|Year : 2019 | Volume
| Issue : 5 | Page : 1011-1017
Two-dimensional dose reconstruction using scatter correction of portal images
Akbar Anvari1, Parham Alaei2, Mohammad Mohammadi3
1 Department of Radiation Oncology, University of Maryland School of Medicine, Baltimore, MD, USA
2 Department of Radiation Oncology, University of Minnesota, Minneapolis, MN, USA
3 Department of Medical Physics, Royal Adelaide Hospital, Adelaide, SA, Australia
|Date of Web Publication||4-Oct-2019|
Department of Radiation Oncology, University of Maryland School of Medicine, Baltimore, MD
Source of Support: None, Conflict of Interest: None
Context: Electronic portal imaging devices (EPIDs) could potentially be useful for patient setup verification and are also increasingly used for dosimetric verification. The accuracy of EPID for dose verification is dependent on the dose-response characteristics, and without a comprehensive evaluation of dose-response characteristics, EPIDs should not be used clinically.
Aims: A scatter correction method is presented which is based on experimental data of a two-dimensional (2D) ion chamber array. An accurate algorithm for 2D dose reconstruction at midplane using portal images for in vivo dose verification has been developed.
Subjects and Methods: The procedure of scatter correction and dose reconstruction was based on the application of several corrections for beam attenuation, and off-axis factors, measured using a 2D ion chamber array. 2D dose was reconstructed in slab phantom, OCTAVIUS 4D system, and patient, by back projection of transit dose map at EPID-sensitive layer using percentage depth dose data and inverse square. Verification of the developed algorithm was performed by comparing dose values reconstructed in OCTAVIUS 4D system and with that provided by a treatment planning system.
Results: The gamma analysis for dose planes within the OCTAVIUS 4D system showed 98% ±1% passing rate, using a 3%/3 mm pass criteria. Applying the algorithm for dose reconstruction in patient pelvic plans showed gamma passing rate of 96% ±2% using the same pass criteria.
Conclusions: An accurate empirical algorithm for 2D patient dose reconstruction has been developed. The algorithm was applied to phantom and patient data sets and is able to calculate dose in the midplane. Results indicate that the EPID dose reconstruction algorithm presented in this work is suitable for clinical implementation.
Keywords: Back projection, electronic portal imaging device, in vivo dosimetry, OCTAVIUS 4D
|How to cite this article:|
Anvari A, Alaei P, Mohammadi M. Two-dimensional dose reconstruction using scatter correction of portal images. J Can Res Ther 2019;15:1011-7
| > Introduction|| |
Electronic portal imaging devices (EPIDs) are not only utilized for patient setup verification but are also increasingly used for dosimetric verification., Among the various types of EPIDs developed, three have been commercialized: amorphous silicon (a-Si) flat panel detectors, ionization chamber (IC)-based detectors,, and fluoroscopic/charged coupled device (CCD) camera-based EPIDs.,,,, One of the first commercially available EPIDs was the fluoroscopic type which can be acquired and added to existing linear accelerators (linacs). This type of EPID has the advantage of low price and long life span compared to other types.
In vivo dosimetry is a valuable treatment verification tool in radiation therapy. Dosimetry using portal imagers allows for fast and accurate dose verification., To use an EPID as a dosimeter, it must be calibrated and its response characterized. As it has been previously reported by our group, excellent reproducibility and stability have been observed for CCD camera-based EPIDs., Other groups have also investigated the fluoroscopic EPIDs and reported similar results.,,,
Several approaches have been proposed for EPID calibration which are categorized in two general methods: experimental methods and Monte Carlo, i.e., conversion of the grayscale pixel value to a dose value  or simulation of the grayscale pixel value., EPID dosimetry methods are categorized in two main approaches, both of which are suitable for pretreatment verification and in vivo dosimetry. In “forward approach”, the radiation fluence map obtained from a portal image is compared to the predicted dose at the EPID plane by the treatment planning system (TPS).,, In the “backward approach” or back projection, portal images are used to reconstruct the dose within the patient or phantom. The back projection method makes it possible to directly compare the calculated and delivered dose distributions in the patient or phantom.,,, While the dose is verified in only one plane with the forward approach, three-dimensional dose reconstruction is potentially possible with the back projection method.
