A two step algorithm for predicting portal dose images in arbitrary detectors

An algorithm has been presented which accurately predicts portal dose images for arbitrary detectors and air gaps. Implementation involves first predicting the primary and scattered photon fluence into a detector, then predicting the dose response of the detector. The algorithm utilizes pre-calculated libraries of scatter fluence kernels and dose deposition kernels, which are obtained through Monte Carlo radiation transport techniques. The algorithm is fast, allows a separation of primary and scatter, and can model arbitrary detector materials. The accuracy of the algorithm was investigated for a 6 MV beam over air gaps of 10-80 cm for a PMMA slab phantom, a PMMA slab with a cork inhomogeneity, and an anthropomorphic phantom. Two different detector configurations were used, involving low and high atomic number buildup material. In most cases (>95%), the difference between predicted and measured doses is within 3%, and penumbra's are within 4 mm. This level of accuracy is within the guidelines set out for treatment planning dose calculation algorithms. It is concluded that this approach represents a fast, accurate, and flexible solution to portal dose image prediction.