EPID‐based in vivo dosimetry using Dosimetry Check™: Overview and clinical experience in a 5‐yr study including breast, lung, prostate, and head and neck cancer patients

Abstract Background Independent verification of the dose delivered by complex radiotherapy can be performed by electronic portal imaging device (EPID) dosimetry. This paper presents 5‐yr EPID in vivo dosimetry (IVD) data obtained using the Dosimetry Check (DC) software on a large cohort including breast, lung, prostate, and head and neck (H&N) cancer patients. Material and Methods The difference between in vivo dose measurements obtained by DC and point doses calculated by the Eclipse treatment planning system was obtained on 3795 radiotherapy patients treated with volumetric modulated arc therapy (VMAT) (n = 842) and three‐dimensional conformal radiotherapy (3DCRT) (n = 2953) at 6, 10, and 15 MV. In cases where the dose difference exceeded ±10% further inspection and additional phantom measurements were performed. Results The mean and standard deviation (μ±σ) of the percentage difference in dose obtained by DC and calculated by Eclipse in VMAT was: 0.19±3.89% in brain, 1.54±4.87% in H&N, and 1.23±4.61% in prostate cancer. In 3DCRT, this was 1.79±3.51% in brain, −2.95±5.67% in breast, −1.43±4.38% in bladder, 1.66±4.77% in H&N, 2.60 ± 5.35% in lung and −3.62±4.00% in prostate cancer. A total of 153 plans exceeded the ±10% alert criteria, which included: 88 breast plans accounting for 7.9% of all breast treatments; 28 H&N plans accounting for 4.4% of all H&N treatments; and 12 prostate plans accounting for 3.5% of all prostate treatments. All deviations were found to be as a result of patient‐related anatomical deviations and not from procedural errors. Conclusions This preliminary data shows that EPID‐based IVD with DC may not only be useful in detecting errors but has the potential to be used to establish site‐specific dose action levels. The approach is straightforward and has been implemented as a radiographer‐led service with no disruption to the patient and no impact on treatment time.


| INTRODUCTION
It is recommended that all radiotherapy centers in the United Kingdom have a protocol for accurately measuring the dose delivered to a patient during a course of radiotherapy and comparing this to the planned dose. [1][2][3][4] This approach, commonly known as transit or IVD, has its origins in the 1980s and 1990s when radiographic and radiochromic films were used for this purpose. More recently thermoluminescent detectors (TLDs), semiconductor diodes, metal-oxide field effect transistors (MOSFETs), and optically stimulated luminescence dosimeters (OSLDs) have been used for point dose measurements. [5][6][7] However, there are inherent difficulties with each of these approaches, which have been comprehensively reviewed in several key publications. [8][9][10][11] Many of the limitations of these dosimeters can be overcome by the use of EPIDs, which although developed primarily for imaging, are now widely used as dosimeters and consequently for treatment verification. [11][12][13][14][15] EPIDs, like the aforementioned dosimeters, can be used for either pretreatment verification without the patient present or, as discussed here, for IVD where the patient is present. The main challenge in using EPIDs as dosimeters is in the mapping between the EPID images into dose, with two techniques commonly used for this purpose. In the first a portal dose image is predicted from the treatment plan and the computerized tomography (CT) images used for planning, which is compared to the measured portal dose image. In the second, the measured portal dose image is combined with a back-projection algorithm to calculate the dose in any given CT voxel and hence received by a patient.
The most widespread use of EPIDs as dosimeters has been in pretreatment verification. [16][17][18] However, there are limitations associated with pretreatment verification for detecting certain errors such as those associated with patient anatomy. 19 Furthermore patientspecific pretreatment verification requires additional quality assurance (QA) procedures and linear accelerator time. Using EPID-based IVD for patient-specific QA overcomes these weaknesses and allows verification of the radiotherapy workflow using an independent technique. This was demonstrated in practice in a large-scale study of There is also compelling evidence for treatment adaptation using dosimetric information acquired by this approach. In a recent study of five patients 3D EPID dose was combined with cone-beam CT (CBCT) imaging to detect atelectasis-induced dose changes in lung cancer patients where plan adaptation would be beneficial. 21 There are now commercial systems available for IVD including:  The DC software has two main technical elements. The first is that there is a mapping between the EPID fluence and the monitor units (MU) that would produce the same exposure at the center of a 10 × 10 cm 2 field at the appropriate reference conditions. The output of this mapping is termed the relative monitor unit (RMU). The second is that the scatter within the EPID housing must be taken into account to allow this new unit of RMU to independently calculate the dose received by a patient. This is done by deconvolution of the EPID fluence with the point spread function (PSF) of the EPID, which produces the RMU in terms of in-air fluence.
In practice the PSF has to take into account the dependence of the EPID on the input beam energy and the additional low-energy scatter radiation reaching the EPID from the presence of a patient in NAILON ET AL. | 7 the beam. In DC this is done at commissioning by calculating the PSF for a beam incident on and exiting water at regular intervals from 5 cm up to a maximum of 60 cm between the EPID and the radiation source.