In this paper, a fluoroscopic EPID has been used for patient dose verification and in vivo dosimetry. A few groups have used fluoroscopic EPIDs for dosimetry purposes by following several steps to convert EPID images into dose distributions. After EPID calibration, they have corrected images for optical cross talk and lateral electron transport by the complicated convolution and shift-invariant kernels.,,,,, Subsequently, EPID response was corrected by other factors such as field size dependency. Heijmen et al., Pasma et al.,, and Vieira et al. developed a method based on interpolation of cross talk contributions of a series of square fields, behind a homogeneous phantom. The kernels were based on the consecutive measurements and needed much time for accurate calculations.
Our developed empirical algorithm is simple, accurate, and without the need for excessive measurements to correct for cross talk, hence easier to implement in clinics. The algorithm introduced here provides two-dimensional (2D) transit dose at the EPID plane and is also able to reconstruct the 2D dose at midplane of phantom or patient to compare with TPS-computed one. In this study, PTW 729 ion chamber array and OCTAVIUS phantom, collectively referred to as OCTAVIUS 4D system (PTW GmbH, Freiburg, Germany), have been used to verify the dose reconstruction [Figure 1]. The accuracy of this detector has previously been studied by our group , and others.,
|Figure 1: The OCTAVIUS 4D system. The 2D ion chamber array is inserted in a cylindrical phantom that rotates synchronously with the gantry|
Click here to view
| > Subjects and Methods|| |
All measurements were performed using a new version of low-elbow fluoroscopic EPID (Cablon Medical TheraView Technology, Leusden, Netherlands) with minimum separation between the mirror and screens of 6.6 cm, coupled to a Siemens Primus linear accelerator (Siemens Medical Solutions, Erlangen, Germany). This CCD camera-based EPID has a 40.6 cm × 40.6 cm detecting surface (approximately 28.0 cm × 28.0 cm at the isocenter) with a matrix of 1024 × 1024 pixels (0.396 mm pixel size). In this work, all EPID images were acquired in dosimetry mode without any image processing at a fixed source-to-detector distance (SDD) of 150 cm with various gantry angles. Linac output was calibrated following the International Atomic Energy Agency Technical Reports Series 398 protocol  and checked on a daily basis.
Two steps are necessary to reconstruct the dose in phantom or patient from the EPID images. First, a dosimetric calibration is needed to establish the dose-response relationship between EPID pixel values and dose values at the position of the imager. Second, parameters of back projection as well as correction factors such as attenuation, and off-axis have to be determined to convert the transit dose map at EPID level to dose inside the patient or phantom.
To determine the relationship between the dose and measured EPID pixel values, the fluoroscopic EPID was exposed to radiation beam. The EPID images without any processing were assessed. Using the same setup, absolute dose value was determined using a calibrated IC on the central axis in water tank. EPID signal and IC readings were analyzed, and a relationship was established for EPID calibration. Statistical analysis was performed using the Software Package for Statistical Analysis (SPSS) version 14.0 (SPSS, Chicago, IL, USA). Level of statistical significance was adjusted for P < 0.05. We have reported the results in detail previously.,
Beam attenuation through phantom or patient and scattered radiation from medium should also be corrected to accurately reconstruct dose from portal images. Slabs of polymethyl methacrylate (PMMA) were used to evaluate EPID's response variation with slab thickness. PMMA slabs of 40 cm × 40 cm and thicknesses of 1 and 2 cm were placed on the couch with overall thickness ranging from 0 to 30 cm and irradiated using 10, 20, and 50 MUs with field sizes of 10 cm × 10 cm and 20 cm × 20 cm.
