Development and validation of a DICOM-based prediction model for instantaneous dose rate and delivery time in VMAT

This study developed and validated a deterministic, constraint-based model for predicting instantaneous dose rates and total delivery times. A constraint-based Python model extracted treatment plan parameters from DICOM-RT Plan files and simulated the dose rate using three mechanical constraints: multileaf collimator (MLC) velocity, gantry rotation speed, and nominal dose rate, and incorporated geometric beam-off constraints such as avoidance sectors. Validation used high-resolution trajectory logs from a TrueBeam STx across 26 clinical VMAT arcs, including flattened and flattening-filter-free modes. Predicted dose rates agreed well with trajectory logs, with a mean absolute error of 60 MU/min and a root-mean-square error of 140 MU/min for the instantaneous dose rate. Predicted and actual delivery times showed a mean error of - 0.8 ± 0.8 s. This deterministic framework provides an accurate and efficient tool for pre-delivery plan assessment and clinical scheduling, as supported by its high agreement with trajectory log data.