Dosimetric evaluation of a three-dimensional treatment planning system.

The computerized treatment planning system plays a major role in radiation therapy in delivering correct radiation dose to the patients within ±5% as recommended by the ICRU. To evaluate the dosimetric performance of the Treatment Planning system (TPS) with three-dimensional dose calculation algorithm using the basic beam data measured for 6 MV X-rays. Eleven numbers of test cases were created according to the Technical Report Series-430 (TRS 430) and are used to evaluate the TPS in a homogeneous water phantom. These cases involve simple field arrangements as well as the presence of a low-density material in the beam to resemble an air in-homogeneity. Absolute dose measurements were performed for the each case with the MU calculation given by the TPS, and the measured dose is compared with the corresponding TPS calculated dose values. The result yields a percentage difference maximum of 2.38% for all simple test cases. For complex test cases in the presence of in-homogeneity, beam modifiers or beam modifiers with asymmetric fields a maximum percentage difference of 5.94% was observed. This study ensures that the dosimetric calculations performed by the TPS are within the accuracy of ±5% which is very much warranted in patient dose delivery. The test procedures are simple, not only during the installation of TPS, but also repeated at periodic intervals.

Introduction

Radiotherapy aims to cure, or locally control the disease, while concurrently minimizing complications in normal tissues. The radiation dose must be delivered within ±5% of the prescribed dose.[1-3] The treatment planning and dosimetry are the main steps in radiotherapy, which includes calibration of the equipments, determination of absorbed dose under reference conditions, phantom measurements under non-reference conditions, calculation of dose distribution in the patient and, finally, treatment delivery via monitor units or treatment time calculation. Consideration of the uncertainties associated with each of the above steps and their propagation increases the demand for accuracy in the dose calculation algorithm employed in the treatment planning. Therefore, quality assurance (QA) is necessary in the commissioning stage of the treatment planning system (TPS) prior to their use in clinical practice.

In the present work, the dosimetric performance of a commercial TPS with a three-dimensional calculation algorithm (Plato V 2.7.2, Nucletron B.V) is studied using a basic beam data set measured for a 6 MV X-ray beam and a set of test case configurations which are based on the TRS-430. The aim is to determine the accuracy of our TPS in dose calculation in a homogenous phantom as well as in the presence of in-homogeneity, and potential limitations of the dose calculation algorithm.

Materials and Methods

Treatment planning system

The TPS agreement with the treatment machine is based on IEC 1217 conventions[4] for specifying gantry angle, collimator angle, table angle, wedge orientation and patient orientation, and TPS software operates in a UNIX environment. The photon beam dose calculation algorithm employed by TPS comprises a convolution based approach where the energy fluence distribution is convolved with a dose pencil beam.[5] The dose pencil beam consists of depth-independent width and depth-dependent relative weight. Thus the dose calculation at any arbitrary depth in a homogeneous water phantom involves only one single convolution step for each of the three components. For other depths, the pencil beam components are added using the depth-dependent relative weights. Convolution of the energy fluence distribution with a Gaussian source distribution kernel allows for optimization of the fit between the measured and the calculated edge of the field. In-homogeneities are taken into account by applying the equivalent tissue-air ratio (ETAR) method introduced by Sontag and Cunningham,[6] as described by Yu andWong.[7]

Instrumentation and technique

Beam data and test point doses were measured for a 6 MV X-ray beam (Quality Index-0.665) of a Linear Accelerator (Primus, Siemens, Germany). Percentage depth dose curves and beam profiles were measured with a fully computerized radiation field analyzer (Blue phantom, Scanditronix-Wellhofer, Germany) equipped with a thimble- type ionization chamber and semiconductor detectors for relative dose measurements. Absolute dose measurements were performed with ionization chambers (0.13 cc, 0.6 cc farmer type and 40 cc parallel plate type, Scanditronix-Wellhofer) connected to an electrometer (Dose 1, Scanditronix-Wellhofer, Germany). The chamber was calibrated for N D, W according to the IAEA TRS-398 dosimetry protocol.[8]

Basic beam data

The basic beam data were measured under reference conditions of source to surface distance (SSDref ) = 100 cm, reference field (FSref )=10 x 10 cm 2 , reference depth (rref )=10 cm. The machine was calibrated to deliver 1 cGy/MU at the depth of maximum dose (Dmax )at 1.7 cm. The beam data used for beam modeling include the following parameters.

