Effectiveness of base‐of‐skull immobilization system in a compact proton therapy setting

Abstract Purpose The purpose of this study was to investigate daily repositioning accuracy by analyzing inter‐ and intra‐fractional uncertainties associated with patients treated for intracranial or base of skull tumors in a compact proton therapy system with 6 degrees of freedom (DOF) robotic couch and a thermoplastic head mask indexed to a base of skull (BoS) frame. Materials and methods Daily orthogonal kV alignment images at setup position before and after daily treatments were analyzed for 33 patients. The system was composed of a new type of thermoplastic mask, a bite block, and carbon‐fiber BoS couch‐top insert specifically designed for proton therapy treatments. The correctional shifts in robotic treatment table with 6 DOF were evaluated and recorded based on over 1500 planar kV image pairs. Correctional shifts for patients with and without bite blocks were compared. Results Systematic and random errors were evaluated for all 6 DOF coordinates available for daily vector corrections. Uncertainties associated with geometrical errors and their sources, in addition to robustness analysis of various combinations of immobilization components were presented. Conclusions Analysis of 644 fractions including patients with and without a bite block shows that the BoS immobilization system is capable of maintaining intra‐fraction localization with submillimeter accuracy (in nearly 83%, 86%, 95% of cases along SI, LAT, and PA, respectively) in translational coordinates and subdegree precision (in 98.85%, 98.85%, and 96.4% of cases for roll, pitch, and yaw respectively) in rotational coordinates. The system overall fares better in intra‐fraction localization precision compared to previously reported particle therapy immobilization systems. The use of a mask‐attached type bite block has marginal impact on inter‐ or intra‐fraction uncertainties compared to no bite block.


| INTRODUCTION
Patient immobilization is critical to the safe and accurate delivery of radiation therapy. This is especially critical in particle therapy because of strong dependence of beam range on even the smallest variation in patient position with respect to the reference conditions. 1,2 Thus, it is important to minimize uncertainties associated with patient motion by understanding limitations of in-room immobilization systems and their performance throughout the course of treatment. Accurate characterization of immobilization system performance can provide useful data for physicists and physicians for determining planning margins, which have significant impact on healthy tissues and organs at risk.
Most commercially available head immobilization devices for proton therapy use a relatively thin layer of carbon fiber composites compared to more rigid construction of their x ray therapy counterparts. 1,2 Moreover, to maintain sharp lateral penumbra, air gap between the aperture and patient needs to be minimal. Therefore, the use of bulky immobilization devices is not desired because they potentially increase likelihood of collision scenarios. It is typically challenging to satisfy two conflicting goals of maintaining rigidity and minimizing material in the beam path. A careful balance of these requirements is attained in the design of immobilizing frames such as the widely used base-of-skull (BoS) carbon-fiber frame (kVue TM BoS insert) by Qfix (Avondale, PA, USA) which is evaluated in this study. Despite the widespread use of BoS frame in proton centers, there are few reports evaluating its effectiveness for immobilization with 6 DOF robotic couches. 3 The purpose of this study is to analyze daily repositioning accuracy of patients treated for intracranial or base of skull tumors in a compact proton therapy system. For this reason, we evaluate interand intra-fractional uncertainties associated with patient movement throughout the treatment. Furthermore, we investigate the effect of utilizing a bite block on inter-and intra-fractional uncertainties. To our knowledge, this is the first study of this kind on a compact proton system setting such as the Mevion S250 system.

| MATERIALS AND METHODS
Our clinic is equipped with a Mevion S250 compact proton therapy system (Mevion Medical Systems, Littleton, MA, USA) accelerator with 190°rotating gantry and a 270-degree robotic couch (KUKA Roboter, GmbH, Augsburg, Germany) with 6 DOF capable of executing submillimeter translational motions and rotational motions.
Robotic couches with 6 DOF are known for their motion accuracy and precision 4,5 which is why they are primarily used for proton therapy. Hsi et al. 6 reported residual target displacement for KUKA robotic couch rotations with respect to isocenter to be within 0.5 mm and nearly 85% of all couch movements were within 0.5 mm in the horizontal plane and within 0.7 mm vector distance from required displacements. The robotic couch in this study is capable of maintaining the position of the isocenter within submillimeter accuracy throughout the treatment and this accuracy is routinely verified on a monthly basis during Monthly QA.
Patient immobilization was achieved using a BoS frame insert with a new type of aquaplast mask and a cushion by Klarity (Newark, OH, USA) as shown in Fig. 1. This new mask is known for its nonstick properties and relatively slow hardening process, which make it easier to prepare and conform to patient's anatomy. The cushion was used to support posterior skull in a comfortable and reproducible position. For certain patients, and per physician directive, a mask combined with a Klarity bite block with malleable thermoplastic material was used with goal of decreasing patient motion within the mask. Subsequently, the bite block was attached to the mask and the mask was indexed to the BoS frame.
Once the patient was simulated, treatment was planned in Pinna- It is worth mentioning that motion accuracy for our robotic couch is consistent as long as isocenter is located inside the so-called "Couch Treatable Volume", which is a space that includes the treatment end of couch-top and enough space above it in which a patient isocenter can be located. The treatable volume covers a rectangular solid volume which moves with couch-top and has dimensions of 50 cm wide, 39 cm tall, and 95 cm long. Couch corrections that will put isocenter outside the Treatable Volume will not be executed in Verity TM software.
In order to assess various components of uncertainties in the immobilization and localization system, we employ the approach adopted from van Herk's population based statistics. 7 Below, the definition of each error category is presented: • M g or Mean group error: where d i is the average inter-fractional displacement for a single patient i throughout the course of treatment, and N denotes the number of patients in the group under study.
• r g or systematic error: where SDðd i Þ is the standard deviation of the average inter-fractional displacement measurements for each patient.
• r p or inter-fraction random error: where SDðd i Þ is the standard deviation of the average inter-fractional displacement measurements for each patients.
• r f or intra-fraction uncertainty : r f is calculated using the same equation as r p , utilizing post-treatment images.
Once the above uncertainties are assessed, appropriate margins are applied to ensure robust delivery of prescribed dose to target volume.
In this study, we analyze inter-fractional translational displacements in addition to rotational alignments on 33 patients including  Fig. 3 and Table 1. Kolmogorov-Smirnov goodness-offit test reveals that the above distributions are of Gaussian nature.
To illustrate this clearly, we overlay a Gaussian fit for every histogram.

