An effective method to reduce the interplay effects between respiratory motion and a uniform scanning proton beam irradiation for liver tumors: A case study

Abstract Purpose For scanning particle beam therapy, interference between scanning patterns and interfield organ motion may result in suboptimal dose within target volume. In this study, we developed a simple offline correction technique for uniform scanning proton beam (USPB) delivery to compensate for the interplay between scanning patterns and respiratory motion and demonstrate the effectiveness of our technique in treating liver cancer. Methods The computed tomography (CT) and respiration data of two patients who had received stereotactic body radiotherapy for hepatocellular carcinoma were used. In the simulation, the relative beam weight delivered to each respiratory phase is calculated for each beam layer after treatment of each fraction. Respiratory phases with beam weights higher than 50% of the largest weight are considered “skipped phases” for the next fraction. For the following fraction, the beam trigger is regulated to prevent beam layers from starting irradiation in skipped phases by extending the interval between each layer. To calculate dose‐volume histogram (DVH), the dose of the target volume at end‐exhale (50% phase) was calculated as the sum of each energy layer, with consideration of displacement due to respiratory motion and relative beam weight delivered per respiratory phase. Results For a single fraction, D1%, D99%, and V100% were 114%, 88%, and 32%, respectively, when 8 Gy/min of dose rate was simulated. Although these parameters were improved with multiple fractions, dosimetric inhomogeneity without motion management remained even at 30 fractions, with V100% 86.9% at 30 fractions. In contrast, the V100% values with adaptation were 96% and 98% at 20 and 30 fractions, respectively. We developed an offline correction technique for USPB therapy to compensate for the interplay effects between respiratory organ motion and USPB beam delivery. Conclusions For liver tumor, this adaptive therapy technique showed significant improvement in dose uniformity even with fewer treatment fractions than normal USPB therapy.


| INTRODUCTION
Particle beam therapy has become an important tool in radiation oncology for cancer treatment due to its improved dose distribution compared to photon beam treatments. 1,2 However, particle beams are generally more sensitive to organ motion than photon beams. [3][4][5] This sensitivity is more pronounced in medium with inhomogeneity such as lung and bones, where inter-or intrafractional organ motion may significantly change the radiological path length (particle range), thus affecting delivered dose distribution.
Due to concerns over neutron dose in proton beam, beam scanning has recently become an attractive choice. [6][7][8][9] However, scanned beams are more susceptible to the perturbation caused by scanning motion and interfield organ motion that should be considered. 10,11 This interplay typically results in under-or overdosage within the target volume, depending on the motion and scanning pattern. Phillips et al. 12 showed that dose uniformity depends on the motion amplitude relative to the direction of beam motion and target motion. Lambert et al. 13 assessed the interplay for two different scan directions in proton beam therapy and concluded that target margins is not effective in compensating for the effects of intrafractional motion in scanned beam therapy. Furthermore, although range-based internal target volume (ITV) is commonly used, complexity of using ITV for particle therapy has also been reported. 14 Recently, several studies have investigated rescanning techniques to reduce interplay effects to improve dose homogeneity. 5,9,15,16 In this study, we developed a simple offline correction technique for uniform scanning proton beam (USPB) delivery to compensate for the interplay between beam scanning and respiratory motion. Here, we demonstrate the effectiveness of this technique for treatment of liver cancer patients.

2.A | Uniform scanning proton beams
The proton beam used in this study is a nearly continuous beam with an initial energy of 208 MeV. The vertical scanning frequency is 144 Hz, whereas the horizontal scanning frequency depends on the field size (i.e., 14.4 Hz for 10 scanning lines). A detailed description of the USPB technique has been described elsewhere. 6,7 Unlike spot scanning for intensity-modulated proton therapy (IMPT), the effects of scanning within the iso-energy layer will be negligible because the scan speed is much faster than the respiratory cycle. To produce a uniform spread-out Bragg peak (SOBP), a range modulator consisting of binary combination of graphite plates was used to pullback the pristine Bragg peak to different ranges. A 0.5-s interval is required to change the energy layer by switching the range modulator and is accounted for in the calculation. Because beam layers are pulled back in sequence, a time delay occurs with each beam delivery, consequently leading to interplay effects with moving organs.
2.B. | Patients, respiratory motion, and treatment planning Under Institutional Review Board (IRB) exemption, the CT and respiration data of two patients who received stereotactic body radiotherapy (SBRT) for hepatocellular carcinoma were used for simulation. In both cases, the planning target volume (PTV) does not include air (i.e., lung volume). Because the density of liver and surrounding soft tissue is uniform, the variation in dose distribution due to respiratory organ motion is relatively small. Table 1 shows the size, region, and maximum motion range of target volume. The gross tumor volume (GTV) was delineated on the four-dimensional (4D)-CT images of the 0%, 25%, 50%, and 75% respiratory phases, and the contours were copied onto the free-breathing CT images.
Clinical target volume (CTV) margin was zero. The original PTV for SBRT treatment was generated by adding a margin to the rangebased ITV, which was the merged volume of the GTVs. Respiratory motion of liver tumors were evaluated with orthogonal cine-MRI images as shown by Akino et al. 17 Sagittal and coronal images were acquired for 30 s with the same immobilization of treatment to evaluate the motion amplitude of diaphragm and respiration stability. The motion vectors between two continuous images were analyzed using an optical flow estimation algorithm known as the pyramidal Lucas-Kanade method. 18,19 After cine-MRI image acquisitions, planning CT and 4D-CT images were acquired within 15 min.
The 4D-CT images were sorted into eight respiratory phases. The phases of 0% and 50% accommodate the end-inhalation and endexhalation phases, respectively.   does not provide dosimetric data from each layer in the current version of the software. We therefore considered a rectangle dose distribution that includes the original PTV for the sake of simplicity.
We also created multiple rectangular target volumes by varying their thickness in 5-mm steps and generated treatment plans for each tar-  After treatment of each fraction, the relative beam weight delivered to each respiratory phase is calculated for each beam layer

