Enhancing the speed of radiotherapy Monte Carlo dose calculation with applications in dose verification
| dc.contributor.author | Townson, Reid William | |
| dc.contributor.supervisor | Zavgorodni, Sergei | |
| dc.contributor.supervisor | Jirasek, Andrew | |
| dc.date.accessioned | 2015-04-21T21:58:35Z | |
| dc.date.available | 2015-04-21T21:58:35Z | |
| dc.date.copyright | 2015 | en_US |
| dc.date.issued | 2015-04-21 | |
| dc.degree.department | Department of Physics and Astronomy | en_US |
| dc.degree.level | Doctor of Philosophy Ph.D. | en_US |
| dc.description.abstract | Monte Carlo (MC) methods for radiotherapy dose calculation are widely accepted as capable of achieving high accuracy. In particular, MC calculations have been demonstrated to successfully reproduce measured dose distributions in complex situations where alternative dose calculation algorithms failed (for example, regions of charged particle disequilibrium). For this reason, MC methods are likely to play a central role in radiotherapy dose calculations and dose verification in the future. However, clinical implementations of MC calculations have typically been limited due to the high computational demands. In order to improve the feasibility of using MC simulations clinically, the simulation techniques must be made more efficient. This dissertation presents a number of approaches to improve the efficiency of MC dose calculations. One of the most time consuming parts of source modeling is the simulation of the secondary collimators, which absorb particles to define the rectangular boundaries of radiation fields. The approximation of assuming negligible transmission through and scatter from the secondary collimators was evaluated for accuracy and efficiency using both graphics processing unit (GPU)-based and central processing unit (CPU)-based MC approaches. The new dose calculation engine, gDPM, that utilizes GPUs to perform MC simulations was developed to a state where accuracy comparable to conventional MC algorithms was attained. However, in GPU- based dose calculation, source modeling was found to be an efficiency bottleneck. To address this, a sorted phase-space source model was implemented (the phase-space- let, or PSL model), as well as a hybrid source model where a phase-space source was used only for extra-focal radiation and a point source modeled focal source photons. All of these methods produced results comparable with standard CPU-based MC simulations in minutes, rather than hours, of calculation time. While maintaining reasonable accuracy, the hybrid source model increased source generation time by a factor of ~2-5 when compared with the PSL source model. A variance reduction technique known as photon splitting was also implemented into gDPM, to evaluate its effectiveness at reducing simulation times in GPU calculations. Finally, an alternative CPU-based MC dose calculation technique was presented for specific applications in pre-treatment dose verification. The method avoids the requirement of plan-specific MC simulations. Using measurements from an electronic portal imaging device (EPID), pre-calculated MC beamlets in a spherical water phantom were modulated to obtain a dose reconstruction. | en_US |
| dc.description.scholarlevel | Graduate | en_US |
| dc.identifier.uri | http://hdl.handle.net/1828/5983 | |
| dc.language | English | eng |
| dc.language.iso | en | en_US |
| dc.rights | Available to the World Wide Web | en_US |
| dc.rights.uri | http://creativecommons.org/licenses/by/2.5/ca/ | * |
| dc.subject | Medical physics | en_US |
| dc.subject | Radiotherapy | en_US |
| dc.subject | Monte Carlo | en_US |
| dc.title | Enhancing the speed of radiotherapy Monte Carlo dose calculation with applications in dose verification | en_US |
| dc.type | Thesis | en_US |