Peter Andersson

Aims. This work aims to develop and investigate the feasibility of a hybrid model combining Monte Carlo (MC) simulations and deep learning (DL) to predict electronic portal imaging device (EPID) images based on MC-generated exit phase space energy fluence maps from dynamic radiotherapy treatment plans. Such predicted images can be used as reference images during in vivo dosimetry. Materials and methods. MC simulations involving a Varian True Beam linear accelerator model were performed using the EGSnrc code package. Two custom variants of the U-Net architecture were employed. The MLC dynamic chair sequence and 17 clinical treatment plans, spanning various cancer types and delivery methods, were used to acquire experimental data, and in the MC simulations. The proposed method was tested through 2D gamma index analysis, comparing predicted and measured EPID images. Results. Results showed gamma passing rates of 38.65%, 74.16% and 96.17% (minimum, median, maximum) for a simpler model variant and 52.72%, 80.61% and 96.80% for the more complex model variant. Conclusion. The study highlights the feasibility of integrating MC and DL methodologies for in vivo dosimetry quality assurance in complex radiotherapy delivery.

Purpose. To investigate the response of the CC13 ionization chamber under non-reference photon beam conditions, focusing on penumbra and build-up regions of static fields and on dynamic intensity-modulated beams. Methods. Measurements were performed in 6 MV 100 × 100, 20 × 100, and 20 × 20 mm² static fields. Monte Carlo calculations were performed for the static fields and for 6 and 15 MV dynamic beam sequences using a Varian multi-leaf collimator. The chamber was modelled using EGSnrc egs_chamber software. Conversion factors were calculated by relating the absorbed dose to air in the chamber air cavity to the absorbed dose to water. Correction and point-dose correction factors were calculated to quantify the conversion factor variations. Results. The correction factors for positions on the beam central axis and at the penumbra centre were 0.98–1.02 for all static fields and depths investigated. The largest corrections were obtained for chamber positions beyond penumbra centre in the off-axis direction. Point-dose correction factors were 0.54–0.71 at 100 mm depth and their magnitude increased with decreasing field size and measurement depth. Factors of 0.99–1.03 were obtained inside and near the integrated penumbra of the dynamic field at 100 mm depth, and of 0.92–0.94 beyond the integrated penumbra centre. The variations in the ionization chamber response across the integrated dynamic penumbra qualitatively followed the behaviour across penumbra of static fields. Conclusions. Without corrections, the CC13 chamber was of limited usefulness for profile measurements in 20-mm-wide fields. However, measurements in dynamic small irregular beam openings resembling the conditions of pre-treatment patient quality assurance were feasible. Uncorrected ionization chamber response could be applied for dose verification at 100 mm depth inside and close to large gradients of dynamically accumulating high- and low-dose regions assuming 3% tolerance between measured and calculated doses.