Beam modeling and verification of a photon beam multisource model

Abstract

Dose calculations for treatment planning of photon beam radiotherapy require a model of the beam to drive the dose calculation models. The beam shaping process involves scattering and filtering that yield radiation components which vary with collimator settings. The necessity to model these components has motivated the development of multisource beam models. We describe and evaluate clinical photon beam modeling based on multisource models, including lateral beam quality variations. The evaluation is based on user data for a pencil kernel algorithm and a point kernel algorithm (collapsed cone) used in the clinical treatment planning systems Helax-TMS and Nucletron-Oncentra. The pencil kernel implementations treat the beam spectrum as lateral invariant while the collapsed cone involves off axis softening of the spectrum. Both algorithms include modeling of head scatter components. The parameters of the beam model are derived from measured beam data in a semiautomatic process called RDH (radiation data handling) that, in sequential steps, minimizes the deviations in calculated dose versus the measured data. The RDH procedure is reviewed and the results of processing data from a large number of treatment units are analyzed for the two dose calculation algorithms. The results for both algorithms are similar, with slightly better results for the collapsed cone implementations. For open beams, 87% of the machines have maximum errors less than 2.5%. For wedged beams the errors were found to increase with increasing wedge angle. Internal, motorized wedges did yield slightly larger errors than external wedges. These results reflect the increased complexity, both experimentally and computationally, when wedges are used compared to open beams.

Monitor unit calculations for external photon and electron beams: Report of the AAPM Therapy Physics Committee Task Group No. 71

“…70,[72][73][74] Additionally, depending on the planning system it may be possible to extract the primary energy fluence transmitted through a flat water phantom to determine POAR. In either case, it is recommended that a few sample measurements be made to confirm these data.…”

Patient Dose Computation

“…The sources included in C hsc depend on the implementation but to explicitly model at least the flattening filter as an extended source (see Figure 4) is common. Ahnesjö et al (2005) included three different sources in C hsc : flattening filter scatter through an area integral over the part of the filter visible through the collimating system from the point of interest; scatter from hard wedges, if present in the beam, with an integration method similar to that for the flattening filter; and collimator scatter by means of a weighted line integral along the field periphery as viewed from the calculation point of interest (Ahnesjö , 1995). Collimator scatter is usually of low magnitude in photon beams -most of the scatter variations with collimator settings stems from changes in flattening filter scatter -and some systems merge those two contributions into one.…”

“…Detailed accounts of system-specific solutions must be sought in system documentation and manuals, but for several systems, overviews have been published in the open literature. The photon beam modeling process in major TPS has been described by Bedford et al (2003) for Pinnacle (Philips Healthcare), Ahnesjö et al (2005) for Oncentra (Elekta Nucletron), and Tillikainen et al (2007) for Eclipse (Varian Medical Systems). The technical details and workflow differ, but the overall approach is similar in which a limited set of free parameters for multisource models are optimized to minimize differences between measured and calculated dose distributions for a limited set of fields in a water phantoms.…”

Fast dose algorithm for generation of dose coverage probability for robustness analysis of fractionated radiotherapy

“…Any suitable, accurate dose engine that can deliver the primary and scatter dose can be used to calculate p ∞ and s ∞ . In this work we use the pencil kernel (Ahnesjö et al 1992) or the collapsed cone (Ahnesjö 1989) algorithm implemented in the TPS (Oncentra® 4.3) (Ahnesjö et al 2005) but e.g. a Monte Carlo dose engine could be used as well.…”