Assessing the Practicality of Using a Single KBP Model for Multiple Linac Vendors
R Douglas*, A Olanrewaju, L Zhang, L Court, University of Texas MD Anderson Cancer Center, Houston, TX
Presentations
SU-F-TRACK 5-2 (Sunday, 7/25/2021) 4:30 PM - 5:30 PM [Eastern Time (GMT-4)]
Purpose: To evaluate the effectiveness of a commercial knowledge-based planning (KBP) tool by applying a single KBP model to three treatment machines with different beam and MLC characteristics (Varian C-arm, Varian Halcyon and Elekta Versa HD).
Methods: A RapidPlan model for head-neck VMAT treatments was developed and finetuned for a Varian 2100 linac. For fifty retrospective head-neck patients, we generated three sets of VMAT plans using an automated treatment planning system (Radiation Planning Assistant). Each plan used a different treatment machine: Varian 2100, Elekta Versa HD, and Varian Halcyon. The Varian 2100 plans were reviewed by radiation oncologists in a separate study. For each patient, we calculated the dosimetric differences (Versa or Halcyon minus Varian 2100) between each plan. We also calculated the percentage of plans that meet established clinical dosimetric recommendations for normal and planning structures.
Results: The mean differences from our Varian 2100 plan for brainstem (Dmax), contralateral parotid (Dmean), ipsilateral parotid (Dmean) and spinal cord (Dmax) were 0.76±1.57Gy, 0.46±0.59Gy, 0.53±0.44Gy, 0.80±1.1Gy, respectively for the Versa plans; and 0.46±3.46Gy, -0.45±0.59Gy, -0.34±0.49Gy, -0.14±1.24Gy, respectively for the Halcyon plans. The mean difference in high dose PTV coverage were less than 3% for prescription dose (V100%) and hotspots (V105% and V110%). 100% of plans for all machines met the clinical dosimetric recommendations for brainstem (Dmax<54Gy), spinal cord (Dmax<45Gy) and high dose PTV (V95%>95%). For contralateral parotid, 41 Varian 2100 plans met recommendation (Dmean<26Gy), compared to 41 Versa and 40 Halcyon plans. For ipsilateral parotid, 23 Varian 2100 plans met recommendation, compared to 20 Versa and 25 Halcyon plans.
Conclusion: We successfully applied a single RapidPlan model to generate comparable plans across three different treatment machines with different beam and MLC characteristics. This supports the use of head-neck RapidPlan models in centralized planning workflows that support clinics with different linac models/vendors.
Handouts
Keywords
Inverse Planning, Treatment Planning, Optimization
Taxonomy
Contact Email