Investigation of a Collapsed Cone Superposition Algorithm for dosimetry in brachytherapy

Sammanfattning: Background & Purpose: The current standard dosimetry in brachytherapy treatment planning, the TG-43 formalism, ignore the presence of non-water media and finite patient dimensions. This can cause clinically relevant errors in dose estimates. To over- come the limitations of the TG-43 formalism, Model-Based Dose Calculation Algorithms (MBDCAs) have evolved. One of the commercial available MBDCAs is the Advanced Collapsed cone Engine (ACE) by Elekta. In ACE, the total dose is divided into three components, the primary, the first-scattered and the multiple-scattered dose, where the two last mentioned are calculated by the means of the Collapsed Cone Algorithm. In this study the performance of ACE has been investigated. The study has been di- vided into 2 parts, where the aim of part 1 was to analyze the relationship between the so called discretization artifacts, caused by the collapsed cone approximation, and the number of dwell positions. The severeness of the artifact is thought to decrease as the number of dwell positions are increased. The second part focus on ACE’s behavior in cortical bone, with the aim to form a hypothesis (explanation and solution) to the previously observed dose underestimation of the dose to bone made by ACE. Materials and Methods: The generic 192Ir source, the Oncentra Brachy (OcB) treatment planning system (TPS) and the Monte Carlo (MC) platform ALGEBRA have been utilized. In the first part of the study, six source configurations, all with a different number of dwell positions, were created and placed in the center of large water phantoms, i.e. under TG-43 conditions in which the TG-43 formalism can be assumed to yield a high accuracy of the estimated dose. The accuracy of ACE has been judged by its’ deviation from TG-43. In the second part of the study, a cubic source configuration, of 27 dwell positions, was positioned at the center of a cubic water phantom. Three cases where constructed, with a small cortical bone heterogeneity positioned at different distances from the source configu- ration. The ACE calculated dose distribution has been divided into its’ three constituents. The accuracy of ACE and TG-43 has been judged by its’ deviation from MC. Results: Part 1 showed that increasing the number of dwell positions does not guar- antee an improved accuracy of ACE. Local dose difference ratios of > 2%, caused by the artifacts, were mainly located outside the 5% isodose line. A general dose underestima- tion was observed in ACE, with an increased magnitude as the dose level decreased. The majority of local dose difference ratios below -4% were found where the multi-resolution voxelization grid of ACE has a voxel size of ≥23 mm3, that is at a distance of ≥8 cm from the closest dwell position when using the ACE standard accuracy level. In part 2, ACE underestimated the dose to cortical bone, with an increased magnitude as the bone was positioned farther away from the source configuration. The TG-43 formalism gave slightly better estimates of the mean dose to bone than ACE, especially at higher dose levels. For a mean dose to the cortical bone heterogeneity equal to 45% of the prescribed dose, TG-43 and ACE underestimated the mean dose with 1% and 4%, respectively. The estimated mean dose to a volume located directly behind the heterogeneity agreed within 1% between ACE and MC. However, an increased amount of positive local dose difference ratios were observed in this volume. Conclusions: Increasing the number of dwell positions cause a ”blurring” effect of the artifact, but may also increase the fluence gradient. In such situations the severeness of the artifact may not be improved. In patient cases the dwell positions are usually added in a more random manner which may favor the ”blurring effect”. The underestimations observed in ACE are thought to be caused by both the multiple- resolution voxelization grid of ACE and the relationship between the dimensions of the phantom in which the multiple-scattered kernel has been generated and the current calcu- lation volume. ACE was unsuccessful to predict the dose to cortical bone, and should hence be used with caution when cortical bone is an organ at risk, as long as the problem remains. The results indicates that the error in ACE is located in the scatter dose calculations and that the heterogeneity cause ACE to displace the dose. The error is thought to be located in the multiple-scattered dose component, which was also shown by Terribilni et al.. A hypothesis is that the problem is caused by the neglected effect of media dependent absorption coefficients in the multiple-scattered dose calculation. A suggested solution, left to be proven, is to use effective attenuation scaling factors.