2.D. | Evaluation of motion interplay effects
To calculate the dose-volume histogram (DVH), the GTV on the 50% (end-exhalation) phase of 4D-CT was evaluated, as patient respiration is most stable in this phase. 20 Instead of evaluating range-based ITV, which is a merged volume at each phase, the GTV was moved along the respiratory motion vector, and the average of dose at each position was evaluated for each voxel of GTV. The GTV dose was calculated as the sum of each energy layer with compensation for displacement due to respiratory motion and relative beam weight delivered in each respiratory phase [ Fig. 1(c)]. The Di, which is the dose delivered to a voxel of target with the 3D coordinates of Xi, was calculated using the following formula: where D k represents the dose of the pristine peak of the kth energy layer; δ t represents the displacement vector due to the respiratory motion at the respiration phase, t; and W t represents the weight of the where D 2% , D 50% , and D 98% represent doses of 2%, 50%, and 98% target volume, respectively. for layers with short beam-on time than long. Although data from runs both with and without adaptation showed fraction-dependent decreases in SD, plans with adaptation showed a much more rapid decrease with increasing fraction number than those without. Figure 4 shows the range of variation in DVH calculated 500 times for 2 and 8 Gy/min dose rates. The DVH for the 2 Gy/min dose rate was steeper than that for the 8 Gy/min dose rate, which showed large variation with a higher maximum dose and lower minimum dose than the 2 Gy/min dose rate, representing hot and cold spots. In contrast, the histogram of patient #2 showed little difference between 2 and 8 Gy/min dose rates.   To overcome the effects of organ motion, the gating technique 23 has been used for moving targets, such as lung 24,25 and liver 26 tumors. Based on a signal from a motion-monitoring device, the beam is delivered only during specific parts of the breathing cycle.

| RESULTS
Since exhalation is the most reproducible respiratory state, the endexhalation phase is often selected as the gating window. 27 For ion beams including proton and carbon beams, the rescanning technique has been investigated in an effort to reduce dose inhomogeneity. 5,15,16 In this technique, treatment delivery is repeated N times within each fraction, with the number of particles reduced to 1/N per rescan. Multiple scans will lead to an averaging effect of the interference pattern if it can be ensured that the motion parameters such as initial phase or respiratory cycle differ for each rescan. Furukawa et al. 28 showed that the phase-controlled rescanning method with a large number of rescans improved dose delivery for moving targets. The benefits of the gating technique include minimized interplay effects and potential to reduce field sizes, leading to desired dose delivery to target. However, treatment time increases with gating due to the frequent interruption of the beam. The offline adaptation proposed in this study is simple with active correction, and the difference in treatment time between plans with and without correction is less than 10 s. The baseline shift of patient respiration results in inappropriate beam delivery and often prolongs the treatment time of gating radiotherapy. 29 In contrast, the baseline shift will not greatly affect our method, especially in terms of treatment time.
With this correction technique, uniform dose delivery will be achievable without reducing efficiency of treatment due to elongated treatment time. will be helpful for beam regulation with patient respiration. For proton beam therapy, surrogate markers placed on patient chest or abdomen to detect respiration signal may affect the dose distributions. A laser-based or optical camera-based devises will be able to provide the signal without affecting the proton beam delivery. 32 In clinical practice, the range of the ion beam will be affected by density variations, especially in regions including air and ribs. Here, we excluded cases with liver tumors near diaphragm dome and examined patients whose PTV volumes did not include air. For lung cancer treatment, dose calculation will be necessary for each respiratory phase of 4D-CT. Because the tissue surrounding a liver tumor is solid and the density uniform, the variation in dose distribution due to respiratory motion will be small. However, skin motion due to respiration may also affect the dose distribution, leading to an increase in uncertainties in this study.
With respect to scan direction, several reports have suggested that scanning planes should be perpendicular to the motion direction. 13,33 In the present study, treatment plans were created using a single beam to generate various sizes of SOBP. Therefore, the dose variation in the lateral and superior-inferior directions is very small, resulting in an underestimation of interplay effects. As shown in Table 1, the motion in the anterior-posterior and lateral direction of patient # 2 is less than 2 mm, resulting in small motion effects.
In addition, the target volume of patient #2 is much larger than that of patient #1. In an actual treatment, longer beam-on time due to a large target volume may lead to modest interplay effects. Generally, two or more beams are used for proton therapy. If two orthogonal beams are used with our methodology, the interplay effects observed in one beam may not appear in another beam because of simple dose distribution. However, if a compensator is applied to the distal edge of a sphere-shaped target volume, the interplay effects could become more complex and larger, due to the complex dose distribution of each layer. In such cases, motion management may be the choice of treatment to ensure accurate proton therapy.

| CONCLUSION
We have developed an offline correction technique for USPB therapy to compensate for the interplay effects between respiratory organ motion and layer-by-layer beam delivery. For the treatment of liver tumors, this adaptive therapy technique showed a significant improvement in dose uniformity with fewer treatment fractions.