Program Information
A GPU-Accelerated Deterministic Algorithm for Heterogeneity Corrections in Low Dose-Rate Brachytherapy
Y Belanger1*, L Beaulieu2 , P Despres1 , (1) Universite Laval, Quebec, Quebec, (2) Centre hospitalier universitaire de Quebec, Quebec, QC
Presentations
SU-I-GPD-T-41 (Sunday, July 30, 2017) 3:00 PM - 6:00 PM Room: Exhibit Hall
Purpose: To implement a fast and accurate deterministic algorithm that takes into account tissue heterogeneities in dose computation of low dose-rate brachytherapy (LDR).
Methods: The analytical heterogeneity correction proposed by Hueso-González et al. (Phys. Med. Biol., vol 60, 2015) was implemented using a GPU-based ray-tracing algorithm. This method, based on the conservation of energy, yields an equivalent TG-43 radial function (g) that takes into account tissue heterogeneities inside the computed volume. The algorithm was compared to a Geant4 Monte Carlo simulation for a typical prostate LDR case, considered here as the accuracy gold standard.
Results: The algorithm shows maximum discrepancies of 1-2% relative to the Monte Carlo simulation in the first few centimeters around the source and 20% at larger distances or around and behind dense heterogeneities (calcifications). The discrepancy with Monte Carlo is due to the lack of consideration for scattered photons in the formalism. These results are similar to those previously obtained on CPU. However, the GPU-accelerated version of this algorithm executes in 0.1 s per source for a 128x128x128 1 mm³ voxels geometry, 200 times faster than the single core CPU implementation of Hueso-González et al.
Conclusion: In conclusion, the GPU implementation of the proposed algorithm can account for tissue heterogeneity effects in LDRbrachytherapy with acceptable accuracy in a clinically relevant area around a source. For a typical LDR plan containing tens of sources, the algorithm computes the dose in less than 10 seconds. This level of performance let envision the integration of fast and efficient scatter corrections to further increase accuracy away from the sources and close to tissue heterogeneities. This can contribute to filling the gap between TG-43 formalism and Monte Carlo simulations.
Contact Email: