32 Comparison of the dose calculation algorithms in a heterogeneous media using experimental design
Résumé
Introduction
This study compares the performances of different dose calculation algorithms in heterogeneous medium. Then, the properties of two algorithms on clinical cases are considered for the thoracic cases in conformal 3D radiotherapy (RC3D).
Methods
(1) A study on the CIRS phantom was performed by comparing the dose distributions measured with Gafchromic EBT3® with those calculated by the AAA [Analytical Anisotropic Algorithm v13, Varian], AXB [Acuros XB v13.7, Varian], PB [Pencil Beam v4.5.5, Brainlab], MC [Monte Carlo v4.5.5, Brainlab] and CCV [Collapsed Cone Volume v2.0.1, Mobius Medical System]. The performances of these different algorithms were tested using an experimental design of Taguchi Math Eq to demonstrate the influence of the following factors, in 18 tests: energy, collimator angle, gantry angle, X and Y field size.
A comparison by subtracting the dose distributions between those measured and computed was achieved for the 18 tests. (2) The impact of the selection of the algorithm was performed on treatment plans initially calculated with AAA, on a set of 15 patients with thoracic cancer in RC3D. By keeping the same UM number, these treatment plans were compared in terms of coverage of the 95% PTV, a conformity index, doses to the PTV and organs at risk. If necessary, the planimetry was modified in AXB (weighting, filtering and normalization point) in order to validate the treatment plan clinically.
Results
(1) In heterogeneous medium, the average dose differences between the measured and calculated doses are, respectively, for the AXB, AAA, CCV, MC and PB: 0.01.8%, 1.22.2%, −0.12.0%, 1.21.9% and 7.04.1%. (2) For treatment plans with similar MU numbers, an average dose difference of −1.2% in the level of the 95% coverage of the PTV and a modification of conformity index from 1.10 (AAA) to 0.76 (AXB) are obtained, attesting to a lower coverage of the PTV with AXB, related to a difference in the evaluation of heterogeneities. Once these treatment plans were modified, the resulting planimetry was similar to the clinically validated one, with a mean difference of MU greater than 1.5% against treatment plans calculated with AAA.
Conclusions
The best performances are shown with the MC and the AXB algorithms. The experimental design allows reliable comparison of the algorithms with 18 tests. The PEX highlights bigger deviations for small field modelling. Regarding the dosimetry study, AXB can be used for RC3D thoracic cancer treatment. A complementary study for the same localization is ongoing for the VMAT treatment technique.