Program Information
Investigation of Various Detector Response Functions and Their Geometry Dependence in a Novel Method to Address Ion Chamber Volume Averaging Effect
B Barraclough1*, J Li2 , S Lebron1, Qiyong Fan2, C Liu2, G Yan2, (1) J. Crayton Pruitt Family Department of Biomedical Engineering, University of Florida, Gainesville, FL, (2) Department of Radiation Oncology, University of Florida, Gainesville, FL
Presentations
SU-C-304-1 (Sunday, July 12, 2015) 1:00 PM - 1:55 PM Room: 304
Purpose: A novel convolution-based approach has been proposed to address ion chamber (IC) volume averaging effect (VAE) for the commissioning of commercial treatment planning systems (TPS). We investigate the use of various convolution kernels and its impact on the accuracy of beam models.
Methods: Our approach simulates the VAE by iteratively convolving the calculated beam profiles with a detector response function (DRF) while optimizing the beam model. At convergence, the convolved profiles match the measured profiles, indicating the calculated profiles match the “true” beam profiles. To validate the approach, beam profiles of an Elekta LINAC were repeatedly collected with ICs of various volumes (CC04, CC13 and SNC 125) to obtain clinically acceptable beam models. The TPS-calculated profiles were convolved externally with the DRF of respective IC. The beam model parameters were reoptimized using Nelder-Mead method by forcing the convolved profiles to match the measured profiles. We evaluated three types of DRFs (Gaussian, Lorentzian, and parabolic) and the impact of kernel dependence on field geometry (depth and field size). The profiles calculated with beam models were compared with SNC EDGE diode-measured profiles.
Results: The method was successfully implemented with Pinnacle Scripting and Matlab. The reoptimization converged in ~10 minutes. For all tested ICs and DRFs, penumbra widths of the TPS-calculated profiles and diode-measured profiles were within 1.0 mm. Gaussian function had the best performance with mean penumbra width difference within 0.5 mm. The use of geometry dependent DRFs showed marginal improvement, reducing the penumbra width differences to less than 0.3 mm. Significant increase in IMRT QA passing rates was achieved with the optimized beam model.
Conclusion: The proposed approach significantly improved the accuracy of the TPS beam model. Gaussian functions as the convolution kernel performed consistently better than Lorentzian and parabolic functions. Further improvement of using geometry-dependent kernels was marginal.
Contact Email: