Assessment of a 2D electronic portal imaging devices-based dosimetry algorithm for pretreatment and in-vivo midplane dose verification.

BACKGROUND: The use of electronic portal imaging devices (EPIDs) is a method for the dosimetric verification of radiotherapy plans, both pretreatment and in vivo. The aim of this study is to test a 2D EPID-based dosimetry algorithm for dose verification of some plans inside a homogenous and anthropomorphic phantom and in vivo as well.
MATERIALS AND METHODS: Dose distributions were reconstructed from EPID images using a 2D EPID dosimetry algorithm inside a homogenous slab phantom for a simple 10 × 10 cm2 box technique, 3D conformal (prostate, head-and-neck, and lung), and intensity-modulated radiation therapy (IMRT) prostate plans inside an anthropomorphic (Alderson) phantom and in the patients (one fraction in vivo) for 3D conformal plans (prostate, head-and-neck and lung).
RESULTS: The planned and EPID dose difference at the isocenter, on an average, was 1.7% for pretreatment verification and less than 3% for all in vivo plans, except for head-and-neck, which was 3.6%. The mean γ values for a seven-field prostate IMRT plan delivered to the Alderson phantom varied from 0.28 to 0.65. For 3D conformal plans applied for the Alderson phantom, all γ1% values were within the tolerance level for all plans and in both anteroposterior and posteroanterior (AP-PA) beams.
CONCLUSION: The 2D EPID-based dosimetry algorithm provides an accurate method to verify the dose of a simple 10 × 10 cm2 field, in two dimensions, inside a homogenous slab phantom and an IMRT prostate plan, as well as in 3D conformal plans (prostate, head-and-neck, and lung plans) applied using an anthropomorphic phantom and in vivo. However, further investigation to improve the 2D EPID dosimetry algorithm for a head-and-neck case, is necessary.

INTRODUCTION

As the complexity of the radiotherapy process increases, the need for verification of accuracy of the dose delivered becomes more important. Image-guided radiotherapy (IGRT) tools have been made available by vendors for geometric verification of patient positioning. In addition, verification of the dose, either pretreatment or in vivo, plays an important part for a quality assurance program for advanced irradiation techniques. The importance of pretreatment or in vivo verification of the dose delivered to the patient has also been increased; due to the increasing complexity of the new radiotherapy techniques.[1234] Pretreatment verification is a commonly used substitute method, whereby, the planned dose is verified before treatment. In pretreatment, various treatment parameters, including the beam energy, number of monitor units, and multileaf collimator (MLC) settings, have to be verified, to ensure an accurate dose delivered to a patient. The above parameters can be used to calculate dose distribution within a homogenous phantom. The drawback, however, is, those errors occurring at the time of treatment cannot be detected and it is not clear how errors detected in pretreatment would translate to errors within a patient. Ideal dose delivery at radiotherapy, especially in vivo, is an alternative that ensures that the dose distribution delivered to the patient at the time of treatment corresponds to the planned treatment.[5]

Portal images acquired by using EPIDs have been addressed as a key element for the above-mentioned verifications.[467891011121314151617] EPIDs are convenient tools for both pretreatment verification and in vivo dosimetry.[51418] An EPID image is converted to a pretreatment or in vivo dose distribution (with a phantom or patient) using an appropriate approach verification of the dose calculation and plan deliverability.

A dose verification method based on an EPID and using a back-projection approach has been developed at the Netherlands Cancer Institute-Antoni van Leeuwenhoek (NKI-AVL). Two-dimensional dose distributions inside a phantom or patient are reconstructed from portal images using the back-projection algorithm. The model requires the primary dose component at the position of the EPID. A parameterized description of the lateral scatter within the imager is obtained from measurements with an ionization chamber in a mini phantom. It is currently used for all IMRT treatments, preferably for in vivo or otherwise for pretreatment verification. The midline dose can be determined by converting the measured entrance (or exit) dose with the depth dose data, taking the inhomogeneities into account.[19202122] However, the drawbacks of these methods are that additional patient information is required to determine the midline dose, and the measurements and calculations are not completely independent of the treatment planning process.[23] The back-projection approach is a method to derive the midplane dose from the transmission dose data measured with an EPID, without additional patient information, that is, independent of a treatment planning system (TPS).[23]

Several articles have reported pretreatment and in vivo EPID-based dosimetry for IMRT plans based on the determined midplane dose. Boellard et al.[23] developed a new method to derive the midplane dose from the transmission dose data, measured with a scanning liquid ionization chamber (SLIC) EPID, which is not the commercialized and tested transmission dose method to obtain the midline dose in homogenous and symmetrical inhomogeneous phantoms. Wendling et al.[5] modified a previously developed EPID back-projection algorithm, but just for prostate IMRT treatments and applied it to an amorphous silicon EPID. McDermott et al.[15] investigated the feasibility of replacing pretreatment verification with in vivo EPID dosimetry for only prostate IMRT plans. To our knowledge, to date, an amorphous silicon (a-Si) EPID has not been employed to determine the midplane dose using an Anthropomorphic (Alderson) phantom with inhomogenities.

In this study, we assay a 2D EPID-based dosimetry algorithm to compare the reconstructed dose distributions with the corresponding TPS calculated dose distributions for a simple 10 × 10 cm2 box technique, IMRT prostate, and 3D conformal plans (head-and-neck, prostate, and lung), using a homogenous slab and anthropomorphic (Alderson) phantoms, and also inside a patient (in vivo) for three conformal plans (prostate, head-and-neck, and lung).

MATERIALS AND METHODS

Accelerator and electronic portal imaging devices

In this study measurements were performed in the Department of Radiation Oncology of the Netherlands cancer institute on various SL20i linear accelerators (Elekta, Crawley, U.K) using 10 MV and 6 MV photon beams for prostate, lung, and head-and-neck plans. All accelerators are equipped with a multileaf collimator (MLC), consisting of 40 leaf pairs with a projected leaf width of 1 cm at the isocenter.[24] A PerkinElmer RID 1680 AL5/Elekta IView GT a-Si EPID was used for all the measurements. The EPID has an intrinsic 1 mm thick copper plate on top of the scintillation layer. An extra 2.5 mm thick copper plate was mounted directly on top of the standard plate, both as an additional buildup material and to absorb the scattered low-energy photons from the phantom or patient.[25] The sensitive area of the EPID is 41 × 41 cm2. Images were acquired using an in-house developed software.[2526]

Electronic portal imaging devices-based dosimetry using a back-projection algorithm

Images were acquired with an amorphous silicon flat panel imager. Details regarding the image acquisition, stability, imager design, and dosimetric characteristics have been reported extensively in literature.[2526] The details of the algorithm used to determine dose images within a patient (or phantom) for the present study has also been described elsewhere,[523] and will be briefly outlined here. The algorithm converts segment images to a 2D distribution of absolute dose in the reconstruction plane of the patient, defined as the plane perpendicular to the beam axis.[52728] Therefore, this plane rotates with the gantry. The pixel values of the transit dose image are processed using scatter kernels (for scatter produced within the EPID and inside the patient reaching the EPID), the scatter-to-primary ratio (for scattered radiation within the patient), the inverse square law factor, and the measured transmission factor, to obtain absolute dose distribution in the isocentric plane of the patient. The measured transmission of the beam through the patient is determined from images acquired for segments of each field both with and without the patient. The location of the reconstruction plane is arbitrary, so a correction is required to account for attenuation of the beam from the isocentric plane to the exit surface. The external contour of the patient's computed tomography (CT) scan is used to obtain the ratio of the geometrical path lengths, which is used to calculate the attenuation per pixel. Reconstructed 2D dose distributions for each field are then the sum of the reconstructed dose distributions of all the segments belonging to that field.

Treatment planning and dose comparison

Dose distributions were calculated with the current clinical version of the TPS (Pinnacle V9.0, Philips Medical Systems, Eindhoven, the Netherlands). All dose distributions were calculated using Pinnacle's adaptive convolution/superposition approach. This was a fast implementation of the collapsed cone convolution/superposition algorithm and included an advanced tissue inhomogeneity correction.[29] Dose distributions reconstructed from EPID images were compared with the planned dose distributions using 2D γ evaluation within the 20% isodose line for the conformal plans and the IMRT cases. The 20% isodose surface was chosen to improve the γ-statistics and masked regions with discrepancies.[15] The following criteria were used for gamma statistics — a dose difference tolerance of 3% and a distance-to-agreement tolerance of 3 mm. An EPID dosimetry report was created for each treatment plan, providing for each beam the mean γ,”the near maximum γ value” γ1%, and the percentage of points with γ ≤ 1 (Pγ≤1 or pass rate), as well as the dose at the isocenter. A green dot in the EPID dosimetry report meant that all alert criteria were within tolerance level (γmean ≤ 0.5, γ1%≤ 2, Pγ≤1≥ 85%and ΔDisoc ≤ 3%). The yellow dot in the report indicated that at least one of the alert criteria was outside the tolerance level, but still within the action level (0.5 ≤ γmean < 1.0, 2 ≤ γ1%<4, 70% ≤ Pγ≤ 1 < 85% and 3% ≤ ΔDisoc < 5%). The red dot indicated that at least for one of the evaluation criteria the error was outside the action level (γmean > 1, γ1%> 4, Pγ≤ 1 < 70% and ΔDisoc > 5%).

Dose verification in the phantoms and the patients

In order to assay the 2D EPID-based dosimetry algorithm, measurements were performed using a homogenous polystyrene phantom (slab phantom of 30 × 30 cm2 area and 20 cm height) and an inhomogeneous anthropomorphic (Alderson) phantom — the former being a more realistic approximation of clinical situations. The Alderson phantom was positioned in a supine position and the three conformal plans (prostate, head-and-neck, and lung) and one seven-field IMRT prostate plan were randomly selected. The plans were recalculated with the TPS, replacing the patient planning CT scan with a phantom CT scan using the same plan parameters — the MLC settings, gantry, and collimator angles, numbers of monitor units and energy. In the case of homogenous slab phantom measurements, the phantom was irradiated with a simple box technique (four-field, 10 × 10 cm2, 10 MV). The source surface distance (SSD) was 90 cm at a gantry angle of 0° and the isocenter was positioned at the middle of the phantom. For all measurements the source-to-imager distance (SID) was 150 cm.

In vivo EPID dose verification was performed for the same fractions as setup verification for prostate, head-and-neck, and lung plans. Additional EPID dosimetry images were acquired at the third fraction to ensure five in vivo measurements within the first two weeks plus one day of treatment. Therefore, EPID treatment images were acquired at the first three fractions and then once per week, as also for any fractions requiring an additional setup verification. The first measured fraction for each patient was included in this study. Plans were first compared at the isocenter, by summing the planned and EPID isocenter doses for each field, respectively. Each field was then compared in 2D at the plane corresponding to the location of the reconstructed dose distribution determined with the EPID.

RESULTS

Phantom measurements

The results of the γ-evaluation for a simple box configuration delivered to a slab phantom analyzed with the 2D midplane EPID dosimetry approach are shown in Figure 1. Within the area of the 20% isodose lines, the percentage of points with γ ≤ 1 for all fields is within the tolerance level. The average pass rate is 96.3 ± 3.1%, which also shows good agreement.

Figure 1

γ-evaluation of a four-field box plan using 2D EPID reconstruction through a homogenous slab phantom. The percentage of γ ≤ 1 within the 20% isodose line for each field is given at the bottom of figures

In Table 1 the average results (1SD) of the γ-evaluation parameters (mean γ and γ1%) and dose difference at isocenter for the simple box field with the slab phantom are presented. For the simple fields, all γ-evaluation parameters and dose difference at the isocenter are within the tolerance level. The mean γ and γ1% values averaged over the four simple fields are 0.4±0.11 (1SD) and 1.0 ± 0.1, respectively. All γ distributions had mean γ and γ1% values below 0.48 and 1.1, respectively. The AP beam (gantry angle of 0°) had the highest γ values, including mean γ, γ1%, and dose difference at the isocenter with respect to the PA beam. The measured dose at the isocenter was 1.8% below the planned dose, which was within the tolerance level for the mentioned criterion. The individual measured fields were compared and had very good agreement with the planned dose distributions achieved.

