Evaluation of clinically applied treatment beams with respect to bunker shielding parameters for a Cyberknife M6

Abstract Compared to a conventional linear accelerator, the Cyberknife (CK) is a unique system with respect to radiation protection shielding and the variety and number of non‐coplanar beams are two key components regarding this aspect. In this work, a framework to assess the direction distribution and modulation factor (MF) of clinically applied treatment beams of a CyberKnife M6 is developed. Database filtering options allow studying the influence of different parameters such as collimator types, treatment sites or different bunker sizes. A distribution of monitor units (MU) is generated by projecting treatment beams onto the walls, floor and ceiling of the CyberKnife bunker. This distribution is found to be highly heterogeneous and depending, among other parameters, on the bunker size. For our bunker design, 10%–13% of the MUs are delivered to the right and left wall, each. The floor receives more than 64% of the applied MUs, while the wall behind the patient's head is not hit by primary treatment beams. Between 0% and 5% of the total MUs are delivered to the wall at the patient's feet. This number highly depends on the treatment site, e.g., for extracranial patients no beams hit that wall. Collimator choice was found to have minor influence on the distribution of MUs. On the other hand, the MF depends on the collimator type as well as on the treatment site. The MFs (delivered MU/prescribed dose) for all treatments, all MLC treatments, cranial and extracranial treatments are 8.3, 6.4, 7.7, and 9.9 MU/cGy, respectively. The developed framework allows assessing and monitoring important parameters regarding radiation protection of a CK‐M6 using the actually applied treatment beams. Furthermore, it enables evaluating different clinical and constructional situations using the filtering options.


| INTRODUCTION
In clinical practice of radiation oncology, staff members as well as persons of the general public need to be protected from ionizing radiation and dose limits according to locally relevant legal laws have to be fulfilled. This leads to the typical task of a medical physicist to optimize the design of bunkers such that radiation protection issues are managed, while keeping the corresponding costs and resources of the bunker construction as low as possible. This is a challenging task for dedicated delivery systems such as the Cyberknife (CK) system (Accuray Inc, Sunnyvale CA, USA). In general, the dose given to a person can be expressed as follows: where: P i = sum over all exposures i _ D i = dose rate due to exposure i t i = time duration of exposure i Based on eq. (1) one approach to realize practical radiation protection is the reduction of t i . Another possibility is to place attenuating material between the source of radiation and the person, e.g., build a bunker, which reduces the dose rate. This bunker shielding problem can be separated into primary and secondary barriers. 1 For the CK system, the primary beam can point in almost any direction such that almost everywhere a primary barrier is needed for radiation protection purposes. 2 Secondary beams are related to leakage radiation as well as to scattered radiation and with respect to this, the CK is not very much different from standard delivery systems such as linear accelerators (linacs).
It is important for the motivation of this work that, a few years ago, a new CK model (so-called M6 model) has been released, which differs from the previous versions 3 in geometrical and dose delivery aspects. First, the CK-M6 version encompasses a more symmetric arrangement between the delivery robot and the couch such that the beam arrangements are also more symmetric than for previous versions. Moreover, the CK-M6 is equipped with a multileaf collimator (MLC) increasing the flexibility and versatility. 4 It is thus the aim of this work to investigate whether shielding considerations for both, primary and secondary radiation have to be revised when switching from a conventional linac to the CK-M6. Furthermore, this work assesses the impact of the novel MLC on the required radiation shielding of the CK by analyzing clinically applied treatment beams.

