Development and validation of a comprehensive patient‐specific quality assurance program for a novel stereotactic radiation delivery system for breast lesions

Abstract Purpose The GammaPod is a dedicated prone breast stereotactic radiosurgery (SRS) machine composed of 25 cobalt‐60 sources which rotate around the breast to create highly conformal dose distributions for boosts, partial‐breast irradiation, or neo‐adjuvant SRS. We describe the development and validation of a patient‐specific quality assurance (PSQA) system for the GammaPod. Methods We present two PSQA methods: measurement based and calculation based PSQA. The measurements are performed with a combination of absolute and relative dose measurements. Absolute dosimetry is performed in a single point using a 0.053‐cc pinpoint ionization chamber in the center of a polymethylmethacrylate (PMMA) breast phantom and a water‐filled breast cup. Relative dose distributions are verified with EBT3 film in the PMMA phantom. The calculation‐based method verifies point doses with a novel semi‐empirical independent‐calculation software. Results The average (± standard deviation) breast and target sizes were 1263 ± 335.3 cc and 66.9 ± 29.9 cc, respectively. All ion chamber measurements performed in water and the PMMA phantom agreed with the treatment planning system (TPS) within 2.7%, with average (max) difference of –1.3% (−1.9%) and −1.3% (−2.7%), respectively. Relative dose distributions measured by film showed an average gamma pass rate of 97.0 ± 3.2 when using a 3%/1 mm criteria. The lowest gamma analysis pass rate was 90.0%. The independent calculation software had average agreements (max) with the patient and QA plan calculation of 0.2% (2.2%) and −0.1% (2.0%), respectively. Conclusion We successfully implemented the first GammaPod PSQA program. These results show that the GammaPod can be used to calculate and deliver the predicted dose precisely and accurately. For routine PSQA performed prior to treatments, the independent calculation is recommended as it verifies the accuracy of the planned dose without increasing the risk of losing vacuum due to prolonged waiting times.


| INTRODUCTION
The GammaPod (Xcision Medical Systems, LLC; Columbia, MD) is a novel breast-specific stereotactic radiotherapy device developed at the University of Maryland that has recently received 510(k) clearance from the U.S. Food and Drug Administration (FDA). 1 Operating similarly to the long-established Gamma Knife, the GammaPod's 25 nonoverlapping cobalt-60 ( 60 Co) sources dynamically paint dose to a breast lesion by rotating the beams around a small target while the couch translates continuously in three axes during beam delivery.
From simulation to treatment, the patient's breast is immobilized using mild negative pressure through a device-specific dual-cup system with stereotactic fiducials. The cup, which functions as a stereotactic frame, is secured to both the simulation and treatment tables,  3 . Details of the GammaPod and the GammaPod TPS can be found in a previously published reports. [1][2][3] Whenever a large stereotactic dose is delivered to a small volume, geometric misses and errors in delivery can have a large impact on the resulting quality of treatment. [4][5][6] Indeed, studies have shown the importance of independent pretreatment verification with similar modalities, such as the GammaKnife. 7,8 The objective of this study is to develop and report a comprehensive patient-specific quality assurance (PSQA) program for the GammaPod. To this end, we perform absolute ionization chamber measurements and relative film measurements, which we then compare to predicted TPS dose distributions for the initial 15-patient cohort of our trial.
This study represents the first validation of the ability of the novel GammaPod dedicated breast stereotactic treatment unit to accurately plan and deliver stereotactic radiosurgery (SRS)-grade radiation dose distributions. It also represents the first PSQA performed using the patient geometry with the actual patient's plan.
2 | ME TH ODS 2.A | The GammaPod breast stereotactic radiosurgery (SRS) system The GammaPod design was originally described by Yu et al. 1 It includes three main systems: the collimator and source carrier, the couch, and the breast cup. The accuracy of the GammaPod system has been reported in Becker et al. 3 The GammaPod's collimator produces and beam with a FWHM that is reproducible to within ≤ 0.2 mm, the couch translation is reproducible within ≤ 0.1 mm, and the registration of the breast cup fiducial system is accurate within ≤ 0.2 mm.

2.A.1 | Collimator and source carrier
The collimator and source carrier include two independent hemispheres that are nested together. The outer hemisphere is the source carrier, which houses the 25 (previously 36 in an earlier design 1,2 ) sources which are aligned in a spiral pattern around the carrier, all focused at an isocenter 380 mm away. Starting at 18°b elow the horizontal plane, the sources spiral down by 1°off the horizon and 60°longitudinally at a time. In contrast, the inner hemisphere plays the role of the collimator. This inner hemisphere holds openings for both 15-and 25-mm diameter field sizes which can align with each of the 60 Co sources, selecting among them through a 20°rotation between the "blocked" and 15-and 25-mm positions.
To deliver treatment, the entire assembly rotates as a single unit to spread the dose from all the sources around the breast. This geometry is illustrated in Fig. 1, a diagram of the source carrier and collimator system. Fig. 2 is a top-down view of the source configuration for both the 36-and 25-source models.