[Figure 2] depicts the experimental arrangement. As seen in the figure, the slabs were irradiated using an isocentric setup to mimic patient treatment. The attenuation correction factor (ATF) was calculated using Equation 1, in which D(t, s) is the transmitted dose through phantom and D(s) is the open field one.
|Figure 2: Schematic of experimental set-up to calculate attenuation correction factor|
Click here to view
Off-axis ratio correction
In the periphery of radiation beam, there is often an oversensitivity of the EPID signal. This effect is exacerbated by the presence of the horns in the periphery of field which are more pronounced at shallow depths [Figure 3]. Greer corrected this oversensitivity by means of off-axis ratio correction. To correct the over sensitivity of the EPID in off-axis regions and improve the accuracy of reconstructed dose map, we used a correction matrix (CMi, j). The CMi, j was calculated following the method by Mohammadi et al.,
|Figure 3: The relative profile of the electronic portal imaging device compared to the two-dimensional array showing an enhanced signal in the periphery of the radiation field. This effect is corrected by means of a correction matrix|
Click here to view
To determine the off-axis correction factor, PTW 729 array was used by placing it at the same distance as that of EPID panel (150 cm) and irradiating it with 10, 20, and 50 MUs with field sizes of 10 cm × 10 cm and 20 cm × 20 cm. The measured signal from 729 was used as the reference signal to correct the EPID signal. The correction matrix (CMi, j) was then calculated using Equation 2. The area of the field in the penumbral region (20% field) was excluded before correction matrix calculations.
Patient Dose Reconstruction Algorithm
The ultimate goal of this work was to determine the dose delivered to patient. Phantom and patient dose reconstruction was performed using back projection method from portal dose map and accounting for the correction factors listed above. To calculate patient midplane dose from transit dose map at EPID level, percentage depth dose curve and inverse square law were used.
The dose at midplane is calculated using Equation 3, wherein D measured(s) is the measured dose at the EPID level and D midplane (d, s) is the patient dose at midplane for a field size of s. Off-axis correction matrix (CMi, j) is independent of depth while the attenuation correction factor (ATF) and percentage depth dose at any position (i, j) are depth dependent.
To validate the algorithm and applied corrections, the OCTAVIUS 4D system was utilized in addition to slab phantom. Considering the cylindrical shape of the OCTAVIUS phantom and its tissue equivalency, it was deemed to be a suitable patient surrogate to evaluate the reconstruction algorithm. The 729 array was placed inside the OCTAVIUS phantom and irradiated at gantry angles of 0, 60, 90, 180, 210, and 270 using 10, 20, and 50 MUs of 6 MV photon beams and different field sizes. The reason to choose these angles was to cover all anterior, posterior, lateral, and oblique gantry angles. At the same time, portal images were acquired and the algorithm was used to reconstruct the dose inside the phantom and compare to that measured with 729.
Comparison with treatment plans
The OCTAVIUS 4D system was scanned using a Siemens Sensation 16 slice scanner (Somatom Sensation 16; Siemens Medical Solutions), and the images were used to create treatment plans using CorePLAN treatment planning system (Seoul CandJ. Inc.). Multiple treatment plans were created on the scan and delivered to the phantom at various gantry angles. The phantom was rotated to have the central axis of the beam perpendicular to 729 array during all irradiations. The delivered fields were measured with the 729 array, and the portal imager was used to capture the images as well. The reconstructed dose at the plane of 729 was then compared to measured dose.
Patient dose reconstruction
The algorithm was used to reconstruct dose from portal images for seven prostate and rectal cancer patients. Patients were treated using a four-field isocentric box technique in both supine and prone positions with 6 MV beams. Measurements were performed at the various gantry angles. For treatment verification, 2D reconstructed dose from the EPID image was compared to the dose calculated by the TPS.
| > Results|| |
Correction factors Attenuation correction
The experimental data were modeled by an exponential regression with determination coefficient of 0.999 (Equation 4). The modeled and measured data matched very well. In addition, comparison of our results to previously published ones for various slab thicknesses indicated very good agreement [Table 1].
|Table 1: The transmitted relative signal of 6 MV photon beams in 10 cm × 10 cm for various polymethyl methacrylate thicknesses|
Click here to view
ATF (t) =0.97 Exp (−0.0561 t) (4)
Using the attenuation correction factor, the signal obtained with various phantom thicknesses was converted to corresponding signal for the open field and dose reconstruction was performed. [Figure 4] shows the acquired profiles before and after the application of attenuation correction factor.
