TARDIS: An updated artificial intelligence model to predict linear accelerator machine parameters at treatment delivery

Abstract

We present an open-source artificial intelligence (AI) model that predicts machine parameters at treatment delivery using trajectory files from prior patients. Predictive models for IMRT and VMAT utilized a boosted and bagged tree, respectively, and predicted MLC errors with a high degree of accuracy (IMRT R2=0.99and 0.98 for high and low-resolution respectively; VMAT R2=0.97and 0.90). Residual error for constructed cases was <0.01 mm with R2 ranging from 0.84 – 0.99. The updated AI model is now made available to predict error in machine parameters at treatment delivery for a new DICOM-RT plan.

Keywords

• Artificial intelligence (AI)
• Online adaptive therapy
• IMRT quality assurance (IMRT QA), Python