Comparison of pancreatic respiratory motion management with three abdominal corsets for particle radiation therapy: Case study

Abstract Background and purpose Abdominal organ motion seriously compromises the targeting accuracy for particle therapy in patients with pancreatic adenocarcinoma. This study compares three different abdominal corsets regarding their ability to reduce pancreatic motion and their potential usability in particle therapy. Materials and methods A patient‐individualized polyurethane (PU), a semi‐individualized polyethylene (PE), and a patient‐individualized three‐dimensional‐scan based polyethylene (3D‐PE) corset were manufactured for one healthy volunteer. Time‐resolved volumetric four‐dimensional‐magnetic resonance imaging (4D‐MRI) and single‐slice two‐dimensional (2D) cine‐MRI scans were acquired on two consecutive days to compare free‐breathing motion patterns with and without corsets. The corset material properties, such as thickness variance, material homogeneity in Hounsfield units (HU) on computed tomography (CT) scans, and manufacturing features were compared. The water equivalent ratio (WER) of corset material samples was measured using a multi‐layer ionization chamber for proton energies of 150 and 200 MeV. Results All corsets reduced the pancreatic motion on average by 9.6 mm in inferior–superior and by 3.2 mm in anterior‐posterior direction. With corset, the breathing frequency was approximately doubled and the day‐to‐day motion variations were reduced. The WER measurements showed an average value of 0.993 and 0.956 for the PE and 3DPE corset, respectively, and of 0.298 for the PU corset. The PE and 3DPE corsets showed a constant thickness of 2.8 ± 0.2 and 3.8 ± 0.2 mm, respectively and a homogeneous material composition with a standard deviation (SD) of 31 and 32 HU, respectively. The PU corset showed a variable thickness of 4.2 − 25.6 mm and a heterogeneous structure with air inclusions with an SD of 113 HU. Conclusion Abdominal corsets may be effective devices to reduce pancreatic motion. For particle therapy, PE‐based corsets are preferred over PU‐based corset due to their material homogeneity and constant thickness.

were compared. The water equivalent ratio (WER) of corset material samples was measured using a multi-layer ionization chamber for proton energies of 150 and 200 MeV.
Results: All corsets reduced the pancreatic motion on average by 9.6 mm in inferiorsuperior and by 3.2 mm in anterior-posterior direction. With corset, the breathing frequency was approximately doubled and the day-to-day motion variations were reduced. The WER measurements showed an average value of 0.993 and 0.956 for the PE and 3DPE corset, respectively, and of 0.298 for the PU corset. The PE and 3DPE corsets showed a constant thickness of 2.8 ± 0.2 and 3.8 ± 0.2 mm, respectively and a homogeneous material composition with a standard deviation (SD) of 31 and 32 HU, respectively. The PU corset showed a variable thickness of 4.2 − 25.6 mm and a heterogeneous structure with air inclusions with an SD of 113 HU.
Conclusion: Abdominal corsets may be effective devices to reduce pancreatic motion. For particle therapy, PE-based corsets are preferred over PU-based corset due to their material homogeneity and constant thickness. becoming an alternative treatment modality for conventional photon-based radiotherapy (XT) in thoracic and abdominal tumours. This is mainly due to their ability to deposit almost all their dose at the end of their track (the so-called "Bragg peak"), thereby minimizing the dose to organs at risk both in the initial beam path and beyond the Bragg peak. 1 Appropriate immobilisation of the target volume is a key component in the treatment process for precise dose delivery. This is particular important for PT due to its higher sensitivity to density variations caused by inter-and intra-fractional changes in patient anatomy. Morphological changes along the beam path due to organ motion, deformation and organ filling can influence the position of the Bragg peak relative to the target volume. Furthermore, the use of active dose delivery techniques that employ scanned beams to volumetrically scan a mono-energetic Bragg peak over the target volume in combination with intra-fractional organ motion may further degrade precise dose delivery due to the interplay effect. [2][3][4][5][6] This may result in over-or underdosage of the target volume and additional unwanted dose deposition in adjacent organs at risk.
For thoracic and abdominal organs, the main source of motion is respiration. To reduce respiration-induced uncertainties in PT, breathing motion can be accounted for by 4D treatment planning and optimizations, 3,6-11 tumour tracking, 12,13 gating scenarios, [14][15][16][17] or by physically reducing respiratory-induced motion by abdominal compression techniques. [18][19][20] For PT of targets in the upper gastrointestinal tract, including the liver and pancreas, motion mitigation by means of abdominal compression bands or pressure plates can, however, exacerbate the range uncertainties due to a poor setup reproducibility and consequential edge effects of the devices used. 21 In particular, pressure plates only allow for limited usability due to their bulky setup.
As a solution to reduce breathing induced tumour motion in patients with pancreatic cancer, polyurethane-based customized abdominal corsets have recently been used in stereotactic XT. 18 Studies using magnetic resonance imaging (MRI) in multi-planar 2Dcine or 4D mode have shown the ability of abdominal corsets to reduce pancreatic motion mainly in inferior-superior direction. [18][19][20] However, the corsets used in XT may not be applicable for PT, since for the latter, the reproducibility of the setup for immobilization devices placed in the beam path and an exact knowledge of their material properties (i.e. thickness and homogeneity) is critical to calculate the beam penumbra and range in the patient. When placed in the beam path, immobilization devices modify the position of the Bragg peak, and therefore it is crucial to understand how the dose distribution is affected. 21 In PT, the water equivalent ratio (WER) of any material placed in the beam path needs to be known to ensure that the treatment planning system is able to accurately take the effects on beam penumbra and range into account. Furthermore, it is unknown to which level of customization the corsets need to be manufactured in order to achieve an adequate level of motion reduction, while a reproducible positioning is still ensured. This needs to be investigated before corsets can be clinically used for PT. Therefore, in this study we compared three types of abdominal corsets that differ in terms of material composition, thickness, homogeneity, size, and degree of patient customization. The goal of this study was to determine which corsets are suitable for PT regarding their motion reduction capabilities and their material properties.

2.B.2 | Water equivalent ratio
The WER is defined as the ratio of the mass thickness of water and the given material (in g/cm 2 ) that leads to the same beam energy loss. 22 Table 1.  4D-MRI showed that all three corsets reduced pancreatic COM motion in both IS and AP direction. For all corsets, the largest absolute motion reduction was found in the IS direction, measuring on average 2.6-8.8 mm with corset as opposed to 13.2 mm without corset. In AP direction, the mean amplitude was reduced from 3.8 mm without corset to 0.5-0.7 mm with corset. In LR direction, mean amplitudes without corset of 1.6 mm were observed, which were decreased for the PU and 3DPE corset to 0.1-0.7 mm and increased by the PE corset to 2.9 mm. The detailed values are illustrated in Fig. 4. Figures 5 and 6 show motion amplitudes of 2D cine-MRI and 4D-MRI with and without the 3DPE corset, respectively. In Fig. 7, a clinically applied PT plan is shown, in which a patient with an adenocarcinoma of the pancreatic head was irradiated with three proton beams with angles between 140°and 208°to deposit a dose of 50.4 Gy within 28 fractions to the internal clinical target volume (iCTV). As illustrated in Fig. 7, the corset was designed such that both the buckles of the hook-and-pile fasteners as well as the edges of the corset opening were placed on the patient's left lateral side and thus they do not interfere with the applied PT beams. With T A B L E 1 Material analysis of the three corsets. The measured thickness of corsets and samples, material homogeneity as well as the average measured water equivalent ratio (WER) for protons at energies of 150 and 200 MeV, respectively, are listed. For corset thickness, sample thickness and material homogeneity, the indicated uncertainties represent the standard deviation of the measured value in multiple measurements (N ≥ 15). The uncertainties of the WER measurement resulted from the uncertainty of the underlying Bortfeld-Fit of the depthdose curve.

Corset
Corset

| DISCUSSION
All three corsets evaluated in this study reduced pancreatic motion, in particular in IS direction. Compared to the study by Heerkens et al. 18 , investigating the motion reduction by use of a PU corset and reporting a mean IS motion reduction of 4 mm with a large inter-patient variability, we observed a mean pancreatic motion reduction of around 8 mm in IS direction in a single subject. We additionally observed an AP motion reduction of 3 mm, which was not the case in the previous study. 18 However, as reported in previous corset studies, [18][19][20] such results are highly patient-specific.
Therefore, from our study no general motion reduction magnitudes can be deduced.
Moreover, as opposed to our corset design, covering the ribs and abdomen in order to reduce both abdominal respiratory motion and chest breathing, the corset used by Heerkens et al. 18 left out the ribs to promote chest breathing. We therefore hypothesize that these different corset designs may contribute to different observed motion reduction patterns.
While Heerkens et al. 18 planned to use the corset for stereotactic XT, we set the focus on its applicability for PT, for which corset thickness and homogeneity are more important than for XT. With regard to the material properties, both the PE and the 3DPE corset show potential to be used in particle therapy since they have a constant thickness and are made of homogeneous material. From a manufacturing point of view, the 3DPE corset is more time-consuming and costly as a consequence of its more patient-specific manufacturing procedure. This, on the other hand, allows for flexibility in the design and hence a positioning of corset opening and buckles, which would avoid beam edge effects. The design of the 3DPE corset considered that beam angles between 40°and 100°were a priori not intended for pencil beam scanned PT planning since the presence of the stomach and the large intestine prevents the usage of left lateral (oblique) beam angles for pancreatic cancer patients. 25,26 Hence, in our study, the left lateral part was chosen as an opening.
The PE corset, on the other hand, shows partially overlapping teners are placed (see Fig. 1). Since for this corset the opening is fixed by design, it compromises the use of anterior beams, since edge effects may occur due to the fact that the beam will pass across edges of the corset layers or the closing buckles. Although the treatment planning system will take the resulting range shifts into account, small shifts in patient position relative to the edges will impact the range of the traversing beam. In order to minimize the impact of air gaps on beam penumbra, the corset should closely However, for XT purposes, an increased skin dose by the corset needs to be considered. 18 For PT, the air inclusions in the PU corset make it unsuitable, since the resulting range shifts lead to increased range uncertainties. 29 The difference in measured thickness between the PE-based sample used for WER determination and the corsets themselves is partially influenced by the different measurement procedures. While depending on the filling of the stomach and bowel. 30 Moreover, substantial weight loss has been observed in these patients during the course of treatment. 31 These factors compromise the setup reproducibility and may require the corset pressure to be adjusted between treatment fractions in order to secure adequate immobilization. This study has shown that pancreatic motion reduction obtained with patient-individualized corsets is similar to that of the semi-specific corset. Therefore, semi-individualized corsets may be sufficient, which would have the advantages of reusability and short manufacturing time. However, substantial inter-patient variability can be expected. This variability may depend on natural breathing patterns (i.e., chest or abdominal breathing), the size and location of the primary tumour, and its infiltration into surrounding tissues. Therefore, a comparative study in a large patient cohort is mandatory to get good statistics and analyse subgroups.

| CONCLUSION
All three abdominal corsets were found to reduce breathing-induced pancreatic motion to a comparable degree in this case study, in particular in inferior-superior direction. Conclusions derived from this case study should be confirmed by a larger study with patients. Due to their well-defined thickness and material homogeneity, as well as the favourable water equivalent ratios, the polyethylene corsets are suitable candidates for abdominal compression in particle therapy.

ACKNOWLEDG MENTS
The authors thank Dr. Patrick Wohlfahrt, Nils Peters (OncoRay) and Martina Jochim (DKFZ) for their support in the CT acquisitions and Nils Peters also for the WER measurements. The authors are further indebted to Joao Seco (DKFZ) for fruitful discussions.

CONSENT FOR PUBLICATION
The patient provided written informed consent to publish treatmentrelated data in an anonymised manner.