|Figure 4: Open and attenuated field profiles (with 10 cm thick polymethyl methacrylate), before and after application of attenuation correction factor|
Click here to view
Off-axis ratio correction
The over sensitivity of the imager at periphery was corrected by means of correction matrix, CMi, j. The CMi, j was calculated for SDD = 150 cm following the method described above. [Figure 5] shows the effect of CMi, j on dose maps. In this figure, the portal imager dose has been compared to that measured using 729 array. Applying the off-axis correction matrix improves the agreement between portal and measured doses. As seen in this figure, the uncorrected reconstructed dose contains errors both in the center and periphery which is reduced after application of the correction matrix.
|Figure 5: The reconstructed portal dose without correction (left), as compared to 2D array measurements (center), and portal dose corrected by means of CMi,j(right)|
Click here to view
A 10 MU delivery resulted in a 4.55 cGy dose at the center of 729 with a 1% increase of dose at the periphery. The same dose as reconstructed at the EPID level was 4.44 cGy which had a 2.5% under-response with an additional 5% over-response in the periphery. Applying the correction factor improved the reconstructed dose and reduced the error to 0.2%.
OCTAVIUS 4D system verifications
Results of comparison of the reconstructed dose in the midplane of OCTAVIUS 4D system with 729 array measurements for different field sizes are shown in [Figure 6]. Using local gamma criteria of 3%/3 mm, analysis indicated 98% ± 1% passing rate for the 10 cm × 10 cm field and 99% ± 1% passing rate for the 20 cm × 20 cm field, both for a 10 MU delivery.
|Figure 6: Comparison of the reconstructed dose in the midplane of OCTAVIUS 4D system using the proposed algorithm with that measured using 729 array for 10 × 10 (top) and 20 × 20 cm2 (bottom) fields and 10 MU delivery. The reconstructed dose map by algorithm (left) and gamma analysis comparison with OCTAVIUS measured dose map (right)|
Click here to view
Comparison with treatment plans
Results of the comparison of the reconstructed and TPS-calculated doses are shown in [Figure 7]. Using acceptance criteria of 3%/3 mm, gamma analysis indicated 98% ± 1% passing rate for the 10 cm × 10 cm field, and 99% ± 1% passing rate for the 20 cm × 20 cm field, both for a 10 MU delivery. Additional gamma analysis based on varying criteria was also performed and tabulated in [Table 2].
|Figure 7: Comparison of the reconstructed dose in the midplane of OCTAVIUS 4D system using the proposed algorithm with that calculated using the treatment planning system for 10 × 10 (top) and 20 cm × 20 cm (bottom) fields and 10 MU delivery. The calculated dose map by treatment planning system (left) and gamma analysis comparison with reconstructed dose map (right)|
Click here to view
|Table 2: Comparison of the reconstructed dose in the midplane of OCTAVIUS 4D system with treatment planning system-generated one using various gamma criteria|
Click here to view
Patient dose reconstruction
[Figure 8] demonstrates the reconstructed midplane dose and its comparison with TPS-generated planar dose using one of patient portal images. Gamma analysis results indicate that 96% ±2% of the points passed 3%/3 mm acceptance criteria. [Table 3] is a compilation of gamma values in an 8 cm × 8 cm area, comparing EPID-reconstructed dose and TPS-generated one, with a 0.44 mean values and only 4 pixels having a gamma index >1.
|Figure 8: Reconstructed dose in patient midplane (left) and results of its comparison with treatment planning system-calculated dose (right) in an 8 cm × 8 cm area with 40 MU delivery|
Click here to view
|Table 3: Gamma values of comparison of patient midplane reconstructed dose with treatment planning system-calculated one in an 8 cm × 8 cm area|
Click here to view
| > Discussion|| |
EPID is becoming a viable tool for patient dose reconstruction and in vivo dosimetry in radiation therapy. The actual delivered dose distribution may deviate from the planned one due to differences in patient position, anatomic changes, or variations of linac output during treatment. In this study, fluence maps were obtained using a fluoroscopic EPID and used to reconstruct 2D dose distributions by applying various corrections. The method was evaluated in phantom and patient with satisfactory results. The developed empirical algorithm is simple, accurate, and easy to implement with minimal amount of needed time.
The outcome of the dose reconstruction after applying relevant correction factors indicated no errors on the central portions of the beams, with the errors exclusive to the periphery of the fields. The fluoroscopic EPID was found to have an over-response in the field edges, same as that reported for a-Si EPIDs. This was accounted for using a correction matrix. Considering the relative importance of central parts of the field compared to the periphery, elimination of the 25% of the peripheral dose improves the accuracy of the algorithm to better than 99%, utilizing gamma analysis.