3.B | Intra-fractional translational and rotational displacements
Histograms of intra-fractional translational and rotational displacements and statistical values associated with them are presented in Fig. 4 and Table 1, respectively. We find the magnitude of both translational and rotational displacements to be smaller and more symmetric for intra-fractional motion compared to inter-fractional displacements and all rotations show similar distribution profiles centered and symmetrical around zero.

3.C | Evaluation of the inter-and intra-fractional uncertainties
A summary of data analysis for all patients for inter-(n = 644) and intra-(n = 84) fractions is presented in Table 2. For all patients included in this study the highest value for M g error was 2.3 mm along PA direction. M g error for inter-fractional motion along LAT direction was 1.6 mm. We found the systematic errors (r g ) along PA and SI direction to be higher (2.8 and 2.6 mm respectively) compared to 1.4 mm along LAT. Random errors (r p ) along these two axes follow the same pattern (1.0 and 0.4 mm, respectively). We found the intra-fractional uncertainty along PA axis (0.4 mm) to be the highest compared to LAT and SI direction (both equal to 0.1 mm).

3.D | Effectiveness of bite block in conjunction with Klarity mask
In this section, we divide the patients into two groups based on whether they had a bite block or not. Table 3 Table 3 indicating that the use of bite block does not improve overall immobilization. Translational and rotational r p and r f for every degree of freedom does not change for cases with a bite block compared to cases without one.

3.E | Phantom measurements
Comparison of the 100 and 200 lbs scans revealed that the couchtop in the simulation room sags due to increased relative weight on couch-top. Once couch coordinates were calculated, the alignment images were analyzed. The mean magnitude of the couch shifts along PA direction was found to be 3.3 mm different between both weights with larger PA shift corresponding to the heavier weight. weighed less than 150 lbs, which is in agreement with the direction of patient shifts that has been observed in this study as shown in

| DISCUSSIONS
The data presented in this study reflect the combined performance of the immobilization system components (mask, BoS frame, cushion, and bite block) and the motion of the robotic couch. This is especially relevant to compact systems such as the one used in this study where limited gantry motion is compensated for using more robotic couch angles compared to a 360-degree gantry room, where coplanar beam delivery capability is readily available without the need for 180-degree couch rotation.

4.A | Inter-and intra-fractional translational and rotational corrections
The  imaging solutions for daily localization. 9 The translational systematic errors (r g ) are comparable to those reported for non-robotic couches 9 and conventional linear accelerators such as the Elekta Synergy MLCi accelerator with carbon fiber tables. 10 In addition, the random errors are significantly smaller than those reported for both above systems. 9,10 In the case of rotational uncertainties, the interfractional systematic and random errors are remarkably smaller than those reported for other particle therapy systems. 10 The intra-fractional translational and rotational uncertainties are considerably smaller than those associated with inter-fractional motion, and are an order of magnitude smaller than those previously reported for a full gantry proton system. 11 We find intra-fractional uncertainty along PA direction to be higher than other directions.

4.B | Effectiveness of bite block in conjunction with Klarity mask
Our results show that the addition of a bite block does not enhance immobilization effectiveness. Translational M g and r g or setup errors associated with cases without bite block show smaller (better) values. On the other hand, r f for cases with bite block is similar to cases without one. The trends are similar for rotational uncertainties.
Lack of enhancement along PA direction can be because the bite block is not directly attached to the couch as shown in Fig. 1. One might conclude that using bite blocks complicate the setup procedure and increases patient discomfort, which in turn may lead to an increase in inter-fraction uncertainty. 12 It is worth mentioning that while the bite block used in this study is relatively malleable, it is not considered a custom bite block that would mold specifically to patient anatomy.

4.C | Phantom measurements
The mean of magnitude of the couch shifts along PA direction was found to be 3.3 mm from nominal position, which suggests that the robotic couch is compensating for the declining of the treatment room couch correctly. The same argument is valid about rotational corrections like the pitch. Mean of the pitch is different by 0.15°f rom nominal, which is caused by couch-top sag. As for lateral shifts, we found that the distribution of lateral shift reflects a systematic difference between BoS frame position in CT room and treatment room.

| CONCLUSION S
Our study presents a comprehensive robustness evaluation of a widely used cranial immobilization systems in proton therapy and specifically in a compact system setting. We show that the BoS frame localization technique in conjunction with thermoplastic mask yields submillimeter and subdegree intra-factional uncertainty for all translational and rotational degrees of freedom, respectively. This level of precision and reproducibility fares better or comparable to most published studies on this topic to date. In addition, we show that the use of a mask-attached type bite block has marginal impact on inter-or intra-fraction uncertainties compared to no bite block.