Purpose: In radiation therapy, there are three primary explanations for why a plan may fail quality assurance (QA) testing: the beam model is inaccurate, the plan instructions are corrupted when they are transferred to the treatment machine, or the machine does not deliver the plan as intended by the treatment planning system. This work aims to address the first reason by computing metrics based on the relative fluence through different regions of a c-arm single layer multileaf collimator, and correlate those metrics to QA results.
Methods: A novel computational tool was developed with Python to analyze DICOM RT-Plan files. First, it identifies zones of different transmission levels, as defined by the RayStation treatment planning system (RaySearch Laboratories, Stockholm, Sweden). For seven distinct regions, the corresponding relative fluence fractions were computed. These were averaged over the entire plan, weighted by the monitor units assigned to each control point. This process was completed for 17 plans. The plan-average fluence fractions were then compared to measured gamma pass rates and median dose deviations for each plan. The analysis was repeated for fluence fractions normalized by the fraction of the beam passing through open field.
Results: There were no relationships between the fluence fractions and gamma analysis pass rates. However, those for the leaf tip and the leaf body zones were strongly correlated with median dose deviation. Normalizing fluence fractions by the open field fraction eliminated any observed correlation.
Conclusion: Some metrics developed in this work were found to correlate with common QA results. Therefore, the tool has predictive power in determining QA results and may be useful in the clinic for forecasting which treatment plans are most likely to fail QA, given a specific RayStation beam model.