Previous works in this area were limited to dose reconstruction at specific depth within patients, which had limitations in EPID dosimetry, especially because of inaccuracies in dose calculations due to the variation in dose as a function of the collimation. This algorithm can calculate 2D transit dose at the EPID plane and use that to reconstruct 2D dose at midplane within phantom or patient for evaluation against the planar dose generated by TPS. The algorithm's ability to calculate the 2D dose assures accuracy of dose delivery.
In the developed algorithm, the field size correction factor has been used. This is suitable for standard treatment techniques such as a four-field box but not for intensity-modulated radiotherapy or volumetric-modulated arc therapy. The utility of correction matrix in complicated plans is under investigation. The planar dose reconstructed by back projection can provide a reliable method to conduct pretreatment quality assurance or in vivo dosimetry. Alternatively, the measured central axis dose by the EPID can be convolved with a point spread function that is determined empirically or by a fitted Gaussian equation between the ion chamber and EPID data.
| > Conclusion|| |
In vivo EPID dosimetry provides an additional patient safety measure by verifying the dose distribution within patient. An accurate back-projection algorithm for 2D in vivo EPID dosimetry has been developed. In this algorithm, a fluoroscopic EPID was calibrated against IC measurements and the proposed algorithm was corrected using the 729 ion chamber array measurements. The validation of the algorithm has been done using slab phantom and OCTAVIUS 4D system. This algorithm can predict 2D transit dose at EPID level precisely and can calculate 2D patient dose accurately at midplane.
Financial support and sponsorship
Conflicts of interest
There are no conflicts of interest.
| > References|| |
Francois P, Boissard P, Berger L, Mazal A.In vivo
dose verification from back projection of a transit dose measurement on the central axis of photon beams. Phys Med 2011;27:1-0.
Fidanzio A, Azario L, Greco F, Cilla S, Piermattei A. Routine EPID in-vivo
dosimetry in a reference point for conformal radiotherapy treatments. Phys Med Biol 2015;60:N141-50.
Baker SJ, Budgell GJ, MacKay RI. Use of an amorphous silicon electronic portal imaging device for multileaf collimator quality control and calibration. Phys Med Biol 2005;50:1377-92.
Mohammadi M, Bezak E. Two-dimensional transmitted dose measurements using a scanning liquid ionization chamber EPID. Phys Med Biol 2006;51:2971-85.
Mohammadi M, Bezak E, Reich P. Comparison of two-dimensional transmitted dose maps: Evaluation of existing algorithms. Australas Phys Eng Sci Med 2006;29:179-87.
de Boer JC, Heijmen BJ, Pasma KL, Visser AG. Characterization of a high-elbow, fluoroscopic electronic portal imaging device for portal dosimetry. Phys Med Biol 2000;45:197-216.
Pasma KL, Heijmen BJ, Kroonwijk M, Visser AG. Portal dose image (PDI) prediction for dosimetric treatment verification in radiotherapy. I. An algorithm for open beams. Med Phys 1998;25:830-40.
Glendinning AG, Bonnett DE. Dosimetric properties of the theraview fluoroscopic electronic portal imaging device. Br J Radiol 2000;73:517-30.
Anvari A, Aghamiri SM, Mahdavi SR, Alaei P. Dose response characteristics of a novel CCD camera-based electronic portal imaging device comparison with OCTAVIUS detector. J Cancer Res Ther 2015;11:765-9.
Anvari A, Aghamiri SM, Mahdavi SR, Alaei P, Mohammadi M. Dosimetric properties of fluoroscopic EPID for transit dosimetry. J Radiother Pract 2015;14:27-34.
Camilleri J, Mazurier J, Franck D, Dudouet P, Latorzeff I, Franceries X, et al.
2D EPID dose calibration for pretreatment quality control of conformal and IMRT fields: A simple and fast convolution approach. Phys Med 2016;32:133-40.
McCurdy BM, Luchka K, Pistorius S. Dosimetric investigation and portal dose image prediction using an amorphous silicon electronic portal imaging device. Med Phys 2001;28:911-24.
