Author: Ergun E. Ahunbay, Abdul Parchur, Eric S. Paulson, Ilaria Rinaldi, Angelo Schiavi, Li Zhao 👨🔬
Affiliation: Sapienza University of Rome, Department of Radiation Oncology (Maastro), GROW School for Oncology, Maastricht University Medical Centre+, Department of Radiation Oncology, Medical College of Wisconsin 🌍
Purpose: Online adaptive replanning is often necessary in Intensity Modulated Proton Therapy (IMPT) due to the sensitivity of proton dose distributions to daily anatomical changes. A rapid, automated decision tool to trigger the need for daily adaptation is critical. We introduce here a fully automated workflow utilizing FRED, an open source, fast Monte Carlo-based dose calculation engine, to rapidly evaluate the need for online adaptive radiotherapy (OART) in IMPT.
Methods: A workflow was developed in the MIM software platform, running FRED remotely on a PC equipped with a 44.5GB NVIDIA RTX 6000 GPU used for secondary dose verification. IMPT reference plans for brain, HN (head and neck), and pancreas patients were created in RayStation. Daily CT-on-rails (CTOR) images, deep learning-based auto-contours, and clinical spatial registration objects (SRO) from the same patients were used to simulate the following automated OART workflow: i) align daily images to reference images using SROs, ii) generate FRED-compatible RTplan with updated isocenter based on the transformation matrix, iii) prepare input files for FRED, including CT, contour masks, and density overrides for the treatment couch, iv) calculate dose with FRED, v) perform 3D gamma analysis and dosimetric comparisons against FRED reference plan dose for clinical decision-making. The total time required for processing and evaluation was recorded.
Results: The automated OART workflow was successfully implemented with execution times of 58, 125, and 46 seconds, including Fred runtimes of 33, 46, and 16 seconds. Gamma passing rates between reference and daily doses were 82.1%, 84.9% and 99.85% for pancreas, HN and brain patients respectively. Dosimetric criteria was violated in pancreas and HN plans, triggering the need for OART.
Conclusion: This fully automated workflow provides a rapid method for dosimetric evaluation on daily CTOR images with reference IMPT plans, enabling efficient decision-making for OART within clinically acceptable times.