Accuracy of dose calculation algorithms for virtual heterogeneous phantoms and intensity-modulated radiation therapy in the head and neck.

This study verified the dose calculation accuracy of the analytical anisotropic algorithm (AAA), Acuros XB version 10 (AXB10), and version 11 (AXB11) installed in an Eclipse treatment planning system, by comparing with Monte Carlo (MC) simulations. First, the algorithms were compared in terms of dose distributions using four types of virtual heterogeneous multi-layer phantom for 6 and 15 MV photons. Next, the clinical head and neck intensity-modulated radiation therapy (IMRT) dose distributions for 6 MV photons were evaluated using dose volume histograms (DVHs) and three-dimensional gamma analysis. In percentage depth doses (PDDs) for virtual heterogeneous phantoms, AAA overestimated absorbed doses in the air cavity, bone, and aluminum in comparison with MC, AXB10, and AXB11. The PDDs of AXB10 almost agreed with those of MC and AXB11, except for the air cavity. The dose in the air cavity was higher for AXB10 than for AXB11, because their electron cutoff energies are set at 500 and 200 keV, respectively. For head and neck IMRT dose distributions, the D95 in the clinical target volume (CTV) for AAA was almost the same as that for AXB10 and was approximately 7 % larger than that for MC. Comparing each approach with MC using a criterion of 3 %/3 mm, the pass rates for AXB10, AXB11, and AAA were 92.4, 94.7, and 90.4 % in the CTV, respectively. In conclusion, AAA produces dose errors in heterogeneous regions, while AXB11 provides calculation accuracy comparable to MC. AXB10 overestimates the dose in regions that include an air cavity.