Table 1

The results of γ-evaluation parameters and the measured/calculated dose difference at the isocenter for the box field in a homogenous slab phantom

In Figure 2 a set of planned X and Y dose profiles is compared with the measured ones acquired from the reconstructed EPID midplane dose image of a slab phantom, which demonstrate very good agreement, especially in the penumbra region. The EPID reconstructed dose on the central axis is 1.8% below the planned dose.

Figure 2

Absolute dose profiles of a square field (G = 0°) reconstructed by EPID in the midplane (blue curve) and calculated by the TPS (red curve)

In Figure 3 the mean γ, γ1%, and the measured and planned dose values of the three delivered clinical conformal plans (head-and-neck, lung, and prostate) in an Alderson phantom are presented, using the AP–PA beams. For the lung plan, the mean γ on anterior–posterior (AP) beam is 12.2%, larger than the tolerance level and in the posterior–anterior (PA) beam. The head-and-neck plan has a larger mean γ value (10.7%) than the tolerance level, while all γ1% values are within the tolerance level for all plans in both AP and PA beams. The lung plan has the highest mean γ for the AP beam, while on the PA beam the head-and-neck has the highest mean. Figure 3b indicates that the planned dose for the AP beam is higher than the EPID reconstructed dose for all plans, while for the PA beam it is the opposite.

Figure 3

The results of (a) γ mean and γ1% and (b) planned and EPID doses for both beam directions, of the prostate, head-and-neck, and lung plans with the Alderson phantom. The dashed lines indicate the tolerance level for Figure 3a

Figure 4 represents the percentage of points with γ ≤ 1, within the area of the 20% isodose lines for the three conformal plans in the Alderson phantom. Figure 4a shows the γ1% of both the AP–PA beams for the conformal plans and Figure 4b represents the γ1% values of a seven-field prostate IMRT plan, field by field: The percentage of points with γ ≤1 are 5.3, 10.2, 7.5, and 7.0% larger than the tolerance level, respectively.

Figure 4

γ-evaluation of (a) the head-and-neck, lung, and prostate plans for both AP–PA beams and (b) a seven-field prostate IMRT plan. The percentage of γ ≤ 1 within the 20% isodose line for each field is given in brackets

In Figure 5 the mean γ and γ1% values of the seven-field prostate IMRT delivered clinical plan in the Alderson phantom is presented. The mean γ values varied between 0.28 and 0.65. Figure 5 also shows that the γ1% for all fields is less than 2 (the tolerance level). The gantry angle of 150° had the highest γ failure rate for both mean γ and γ1%.

Figure 5

γ mean and γ1% results of a seven-field prostate IMRT plan. The dashed lines indicate the tolerance level

Patient measurements

The example presented in Figure 6 is one fraction of an in vivo verification EPID dosimetry reported for three-dimensional conformal plans, including prostate, lung, and head-and-neck cancer treatments. The dose distributions reconstructed by the 2D EPID dosimetry algorithm were compared with the planned dose distributions. The difference between the measured and planned doses at the isocenter was less than 3% for all plans except for the head-and-neck plan, which was 16.7%. The mean γ value for all plans was less than 0.5 except for one field (G = 120°), which was about 18% higher than the tolerance level and a warning was raised. On an average, over all fields, the mean γ value was found to be 0.41 for all plans. Within the area of the 20% isodose lines, the near maximum γ value averaged over all fields was 1.20, 1.27, and 1.56 for the prostate, lung, and head-and-neck plans, respectively.

Figure 6

In vivo EPID dosimetry reports of the (a) prostate, (b) lung, and (c) head-and-neck plans. The symbols above the report indicate the gantry angle (G), collimator angle (C), table position (T), and beam energy (E)

