angle correction (0.5 T +5°and 1.5 T +15°).Dose distributions in the static planned case (left), without rescanning (VS1) (centre) and 8-times rescanning (VS8) (right) are shown.The CTV and PTV are delineated in white and for each scenario a sagittal and axial plane is shown.

Figure 5 .
Figure 5. DVHs for CTV of Phantom A (left) and Phantom B (right) for 0 T (black), 0.5 T (red) and 1.5 T (blue).Dashed red and green lines represent plans 0.5 T and 1.5 T with gantry rotated by +5°and +15°, respectively.Results for the static treatment plan (top), no rescanning (middle) and 8-times rescanning (bottom) are shown.

Table 1 .
DVH parameters for CTV of Phantom A for different magnetic field strengths for the planned static case, without rescanning (VS1) and with 8-times rescanning (VS8) and their differences.For VS1 and VS8 mean values and standard deviation for 10 different starting phases are stated [%].

Figure 8 .
Figure 8. DVHs for CTV of Liver 1 (left) and Lung 1 (right) for 0 T (black), 0.5 (red) and 1.5 T (blue).Dashed red and green lines represent plans 0.5 T and 1.5 T with gantry rotated by +5°and +15°, respectively.Results for the static treatment plan (top), no rescanning (middle) and rescanning (bottom) are shown.