2.A.1 | Relative monitor units
To obtain the absorbed dose (cGy) from the EPID, integrated EPID images are first mapped to RMU, which was defined by Renner as the number of MU that produces the same EPID pixel gray levels as a well-controlled calibration condition. 24 This is usually the 10 × 10 cm 2 reference field that is used to define the output of a linear accelerator, typically as, "1.0 cGy/MU at 10 × 10 cm 2 field size at 100 cm from the surface of water at 1.5 cm depth for 6 MV x rays a ." In the case of open square fields this may be thought of as the collimator scatter factor (S C ) multiplied by the output (MU).
The first step in converting an integrated EPID image into RMU is to establish the relationship between the EPID signal at the central axis of a 10 × 10 cm 2 field and the corresponding MU required to obtain this signal. To account for points not on the central axis, or off-axis points, the in-air off-axis ratio (OAR) along the diagonals of a 40 × 40 cm 2 field are measured to obtain the average OAR.
Multiplying EPID pixels at a distance r cm from the central axis by this value restores the horns on a crossbeam profile, which arise as a result of using a flattening filter.

2.A.2 | In vivo dose evaluation
From this knowledge of the fluence at each pixel of the EPID, which is in RMU, and the beam geometry and patient CT it is possible to ray trace from the x-ray source through the equivalent thickness of water that would produce this fluence. The same principle is applied when the planning CT is used in place of water and ray tracing is used to establish the dose at a point in the CT and hence the patient. The fluence map collected by the EPID image is the source of input for the DC dose calculation engine. This fluence map is used to parameterize the independent pencil beam dose calculation (PBC) algorithm that is used by DC. The dose calculated by DC is next compared to the dose matrix calculated by the treatment planning system. Quantitative evaluation of the difference between the planned and measured dose distribution is carried out in DC using either whole volume or partial volume gamma analysis or by a point dose comparison. 30 The DC software platform has been used for IVD at the Edinburgh Cancer Centre since 2011 to monitor 3795 patients treated with VMAT and 3DCRT at beam energies 6, 10, and 15 MV. This included the cancer sites abdominal (n = 38); brain (n = 256); breast (n = 1215); genitourinary (n = 246); pelvic (n = 318); H&N (n = 636); lung (n = 664); prostate (n = 345); other (n = 77).

2.B | Dosimetry Checkclinical implementation
To configure the DC software for routine clinical use and to measure the dose received by a patient, EPID Sc measurements must be obtained at different beam energies at all available field sizes and at different water equivalent depths. Integrated EPID images of water were acquired on 10 field sizes ranging from 2 × 2 to 28 × 20 cm 2 and 10 water depths in the range from 5 to 60 cm at a focus to imager distance (FID) of 150 cm and 100 MU. This procedure was repeated on the three C-series and two Truebeam linear accelerators used at Edinburgh Cancer Centre. The FID was held constant and the treatment isocenter was always at the center of the water phantom. A fitting program, which is a standard module within DC, was used to fit the measured collimator scatter (S MEAS C Þ to the calculated collimator scatter ðS CALC C Þ using a downhill search optimization algorithm. 25 Convergence was obtained when the variance, defined by, over m data points was within 2%. Figure Table 1 shows the range of treatment sites, the total number of treatment plans verified and the alerts produced by the system. The mean and standard deviation ðl AE rÞ of VMAT cases was found to be T A B L E 1 Details of the range of treatment sites, total number of treatment plans verified and the alerts produced above the 10% threshold from 2011 to 2016. between the planned and delivered (in vivo) dose. Of those cases that exceeded the ±10% tolerance the majority were found to be in patients with breast, prostate, and H&N cancer. The reason for the alerts in the breast group was due to several compounding factors.

3.B | Radiotherapy courses and patients
These were (a) changes in the volume of breast irradiated at each fraction due to the inherent difficulties in positioning of the breast; (b) chest wall irradiation, particularly the impact of rib structures in the field and the resulting uncertainties in the dose; (c) in nodal breast irradiation where the EPID imager position has to be shifted, no off-axis correction is currently applied in the calculation; (d) currently breast patients are treated in free-breathing mode, adopting a breath-hold technique will improve positioning and reduce dosimetric uncertainty.
In the prostate cohort the failures were due to bladder and rectal filling and in the H&N cancer group the alerts were as a result of weight loss and choice of the reference point used for analysis.
In the future the availability of IVD data, such as the data pre- The percentage difference between the measured and calculated Sc was found to be within ±1.0% for field sizes between 5 × 5 cm 2 and 20 × 20 cm 2 . The largest difference was found at the smallest (2 × 2 cm 2 and 3 × 3 m 2 ) and largest (28 × 20 cm 2 ) field sizes investigated. With the substantial increase in the use of small radiotherapy fields, particularly for hypofractionation techniques where high doses per fraction are delivered, it is important to obtain a better match of Sc. 37 The role of checking has been identified as a key element when independently verifying the integrity of a treatment plan. 38  an excessive amount of bowel gas at the time of the treatment planning CT, which was not present at the time of treatment. The patient was rescanned, replanned, treated, and DC IVD used to confirm that the changes made were indeed appropriate. Figure 5 shows the 1D profile through the bladder in the anterior posterior (AP) direction in which it is clear that there is a much better match between the measured and planned dose after the rescan. The use of DC IVD for this case proved particularly helpful.

| CONCLUSIONS
The ability to perform patient-specific QA is now an accepted requirement in modern radiotherapy. This paper presents preliminary data, with a focus on safety, showing that EPID-based IVD with DC has significant potential for this. From knowledge of the expected difference between the in vivo dose and the planned dose, collected on a large number of cases, it may be possible to set site-specific alert criteria for a given treatment site. Furthermore, this approach has the potential to identify suboptimal treatments much earlier than is currently possible.

CONFLI CT OF INTEREST
The authors report a continuing nonfinancial collaboration with Math Resolutions. The authors report no other conflicts of interest. The authors alone are responsible for the content and writing of the paper.

N O T E S
a This is the definition of machine output at the Edinburgh Cancer Centre.