A generalized a priori dose uncertainty model of IMRT delivery

Multileaf collimator‐based intensity modulated radiation therapy (IMRT) is complex because each intensity modulated field consists of hundreds of subfields, each of which is associated with an intricate interplay of uncertainties. In this study, the authors have revised the previously introduced uncertainty model to provide an a priori accurate prediction of dose uncertainty during treatment planning in IMRT. In the previous model, the dose uncertainties were categorized into space‐oriented dose uncertainty (SOU) and nonspace‐oriented dose uncertainty (NOU). The revised model further divided the uncertainty sources into planning and delivery. SOU and NOU associated with a planning system were defined as inherent dose uncertainty. A convolution method with seven degrees of freedom was also newly applied to generalize the model for practical clinical cases. The model parameters were quantified through a set of measurements, accumulated routine quality assurance (QA) data, and peer‐reviewed publications. The predicted uncertainty maps were compared with dose difference distributions between computations and 108 simple open‐field measurements using a two‐dimensional diode array detector to verify the validity of the model parameters and robustness of the generalized model. To examine the applicability of the model to overall dose uncertainty prediction in IMRT, a retrospective analysis of QA measurements using the diode array detector for 32 clinical IM fields was also performed. A scatter diagram and a correlation coefficient were employed to investigate a correlation of the predicted dose uncertainty distribution with the dose discrepancy distribution between calculation and delivery. In addition, a γ test was performed to correlate failed regions in dose verification with the dose uncertainty map. The quantified model parameters well correlated the predicted dose uncertainty with the probable dose difference between calculations and measurements. It was visually validated with the scatter diagrams. The average correlation coefficient between uncertainty and dose difference of 108 verification measurements was 0.80 ± 0.04 , indicating a strong linear correlation. In the clinical IM field studies, the dose uncertainty map mimicked the probable dose difference distribution. The average correlation coefficient between the overall dose uncertainty and the dose difference of 32 QA measurements (total 13 184 comparison points) was 0.75 ± 0.07 , which also indicated a strong linear correlation between them. The failed regions of the γ test remarkably corresponded to relatively high dose uncertainty. In conclusion, the dose uncertainty map was able to highlight high dose uncertainty regions, where more care should be taken during the treatment plan. The a priori accurate prediction of dose uncertainty in IMRT will significantly improve the treatment plan evaluation process, thus improving the quality of radiation treatments.