Heijmen BJ, Pasma KL, Kroonwijk M, Althof VG, de Boer JC, Visser AG, et al.
Portal dose measurement in radiotherapy using an electronic portal imaging device (EPID). Phys Med Biol 1995;40:1943-55.
Kirby MC, Atherton S, Carson P, McDonagh C, Sykes JR, Williams PC, et al.
The performance of a fluoroscopic electronic portal imaging device modified for portability. Br J Radiol 1999;72:1000-5.
Greer PB, Vial P, Oliver L, Baldock C. Experimental investigation of the response of an amorphous silicon EPID to intensity modulated radiotherapy beams. Med Phys 2007;34:4389-98.
Pasma KL, Kroonwijk M, de Boer JC, Visser AG, Heijmen BJ. Accurate portal dose measurement with a fluoroscopic electronic portal imaging device (EPID) for open and wedged beams and dynamic multileaf collimation. Phys Med Biol 1998;43:2047-60.
Pasma KL, Vieira SC, Heijmen BJ. Portal dose image prediction for dosimetric treatment verification in radiotherapy. II. An algorithm for wedged beams. Med Phys 2002;29:925-31.
Wendling M, McDermott LN, Mans A, Sonke JJ, van Herk M, Mijnheer BJ, et al.
A simple backprojection algorithm for 3D in vivo
EPID dosimetry of IMRT treatments. Med Phys 2009;36:3310-21.
Spreeuw H, Rozendaal R, Camargo P, Mans A, Wendling M, Olaciregui-Ruiz I, et al.
Portal dosimetry in wedged beams. J Appl Clin Med Phys 2015;16:5375.
Vieira SC, Dirkx ML, Pasma KL, Heijmen BJ. Dosimetric verification of x-ray fields with steep dose gradients using an electronic portal imaging device. Phys Med Biol 2003;48:157-66.
Franken E, De Boer H, Heijmen B, editors. A closed system for dosimetric imrtverification with an EPID, eliminating ion chamber calibration measurements. Radiother Oncol 2006;81:S157.
Anvari A, Aghamiri S, Mahdavi S, Alaei P. SU-E-T-65: Characterization of a 2D Array for QA and Pretreatment Plan Verification. Medical Physics 2014;41:236-7.
Stathakis S, Myers P, Esquivel C, Mavroidis P, Papanikolaou N. Characterization of a novel 2D array dosimeter for patient-specific quality assurance with volumetric arc therapy. Medical physics 2013;40:071731.
Van Esch A, Basta K, Evrard M, Ghislain M, Sergent F, Huyskens DP. The Octavius1500 2D ion chamber array and its associated phantoms: Dosimetric characterization of a new prototype. Medical physics 2014;41:091708.
Andreo P, Burns D, Hohlfeld K, Huq MS, Kanai T, Laitano F, et al
. IAEA TRS-398. Absorbed dose determination in external beam radiotherapy: an international code of practice for dosimetry based on standards of absorbed dose to water. International Atomic Energy Agency 2000.
Wendling M, Louwe RJ, McDermott LN, Sonke JJ, van Herk M, Mijnheer BJ. Accurate two-dimensional IMRT verification using a back-projection EPID dosimetry method. Medical physics 2006;33:259-73.
Greer PB. Correction of pixel sensitivity variation and off-axis response for amorphous silicon EPID dosimetry. Medical physics 2005;32:3558-68.
Jeon H, Park D, Lee J, Nam J, Kim W, Ki Y, et al
. An Experimental Method for the Scatter Correction of MV Images Using Scatter to Primary Ratios (SPRs). Progress in Medical Physics 2014;25:143-50.
Pouliot J. Megavoltage imaging, megavoltage cone beam CT and dose-guided radiation therapy. Frontiers of Radiation Therapy and Oncology 2007;40:132-42.
Hsu SH, Moran JM, Chen Y, Kulasekere R, Roberson PL. Dose discrepancies in the buildup region and their impact on dose calculations for IMRT fields. Medical Physics 2010;37:2043-53.
[Figure 1], [Figure 2], [Figure 3], [Figure 4], [Figure 5], [Figure 6], [Figure 7], [Figure 8]
[Table 1], [Table 2], [Table 3]