Experimental Validation of the Eclipse AAA Algorithm

Abstract

This study evaluates the performance of a newly released photon beam dose calculation algorithm that is incorporated into an established treatment planning system (TPS). The Analytical Anisotropic Algorithm (AAA), pre-configured with the “Golden Beam Data” for Varian linear accelerators was compared with measurements performed at two institutions using 6 and 15 MV beams. TG-53 evaluation regions and criteria were used to evaluate profiles measured in a water phantom for a wide variety of clinically relevant beam geometries. The total scatter factor (TSF) for each of these geometries was also measured and compared against the results from the AAA algorithm. At one institute, TLD measurements were performed at several points in the neck and thoracic regions of a Rando phantom while ion chamber measurements were performed in a CIRC inhomogenous phantom at the other institution. These phantoms were imaged using CT and the dose calculated using the AAA algorithm at corresponding detector locations. Evaluation of measured relative dose profiles revealed that 97%, 99%, 97% and 100% of points at one institute and 96%, 88%, 89% and 100% of the points at the other institute pass TG-53 evaluation criteria in the outer beam, penumbra, inner beam and buildup regions respectively. Poorer results in the penumbra and inner beam regions at one institute are attributed to a fewer number of points analyzed, insensitivity of AAA modeled penumbra to the measurement data and a possible mismatch of the accelerator output with the “Golden Beam Data”. For the validation of the monitor unit calculations, the mean difference between measured and calculated TSFs was <0.5%; test cases involving physical wedges had, in general, >1% differences. Point measurements performed in inhomogeneous phantoms compared reasonably well with AAA output and were within TG-53 guidelines. By intent, the methods and evaluation techniques were similar to a previous investigation involving another convolution-superposition photon beam dose calculation algorithm in another TPS, so that our analysis allowed an independent comparison between the two algorithms. The AAA algorithm, within the frame work of measurements and analysis performed in this study, compares very well with the evaluation reported in the previous study.