2.A.2 | Patient couch
The second component of the GammaPod system is the patient couch. This couch system holds and immobilizes the patient and locks her into place via the breast cup. The table lowers the patient into position so that the radiation isocenter is within the breast & PTV. The couch then translates the patient along the planned path so that the target receives the desired dose distribution. This is quite similar conceptually to the use of multiple isocenters in the Gamma-Knife, with the exception that the GammaPod table moves continuously between tens to hundreds of isocenters during treatment, while GammaKnife delivers radiation only in discrete isocenters. This dynamic delivery method enables a much more uniform dose distribution inside the target compared to the "sphere packing" technique characteristic of the Gamma Knife. 9,10

2.A.3 | Breast cup system
The third and final component of the GammaPod is the breast cup system, which serves as an immobilization device and provides the stereotactic frame. It consists of an outer cup, an inner cup, and a flange (Fig. 3). Outer cups are in three diameters (small, medium, and large) and are made of a hard and rigid polycarbonate. The small, medium, and large sizes correspond to diameters at the base of 93.7, 121.7, and 153.7 mm, respectively. This outer cup locks to the table to keep the patient immobilized. Embedded in the cup is a fiducial wire (recognized by the TPS) that serves to determine the laterality of the treatment and the stereotactic coordinate reference frame. The outer cup is hermetically sealed, except for a tube that connects to the vacuum pump. The inner cup, constructed with a thin layer of polyethylene, is available in the same three diameters as the outer cups and in 10 sizes for each diameter, based on chestto-apex distances (Fig. 4). This inner cup is selected based on the patients' breast size so that the entire cup is filled, except for a small air gap at the apex of the breast, ensuring a predictable treatment geometry for planning purposes. A seal between the inner cup and the patient's skin is achieved using a silicone flange fitted to the inner cup. The flange is then attached to the outer cup. When the vacuum is applied between the outer and inner cups, aeration holes close to the apex of the breast produce suction that immobilizes the breast and the chest wall for simulation and treatment delivery.

2.B | Original clinical trial
This study was conducting using the planning and measurement information of 15 patients who underwent GammaPod treatments under the IRB approved protocol to collect safety and feasibility data in preparation for submission for FDA 510(k) clearance. 11 These 15 original patients were treated with a single-fraction 8-Gy boost to the lumpectomy cavity + a 1-cm margin. 12 After GammaPod treatment, patients proceeded to receive whole-breast radiotherapy using a hypofractionated course of therapy consisting of 40.05 Gy in 15 fractions or 42.56 Gy in 16 fractions. Table 1 displays the target and plan details for each patient.

2.C | Patient-specific QA program
We validated the ability of the GammaPod SRS device to accurately and precisely deliver conformal high-dose distributions. Three components were employed: absolute ionization chamber measurements, relative film measurements, and an independent calculation program.

2.C.1 | Absolute measurements in water
We measured the absolute dose using the TG-21 formalism under two geometries: a water cup matching the patient, and a poly-  Standard Imaging, WI). The chamber was chosen since it was made specifically for small field sizes that are utilized by the machine.
The full width half-maximum for the smallest collimator is 22.0 mm.
The first measurement geometry (e.g., the water cup) was chosen to utilize the specific cup used to immobilize and treat the patient.
To prevent spills, ventilation holes located at the tip of the cup are taped closed (typically under suction to immobilize the breast), and the inner cup is filled with water. We designed and fabricated a cus-    Only the table (x, y, z) positions are adjusted to move the centroid of the GTV to the ion chamber location.

2.C.3 | Relative dose distributions measured by
Gafchromic film The PMMA phantom described in the previous section is also capable of accommodating film. This is accomplished by removing the ion chamber insert and replacing it with the film holder. Using EBT3 Gafchromic (Ashland Inc., Wayne, NJ) film placed in the XZ plane, we measured relative dose distributions and compared these to those predicted by the TPS (Fig. 7). These measurements utilize the same QA plan created in the subsection above for the absolute point dose measurements in the same phantom.
The film was previously calibrated using a 6-MV X-ray beam with 12 dose points between 0.01 and 10 Gy. All films were cut and scanned using standard Gafchromic film procedures (i.e., using the red channel only). 13 The film holder has four pins to mark the film for registration. When a QA plan is generated from a patient's plan, four marks are burned to dose planes to allow for image registration with the film. Fig. 7 shows the comparison between a film and calculated dose profile. The film dose was normalized to a plateau region in the target and analyzed with 3%/1 mm gamma criteria. This criteria is the standard for SRS PSQA. 13

2.C.4 | Independent point dose calculation software
Our third PSQA component is a semi-empirical independent point dose calculation (SEIPDC) to verify the integrity of the TPS calculation. The calculation revolves around using a kernel approach, utilizing predefined isocenter dose rates and off-center ratios (OCRs). The isocenter dose rates are the dose rates at isocenter for each possible position in the breast. If the isocenter is positioned near the chestwall of a large breast, the dose rate will be less than when the  the measurements. It is also a dimensionless factor. Then dose to the reference point can be acquired as following; V f is a volume dependency correction factor which is used to correct for the distortion of the dose distribution when it is off from the center of the cup.