DISCUSSION

For the inhomogenous phantom dose verification, the dose distributions resulting from the EPID measurements and the TPS agreed within the γ criteria of 3% and 3 mm for at least 93.3% of the points within the 20% isodose surface. We assumed that the linear accelerator (Linac) settings, such as, the number of monitor units and MLC parameters, were accurately transferred from the TPS to the treatment machine. Moreover, it proved that for a homogenous object, the 2D dose-reconstruction algorithm based on EPID transmission images was accurate within the γ criteria of 3% and 3 mm. Note that the dose verification for a 10 × 10 cm2 field was completely satisfied based on the 3%/3 mm criteria [Table 1]. A similar observation was made by Wendling et al.[5] On account of the excellent agreement between reconstructed and calculated dose values in the homogenous phantom, the 2D EPID dosimetry algorithm could potentially be extended for situations where inhomogeneities, such as air cavities, were present, for example, in an anthropomorphic phantom or patient (in vivo EPID-based dosimetry). To illustrate that the 2D in vivo EPID dosimetry algorithm in the presence of inhomogeneties was more useful than a pretreatment verification using a phantom or a fluence check, in-Alderson phantom measurements were performed. In vivo EPID dosimetry did not test the inhomogenity correction of the TPS; the results showed that for the dose verification by Alderson phantom, the dose distributions resulting from the EPID measurements and the TPS agreed within the γ criteria of 3% and 3 mm. Moreover, most of the mean γ values were less than 0.5, except for the lung plan in the AP beam, the head-and-neck plan in the PA beam, and the IMRT prostate plan, only for the gantry angle of 210°. In case of the IMRT prostate plan, this discrepancy for a seven-field was ignorable and for the 3D conformal plans (the lung plan in the AP beam and the head-and-neck plan in the PA beam) it might be due to a larger sensitivity for setup variations and patient anatomy changes of the lung and head-and-neck plans, as compared to the prostate plan. Another reason might be that the use of a calculated transmission introduced an additional uncertainty.[30]

By looking in the EPID-based in vivo dosimetry reports [Figure 6], it can be seen that for the in vivo dose verifications, the EPID-based dose values differ only to a small extent from the planned dose values, for the head-and-neck case. There are several possible explanations why this deviation has occurred. First, the uncertainty of the reconstructed dose, in a low-dose, low-gradient region, using the back-projection algorithm is1.6% (1SD), which has been determined by repeating a measurement for a head-and-neck case for over a nine-month period, at different accelerators. Furthermore, some of the segments are very irregular in shape in the high-dose regions. The other data presented in Figure 6 illustrate that the 2D EPID-based dosimetry algorithm is accurate in dose verification for at least the prostate, head-and-neck, and lung plans, however, further investigation to improve the 2D EPID dosimetry algorithm for the head-and-neck case, is necessary. The accuracy of the 2D-EPID based in vivo dosimetry algorithm is in agreement with the observations reported by McDermott et al.[15] Moreover, the advantages of in vivo dosimetry are that a check and record of the actual treatment is obtained in the measurement time, without additional cost. This is only an advantage if the method can be shown to be as accurate, and provide as much information as necessary, during pretreatment dosimetry. The number of measured in vivo fractions required to replace pretreatment verification is a balance between early detection and workload. Another big advantage of in vivo EPID dosimetry is that the workload of treatments is considerably reduced compared to the pretreatment verification.[15] The only aspect that is not verified in this approach is the accuracy of the TPS tissue inhomogenity correction (including the correct selection of the CT density table). We assume that the TPS tissue inhomogenity correction is checked during the commissioning process of the TPS and that for the correct selection of the CT density tables, other QA procedures are required. Other groups have used advanced dose calculation algorithms such as the Monte Carlo methods, for EPID-based dose verification.[31] These approaches also give accurate results within the inhomogeneous tissue regions such as the lung. However, the use of these methods is counterbalanced by their complexity and the considerably longer time needed for dose calculation.

CONCLUSION

We have shown that a 2D EPID-based dosimetry algorithm provides an accurate method to verify the dose of simple 10 × 10 cm2 fields inside a homogenous slab phantom, as well as the IMRT prostate plan and 3D conformal plans (prostate, head-and-neck, and lung plans) inside the anthropomorphic phantom. The model for 2D in vivo dose reconstruction from EPID images for 3D conformal treatments of a prostate, a head-and-neck, and a lung cancer patient is accurate within the γ criteria of 3% and 3 mm inside the patient, when the external patient contour is known. The algorithm is fast and is independent of the TPS. It provides a safety net for patient treatment.