2.A | Evaluated situation
For this study, clinical treatment plans delivered by a CK-M6 at our center were analyzed. This robot-based stereotactic radiation therapy system was initially equipped with a fixed field size interchangeable cone collimators (Fix) and an Iris (Iris) collimator system 5 allowing the collimation of the beam shaped into circles or dodecagons, respectively. The twelve available diameters or circumdiameters for these collimators range from 5 to 60 mm, defined at a source-to-axis distance (SAD) of 800 mm. Since 2015, a third collimation device, the MLC is available in our clinic allowing to deliver field sizes up to 100 9 115 mm (again defined at SAD of 800 mm). For the CK-M6, a treatment plan consists of several beams, which originate from a discrete set of robot positions, called nodes, and point toward different positions within the target volume.
For this study, we retrospectively analyzed a database of 364 CK treatments performed at our center. Altogether, these patients received 1115 treatment fractions, which lead to a total number of 166'125 applied beams. An overview of treated tumor sites and used collimator types is provided in Table 1.
The primarily used bunker design in this work is illustrated by

2.B | Framework
The CK data management system stores information about all applied beams into a database. This information can be extracted as a log file in xml format. In a first step, the newly developed framework reads for each patient the prescribed dose and the log files from all delivered patient treatments and creates a file containing the following information for each delivered beam: robot position as well as beam direction, number of delivered monitor units (MUs), collimator system used, applied tracking mode, and field size at SAD 800 mm. Note, that for the MLC the field size is defined as the area that is not covered by the leaves. All those parameters, together with the treatment site, which is mapped from a separate database, are recorded and stored into the treatment list. This treatment list is then purged from all sensitive information and serves as a completely anonymized repository of the clinically applied treatments. In a second step, the CK bunker is defined as a rectangular geometry with freely selectable dimensions. These routines allow an analysis of the mentioned parameters as well as projecting the beams onto the inner surfaces of the considered bunker design by simply ray-tracing the beam on its central beam axis.  The 'top-views' in Fig. 3 show the overview of the MU distribution for all barriers simultaneously. In Fig. 3(a), the MU distribution including all cases is shown. In contrast to the results for the CK-G4  Table 2, the difference between the applied MUs to the left and the right wall is smaller for the cranial than for the extracranial irradiations: 0.3% vs. 1.7%. In order to determine whether the  patient's feet depends highly on the ratio of the bunker length from the origin on (x-direction) and bunker width (y-direction). As the bunker height is kept constant, the number of MUs hitting the ceiling is mainly dominated by the distance between the origin and the wall at the patient's feet.

2.C | Evaluated parameters
The analysis of the applied field sizes, represented by histograms of the field diameters in Fig. 4, shows the continuous field sizes applied with the MLC versus the discrete field sizes for the two other collimation devices. As expected, Fix collimators are mainly used for the smallest available field sizes (5 and 7.5 mm diameter).
Furthermore, there are MLC fields applied with openings that are beyond the largest possible Fix and Iris field size (Fig. 4).
Finally, MFs of the treatment plans for the different collimation devices and treatment sites are compared. While the mean MF for the Fix collimator is the highest, the MLC shows the smallest MF ( Fig. 5). Although the extracranial treatments generally encompass larger field sizes, the mean MF is still higher than for cranial treatments (Fig. 5). Investigating the MF of the different extracranial treatment sites reveals that there are large differences between the mean MF for spine treatments (15.4 MU/cGy), prostate treatments (9.9 MU/cGy), liver treatments (7.2 MU/cGy), and lung treatments (7.5 MU/cGy).  Regarding the secondary radiation, the mean MF for the Fix and Iris collimators (8.6 and 8.3 MU/cGy) are larger than the mean MF for the MLC of 6.4 MU/cGy. This is due to the segmented manner of delivery as well as the larger field sizes, which are offered by the MLC and also used during the treatments (Fig. 5). In addition to the collimator system choice, the treatment site has a major influence on the MF.

| DISCUSSION
So far, all clinically delivered beams at our institution are included in the database. By always incorporating the most recent treatments, it is possible to monitor the radiation protection issues in almost real-time. Furthermore, the whole framework was developed in a two-step approach, in which the creation of the treatment list removes any sensitive information. This allows easily comparing data for different centers in future works.

| CONCLUSIONS
The developed framework allows analyzing and monitoring radiation protection parameters for the present situation as well as filtering for collimators or treatment sites and exploring different bunker sizes.

CONFLI CT OF INTEREST
The authors declare that they have no conflicts of interest.