| RESULTS
We developed PSQA tests to verify planned dose distributions in the initial 15-patient study. These patients presented with a range of breast (and thus, cup) and tumor sizes, averaging 1263 ± 335.3 cc and 66.9 ± 29.9 cc, respectively.
Ion chamber measurements in water and in PMMA had an average agreement to those of the TPS of −1.3 ± 1.0% and −1.3 ± 0.5%, respectively. The largest outliers for each were −1.9% and −2.7% respectively. The profile measurements had an average gamma pass F I G . 6. PMMA breast phantom with ion chamber installed (left) and film insert (right). The PMMA phantom can accommodate either insert with no effect on the measured dose distribution. Of particular note, the point dose measurements, both in the PMMA phantom and water-filled breast cup, were consistently below TPS calculations. We believe the cause is related to small uncertainties in the collimator openings which lead to a narrowing of the profile width in a process analogous to the dynamic leaf gap (DLG) commonly used in intensity-modulated radiation therapy. In the DLG, a small error in field size (i.e., in the gap between two multi-leaf collimator leaves) has a negligible effect on the output factor, but can have a disproportionate effect on the total absolute dose delivered by an IMRT plan with a sliding-window delivery. This is due to the fact that the gap is scanned across the target in many occasions, and that dose to a point is related to the time it is "seen" by the MLC gap, which increases when the DLG is larger. Likewise for the GammaPod, the planned dose distribution is delivered through a series of isocenters, and radiation delivery continues intransit while the patient is moved between isocenters. Therefore, the absolute dose is related to the total area under the curve of a radiation profile. A simplistic approximation shows that, for a 15 mm collimator, an error of 0.2 mm (which would be very challenging to measure experimentally) will cause a difference of 1% in area under the curve, and the absolute dose difference will depend on the final isocenter placement and the particular parameters of the plan. Fig. 10 shows two theoretical profiles with a 0.2mm width difference and it also displays the sum under the curve which represents total dose collected from scanning across the profile. As one can see the 0.2mm difference is not visible but the 1% difference in sum is. This does not affect the output calibration of the machine since the measurement is performed with a static beam centered on the isocenter.
We are currently investigating various methods by which to empirically detect this possible cause of the discrepancy between planned and delivered dose.
One of the novel aspects of this study is that the use of a water cup represents one of the first uses of the exact patient geometryalong with the specific patient planto measure the dose delivered to the patient. It does not involve using another, non-representative phantom geometry, or even a recalculation of the dose. For this reason, our study is able to detect the real impact on the absolute dose delivered to the patient arising from such uncertainties in a way that calculation or non-representative phantoms cannot.
where, v is the volume of the target. Considering standard deviation (SD) of 1.4% between the phantom measurements and the calculation, more than 2.7 SD or less than 1% of the cases are expected to have larger than 3.8 % error to the measurements, which requires the measurement-based PSQA before treatments. There were no Type I or Type II errors for the 3% passing criterion of the dose difference and 0.13 Type II error for the 2 mm criterion. Eq (2) provides no false pass for 3% or higher passing criterion.
Further development of the SEIPDC software towards 2-D or 3-D gamma comparison will even increase more confidence level.
Given the relatively few moving parts involved in the GammaPod system and its reliance on well-characterized radioactive sources, catastrophic random errors in dose distribution are unlikely. As with Gamma Knife treatments, it is unclear that maintaining a long-term PSQA will be necessary. However, when introducing a new technique in any center, it is vital to establish its feasibility across a wide variety of geometries and situations. Therefore, any new center implementing a GammaPod program could benefit from introducing a PSQA program such as that described in this study. This is further supported by recent trends toward failure modes and effects analysis of QA procedures, as have been outlined in publications based on the TG-100 report. [26][27][28][29][30][31][32] This study was conducted for 15 patients, which were scanned, simulated, planned and treated as part of the device's initial trial sub-

| CONCLUSION
We successfully implemented a PSQA program for the initial 15-patient study used to obtain FDA 501(k) approval for the GammaPod SRS system. Using absolute ionization chamber measurements performed in water and in a PMMA phantom, as well as film relative-dose measurements and an independent dose calculation, we validated not only the ability of the GammaPod to accurately deliver precise highdose distributions but also the ability of the GammaPod TPS to accurately plan and calculate these highly conformal treatment plans.
Considering the accuracy of the SEIPDC software and the clinical reason of avoiding the increased risk of losing vacuum, the calculation based PSQA is recommended. This study represents the first implementation of a PSQA for the GammaPod system. It also represents the first comprehensive validation of the GammaPod TPS, which is shown to be able to accurately calculate delivered dose distributions.