An electronic portal image-based monte carlo approach to in-vivo dosimetry for intensity modulated radiation therapy

• Main Author: Atwal, Parmveer Singh
• Format: Others
• Language: English
• Published: University of British Columbia 2009
• Online Access: http://hdl.handle.net/2429/13711

Radical radiotherapy aims to concurrently achieve high tumour control probability and low normal tissue complication probability. Intensity-modulated radiation therapy (IMRT) provides the highly localized radiation dose delivery necessary to reach this goal. As highly conformal techniques become more prevalent, the importance of determining, and accounting for, treatment-planning, patient-setup, and delivery errors, which result in discrepancies between the calculated and actual delivered dose, also increases. Accurate Monte Carlo-based modeling of the equipment overcomes some of these deficiencies. Unfortunately, some sources of delivery errors, such as mis-calibration of the beam-modulating system, cannot be easily incorporated in the model. Use of the amorphous-silicon detector (or EPID, for Electronic Portal Imaging Device), available on many linear accelerators, provides a solution. We hypothesize that non-transit dose images from the EPID provide us with information regarding certain delivery errors. To obtain this information, we first capture non-transit EPID dose images of the treatment field. Next, removal of intra-EPID scatter via iterative Richardson-Lucy deconvolution converts the dose image to a fluence matrix. Projected back to the height of the beam-modulating system, this matrix can be used to modulate the statistical weight of photons in a phase-space file simulating the linear accelerator from the source to this height. The modulated phase-space can be used to run Monte Carlo calculations through simulated phantoms. Assumptions regarding the EPID's electromechanical behaviour, as well as regarding beam divergence, were validated. This method was compared and validated against our centre's treatment planning system, for various configurations of the beam-modulating system, in two non-patient phantoms (water and anthropomorphic). The new procedure matched well with film measurements, consistently providing a higher percentage (~10%-15% higher) of pixels with Gamma-Dose (3mm Distance-To-Agreement, 3% Dose-Difference criteria) less than 1, versus the TPS-based dose distributions. This indicates that the EPID-based fluence is more accurate than the TPS-based fluence. This so-called MCEF (Monte Carlo with EPID-based Fluence) procedure can be extended by utilizing Cone-Beam CT (CBCT) to account for any setup errors or physiological changes in the patient. By coupling the EPID-based fluence with CBCT-based phantoms, we believe this method will accurately mimic true 3D in-vivo dosimetry.