Measurement‐based study on characterizing symmetric and asymmetric respiratory motion interplay effect on target dose distribution in the proton pencil beam scanning

Abstract Pencil beam scanning proton therapy makes possible intensity modulation, resulting in improved target dose conformity and organ‐at‐risk (OAR) dose sparing. This benefit, however, results in increased sensitivity to certain clinical and beam delivery parameters, such as respiratory motion. These effects can cause plan degeneration, which could lead to decreased tumor dose or increased OAR dose. This study evaluated the measurements of proton pencil beam scanning delivery made with a 2D ion chamber array in solid water on a 1D motion platform, where respiratory motion was simulated using sine and cosine4 waves representing sinusoidal symmetric and realistic asymmetric breathing motions, respectively. Motion amplitudes were 0.5 cm and 1 cm corresponding to 1 cm and 2 cm of maximum respiratory excursions, respectively, with 5 sec fixed breathing cycle. The treatment plans were created to mimic spherical targets of 3 cm or 10 cm diameter located at 5 cm or 1 cm depth in solid water phantom. A reference RBE dose of 200 cGy per fraction was delivered in 1, 5, 10, and 15 fractions for each dataset. We evaluated dose conformity and uniformity at the center plane of targets by using the Conformation Number and the Homogeneity Index, respectively. Results indicated that dose conformity as well as homogeneity was more affected by motion for smaller targets. Dose conformity was better achieved for symmetric breathing patterns than asymmetric breathing patterns regardless of the number of fractions. The presence of a range shifter with shallow targets reduced the motion effect by improving dose homogeneity. While motion effects are known to be averaged out over the course of multifractional treatments, this might not be true for proton pencil beam scanning under asymmetrical breathing pattern.


| INTRODUCTION
Highly conformal proton pencil beam scanning (PBS) dose distributions generated with intensity-modulated proton therapy (IMPT) improve the therapeutic ratio, achieving highly conformal target doses while reducing toxic doses to surrounding organs-at-risk (OARs). However, since there is interference between PBS delivery and moving target, known as the interplay effect, 1,2 and typical breathing cycles are on the same scale as the time required for the switch between two adjacent energy layers, 3 the superior dose distribution is more sensitive to respiration-induced organ motion for treatment sites such as lung, liver, and mediastinum, which can cause temporal displacement of the target volume and thus degrade the proton dose distribution significantly. 4,5 It was shown that during dynamic proton beam scanning, intrafractional organ motion induces up to 100% of the target to receive a dose outside the International Commission on Radiation Units and Measurements (ICRU) recommended limits with a minimal dose down to 34% of the prescribed dose in the extreme cases. 6 To mitigate the motion interplay effect, several methods including respiratory gating, breath hold, tumor tracking, and repainting have been investigated or clinically implemented. [7][8][9][10][11][12] There were several studies on interplay effects for dynamic delivery of charged particles. Fundamental water phantom-based computer simulation study of dose distribution in the presence of respiratory motion with extensive parameters was performed in the Paul Scherrer Institute, 13 which showed rescanning the target volume with fractionation improves the dose uniformity. Homogeneity degradation of dose distribution with increasing motion of moving target was shown using radiographic film measurements and confirmed by real patient 4DCT-based treatment planning study. 1 Further studies including motion alleviation techniques such as increasing number of fractions or number of scanning of the target in order to mitigate interplay effects were investigated. The different repainting techniques 3 as well as different scanning modes 2,14 in PBS were also investigated. Gating and rescanning combined phase-controlled rescanning has also been studied for carbon spot scanning. 15 Boria et al 16  dose calculation routine for PBS using the time structure of the pencil beam spot delivery from a system log files with validation via 2D array ion chamber and motion platform. Fraction-wise retrospective dose reconstruction and accumulation was investigated using machine log files in combination with the patient's breathing patterns from a pressure belt system and 4D CT datasets through entire treatment course. 18 The deforming grid 4D dose calculation techniques have been employed to predict and validate the pattern of 4D dose distribution 19 and to evaluate different PBS rescanning techniques for moving targets. 20 Several studies have investigated 4D robust optimization for mitigating the interplay effects in scanned particle beam therapy. [21][22][23][24] Whereas most of the studies were performed using simulation models in planning data or in phantom model, our investigation is solely measurement-based study by delivering PBS plans, where the actual fractional dose of 200 cGy was delivered multiple times for a given number of fractions. The goal of this study is to investigate the interplay effect of PBS for different breathing patterns: sinusoidal symmetric motion vs more realistic asymmetric motion, which are simulated by a commercially available respiratory motion plat-    In other words, fractionation was measured using a random initial phase for each fraction. Then, the number of single fraction measurements was added up accordingly to make a multifraction measurement data.

2.D | Evaluation metrics and analysis
In order to provide a dose conformity score, instead of using a conformity index defined as the ratio of reference isodose volume to target volume, which is described in the ICRU Report 62 25 we used a metric called conformation number (CN) introduced by van't Riet et al 26 : where TV RI is target volume covered by the reference isodose, TV is target volume, and V RI is volume of the reference isodose. In this study, 95% of the prescribed dose is used as the reference isodose.
The first term in the right-hand side of Eq. (1) is a modified conformity index that correctly determines the quality of irradiation of the target volume, where 0 and 1 indicate that none of the target volume is located inside the prescription isodose and entire target volume is covered with the prescribed dose, respectively. The second term measures indirectly the volume of surrounding normal tissues involved in the reference isodose in terms of the degree of concordance between target volume covered with the reference isodose and the reference isodose volume, ranging from 0 (no protection of OARs) to 1 (all OARs below the reference isodose). For this study relative CN was used, which was normalized to the CN with no motion since the coverage or conformity for each plan depends on its margin, which is not identical for each different target size and depth.
Homogeneity index (HI) was also used to analyze the uniformity of dose distribution in the target volume. Among various formulae, we chose one defined as follows 27 where D Rx is prescribed dose and D 2% and D 98% are the minimum doses to the 2% and 98% of the target volume, respectively. These are also considered to be maximum and minimum dose, respectively.
In this study, we used D max for D 2% and D min for D 98% , which are practically equivalent considering 2D array ion chamber detector resolution (~0.78 cm) and the number of voxels covered by targets.
Lower HI values indicate more homogeneous target dose distribution.

| RESULTS
In order to estimate the degradation of target dose coverage due to respiratory motions, we evaluated dose conformity and uniformity of each measurement dataset at the center plane of each size of moving targets simulated by the motion platform. Both conformity and homogeneity, however, were significantly recovered as fractionation increased as shown in Fig. 3. To further quantify two different simulated breathing motions, interplay effects of various parameters, and how fractionation mitigates the degradation of dose conformity and uniformity differently for those breathing patterns, CN and HI metrics were used in following subsections.

3.A | Impact of motion on dose conformity
Absolute CNs for all dataset are listed in Table 1 including CNs with no motion. Figure 4 shows relative CNs in percentage, which were

3.B | Impact of motion on dose homogeneity
HI values were shown in Table 2. Figure 5 shows homogeneity as an estimate of the dosimetric influence due to the patterns of respiratory motion as a function of fraction where green lines were added for each plot to represent HI values for static targets as a reference.
Overall homogeneity improved as more fractions were added. For    We found that fractionation alone has some limitation in mitigating the interplay effect in terms of dose conformity, especially in non-sinusoidal asymmetric motion compared to symmetric motion. In addition, Bert et al 29 showed that the most dominant parameter influencing interplay patterns is the motion amplitude and changes in motion parameters or scanning parameters resulted in almost unpredictable dose heterogeneity. Therefore, a well-defined selection of adequate margins and motion management are required to minimize the interplay effect of PBS treatment.
There are few drawbacks of our experimental setup. The study design was to investigate PBS interplay effect with a simple geometric shape of moving target in a homogeneous water phantom and to evaluate the motion-affected dose distributions in 2D plane measurements. This could only represent a surrogate to quantify the volumetric plan quality. Moreover, the interplay effect of PBS delivery with irregular target geometry under realistic patient-specific breathing motion in high degree of heterogeneity of real patient body may be much more complicated to quantify.

| CONCLUSIONS
In this study, the dosimetric interplay effects of different breathing patterns and spot sizes were investigated based on physical measurements. More fractions generally helped mitigate the degradation of dose conformity as well as homogeneity due to respiratory motions. Our study has also confirmed that it is possible to treat moderately moving targets with motion amplitude less than 5 mm using pencil beam scanning in standard fractionation. For the case of small target under relatively large motions, especially with irregular breathing patterns or asymmetric motions where relatively significant amount of time is spent on a certain breathing phase such as end exhale, care must be taken such as further study on breathing patterns for each patient as well as adequate use of motion management.

CONF LICT OF I NTEREST
The authors report no conflicts of interest.