Depth dose data

Open beam depth dose data along the central axis of square field sizes of: 3, 5, 8, 10, 12, 15, 18, 20, 25, 30, 40 (cm x cm) for depths from 0 up to 30 cm were measured.

Off-axis beam profiles

For each of the above-mentioned square fields, five open beamprofiles at various depths of Dmax , 5 cm, 10 cm, 15 cm, and20 cm were acquired.

Wedge field data

Depth dose data and five beam profiles for square field sizesof 3, 5, 10, 15, and 20 (cm x cm) were used for each wedge of 15°,30°, 45°, and 60° nominal angles. Differences betweencalculated and measured profiles were minimized by adjustmentof weight factors according to the comparison of the calculatedwedged beam profiles at depths of 3, 5, 10, 15, and 20 cmwith corresponding measured values. The comparison and evaluationof differences between calculated and measured beam data wereperformed in a trial and error fashion using the Beam Data AnalysisSoftware (BDAS).

The output factors, wedge factors, tray transmission factors, and block transmission factors also acquired for the minimum to maximum field size as an input data for the TPS.

Test data

The selected test cases representing different aspects of the dose computation process as proposed by TRS 430[9] and other authors[10-13] were created. Point dose measurements were performed, and the measured doses were compared with that of TPS calculated values. Test point measurements correspond to different depths ranging from 0.2 to and 15 cm along the central axis for the reference SSD ref =100 cm, unless otherwise stated.

Test case 1

Square fields ranging from 3 x 3 cm2 (the smallest used in our department) up to 28 x 28 cm2 at the depth of 10 and 15 cm along the central beam axis.

Test case 2

Rectangular fields were produced by exchanging the x and y jaws (x × y and y 0× x) without collimator rotation. Rectangular fields and equivalent square fields were also examined along the beam axis.

Test case 3

SSD variation: 13 test points of isocentric setup were investigated in the beam axis, it includes the Anterior and posterior, box and tangential arrangements of clinical situations with symmetric as well as asymmetric fields.

Test case 4

Wedge filter: square fields of 10 x 10 cm2 and 20 x 20 cm2 modified with 15°, 30°and 45° wedge filter were investigated at the depth of 10 cm. Two measurements were performed along the beam axis for each of the possible wedge orientations. The average measured dose value was then compared with the corresponding calculated value.

Test case 5

Central block: a diverging cerrobend block of 4 x 16 cm2 dimension at the isocenter and 7.5 cm thickness resulting in 95% effective attenuation was investigated. Point dose measurements were performed for square fields of 10 x 10 cm2 , 15 x 15 cm2 at the depth 10 cm for a SSD of 100 and 90 cm and at a distance of 0.5 and 1 cm away from the shielding block (2.5 and 3 cm from the central beam axis). The dose values were compared with the corresponding calculated values.

Test case 6

Off-center planes: point dose measurements were performed fora variety of square fields and off-centred planes, i.e. [5 cmx 5 cm, 2 cm], [7 cm x 7 cm, 3 cm],[10 cm x 10 cm, 4 cm], [13 cm x 13 cm,5 cm], [15 cm x 15 cm, 6 cm].The average of the four off-center dose points in the cross-planeand in-plane directions was used as the mean off center dosevalue in both measurements and calculations.

Test case 7

Oblique incidence: the aim of this test was to check the abilityof the TPS to account for oblique incidence and skin contourvariation. Using an isocentric setup and gantry angles of ±20°, ±30°, the dose was determined attwo depths of 5 cm and 10 cm along the central beamaxis for FSD (Gantry angle=0°) =95 cm and FSD (Gantry angle=0°) = 90 cm. Field sizes perpendicular to the beamdirection were 10 cm x 10 cm, 15 cm x 15 cmand 20 cm x 20 cm respectively.

Test case 8

Inhomogeneous medium: An air gap of 3 cm height was created in the solid water phantom to checkthe ability of the TPS to account for the presence of in-homogeneities.The in-homogeneity was perpendicular to the beam axis and shape with a 20x 20 cm2 side area and 3 cm thickness. Point dose was measured at 10 cm depth, SSD = 100 cm along the centralaxis for square field sizes of 5 x 5 cm2 , 8 x 8 cm2 , 10x 10 cm2 , 15 x 15 cm2 .

Test case 9

Asymmetric wedged fields: 16 number of test points were created at the depth of 10 and 15 cm along the beam axis with asymmetric field sizes {10 cm x 15(Y1:5,Y2:10) cm}, {3(X1:2,X2:3) cm x 10 cm}, {2 cm x 12(Y1:8,Y2:4) cm}, {13(X1:9,X2:4) cm x 7 cm} with universal wedges of 15° and 45° with both way of insertion.

Test case 10

Build-up region behavior: 75 number of test points were created in the build-up region form 0.2 cm to 1.0 in steps of 1 mm and 1 cm to 2.2 cm in steps of 2mm for the field sizes of 5, 10, and 20 cm2 with open, shielding tray, and wedge. The doses were measured using a parallel plate chamber.

Test case 11

Shaped fields: Three irregular shapes were created using cerrobend blocks as to simulate 3D conformal treatment and 11 number of test points were created at the depths of 3, 5, 10, and 15 cm for simple and complex geometries.

Results

Calculated (Dcalc) and measured (Dmeas) dose values were comparedfor each of the eleven test cases. The TPS dosimetric performancewas evaluated by calculating the deviation (δ) at the specific depth,[14-16] using

In total, 201 test point measurementsand calculations were compared for the 6 MV photon beamof the linear accelerator. The final outcome of the comparison is summarizedin Table 1, in the form of mean, maximum, and minimum deviations.

Table 1

Results of test cases

In test case 1, the TPS calculations for open square fieldsare in good agreement with measured values presenting a minimum deviation of -0.28 % and a maximum deviation of 1.57 % [Figure 1] and none of the test points exceed the recommended tolerancelevel of 2%. The results also reveal thatas field size increases, deviations become negative,since the TPS tends to underestimate dose in relatively largefields. It may be the limitation of the algorithm that is used, and our results are agreeing with that of Sandiloset al.[16]

Figure 1

Dose comparisons: Histograms of the differences between calculated and measured values in percentage for test case 1.

In test case 2, open rectangular fields were investigated. The minimum and maximum deviations are found to be -0.08% and 2.38% [Figure 2]. Except in two points, all of the test points satisfy the tolerance level of 2%. This finding, combined with the dosimetric verification of calculationsfor rectangular fields, supports the adequacy of the equivalentsquare method. Moreover, a trend of TPS dose underestimationwith increasing field size is also observed for rectangularfields.

Figure 2

Dose comparisons: Histograms of the differences between calculated and measured values in percentage for test case 2

In test case 3, the influence of SSD variation on TPS dose calculationswas investigated. The maximum and minimum deviations are found to be 0.15% and -1.44% [Figure 3]. None of the 13 test point measurementsexceeded the tolerance level of 3%. As SSD decreases, absolutedeviations were found to increase for all field sizes, beingwithin the acceptable tolerance level. TPS dose calculationswere smaller than measured values for most of the points.

Figure 3

Dose comparisons: Histograms of the differences between calculated and measured values in percentage for test case 3

In test case 4, the influence of introduction of wedge filters in the field were investigated. A minimum deviation of 0.24% and a maximum of 4.09% were observed [Figure 4]. Four of the 12 test points, the deviation are found to be more than the prescribed limit of 3% by Venselaar et al.[15] TPS was found to overestimatedose and also the deviation between the measured and calculated values were increasing with the wedge angles. In test case 5 the influence of introduction of central block in the beam was investigated. Four test points were measured in the inner beam and[9] maximum and minimum deviations of 0.13% and -1.92% were observed [Figure 5]. The results for all testpoints presented positive deviation, which were well withinthe recommended tolerance limit of 3%. This implies proper configurationof the TPS shielding block.

Figure 4

Dose comparisons: Histograms of the differences between calculated and measured values in percentage for test case 4

Figure 5

Dose comparisons: Histograms of the differences between calculated and measured values in percentage for test case 5

In test case 6, the accuracy of dose calculation in the off-centred plane were investigated, presenting a maximum and minimum deviation of 0.4% and -2.64%. Except in two points all other points are well within the tolerance level of 3%. The deviations are presented in Figure 6. The test case 7 investigates the influence oblique incidence of beam in the dose calculation algorithm, the deviations were found to be a minimum of -0.92% and a maximum of 0.81% [Figure 7], which is well within the acceptance limitof 3%. The deviations were found to increase with an oblique angle.

Figure 6

Dose comparisons: Histograms of the differences between calculated and measured values in percentage for test case 6

Figure 7

Dose comparisons: Histograms of the differences between calculated and measured values in percentage for test case 7

The test case 8 investigates the accountability of density correction by the treatment planning system; the deviations were found to be in the range of 0.93% to 2.67% [Figure 8]. None of the testpoints exceeded the criterion of 3%.In test case 9, the influence of wedge filters in the asymmetric field were investigated, except in 2 test points of small field sizes all other point the deviations are well within the tolerance level of 4% as shown in the Figure 9.

Figure 8

Dose comparisons: Histograms of the differences between calculated and measured values in percentage for test case 8

Figure 9

Dose comparisons: Histograms of the differences between calculated and measured values in percentage for test case 9

In test case 10, the build-up region behavior was analyzed by measuring dose at 75 test points for open, wedged and shielding tray fields using the parallel plate chamber (PPC40, Scanditronix-Wellhofer, Germany). The build-up region has been divided into three parts as 20-30%, 30-80%, and 80-100% dose regions. In the region of 0.2 cm to 0.5 cm (20-30% dose), a maximum deviation of 14.28% and a minimum deviation of -10.82% is observed, and from 0.6 cm to 1.2 cm (30-80% dose region) of build-up region a maximum deviation of 13.16% and a minimum deviation of -3.51% is observed, in the region of 1.2 cm to 1.6 cm a maximum deviation of 15.63% and a minimum deviation of 13.16% is observed for the field size of 10 cm2 [Figure 10]. The same pattern is observed for 5 and 20 cm2 field sizes. The deviation were less only in the region of 30% to 80% of build-up dose region. All the deviations are within the tolerance limit of 20%.[9] The introduction of shielding tray increases the deviation in the 20-30% dose region, and the deviations were found to be in the negative side. In the wedge field, at 30-80% of build-up dose region a maximum deviation of 12.61% and a minimum deviation of -1.68% is observed [Figure 11]; it may be due to the beam hardening. The deviations are found to be well below the tolerance limit of 50%.[9] In test case 11, cerrobend block-shaped field were used, a minimum deviation of -0.44% and a maximum deviation of 5.94% are absorbed [Figure 12], and the maximum deviation is absorbed in the complex geometry.[9] All other points are well within the tolerance limit of 2%.

Figure 10

Dose comparisons: Histograms of the differences between calculated and measured values in percentage at different depths for 10 cm2 field size for test case 10

Figure 11

Dose comparisons: Histograms of the differences between calculated and measured values in percentage at different depths for 10 cm2 field size with Shielding tray and wedge for test case 10

Figure 12

Dose comparisons: Histograms of the differences between calculated and measured values in percentage for test case 11

Discussion

The first criteria published by Van Dyk et al. in 1993[10] are characterized by increased tolerance limits dueto the fact that most of the TPS were using two-dimensionalalgorithms at the time. The recommendations of AAPM TG53 reportin 1998,[17] report 7 of the Swiss Society for Radiobiologyand Medical Physics (SSRMP) in 1997[14] and Venselaar et al.in 2001[15] are generally more strict, but realistic for a properlyfunctioning dose calculation algorithm. When the complexityof the geometry increases, however, tolerance limits may haveto be less strict relative to beam modelling geometry. In thiswork the set of tolerance limits proposed by Venselaar et al. and TRS 430 is followed. The test cases used can be divided in three groups in terms of increasing complexity of the test configuration. The first group includes simple geometrical test cases (square and rectangular fields, SSD variation, off centre plane and oblique incidence) where dose calculations are performed in a homogeneous phantom for fields without special accessories. The second group includes complex geometrical test cases (wedge, central block, and in-homogeneities). The third group consists of more complex geometries which include combination of first and second groups. The first group of checks (test cases 1, 2, 3, 6, 7 and 11) has also been studied by Alam et al.,[1] Venselaar et al.,[15] and Sandilos et al.[16] for the older PLATO versions 1.21, 2.01, and 2.2.3, respectively. The Nucletron Plato version 1.21 employs a two-dimensional dose calculation algorithm, while versions 2.01 (Venselaar et al.), 2.2.3 (Sandilos et al.), and 2.7.2 (present study) employ a 3D dose calculation algorithm. Comparing results of these previous studies[11516] with those of the present work for version 2.7.2, a continuous improvement of the system is evident. Although older TPS versions also met the tolerance limit for these test cases, a reduction of calculated to measured dose deviations is reported here. The above conclusion assumes ideal modeling of these systems.

In the second group of checks (test cases 4, 5 and 8), differences between the calculated and measured values are well within the tolerance limit of ±3% except for some points in the wedged fields. The present results are comparable with that of Sandilos et al.[16] Third group of check (Test case 9) examines more complex geometries, and most of the test point results fall within ±4% of tolerance.

However, it should be noted that only theaccuracy of the dose calculation algorithm has been investigatedwithout examining other potential inaccuracies associated withthe geometry in the TPS (CT image acquisition and transfer,graphical display of 3D radiation beams, etc.). In addition,results of this work are limited to the 6 MV of the Linear accelerator photon beams.

Conclusion

An attempt has been made to study the performance of the TPS (PLATO V 2.7.2) by using 11 numbers of test cases and 201-point dose measurements for a 6 MV photon beam. The measured- and TPS-calculated point doses are well within the tolerance of ±5%. The study concludes that (i) for higher field sizes the TPS tends to underestimate the dose for both square and rectangular field sizes, (ii) for smaller SSDs the deviations were found to increase, (iii) TPS was found to overestimate the dose for increasing wedge angles, (iv) the deviation were found to increase with an increase of obliquity of a beam angle, (v) in the build-up region analysis, the deviation were less only in the 30--80% of build-up dose region, the deviation were found to be less in wedged fields when compared with open fields and this may be due to the beam hardening effect.

The study has ensured the correctness of the beam data entered in the TPS during the commissioning. The usefulness of test data provided by TRS 430 and Venselaar et al. are verified for QA and inter-comparison of new radiotherapytreatment planning systems.Nevertheless, the beam modelling and basic data entered in eachsystem depend on the user and the particular features of eachsystem. This present methodology may be used to inter compare TPS, in various hospitals. This study also ensures that dosimetric calculations performed by the TPS is very accurate and enables the user to achieve the accuracy of ±5%, which is very much warranted in patient dose delivery.