Evaluation of Varian RapidPlanPT DVH Estimates Against Clinically Delivered Plans for Head and Neck Cancer Treated with Proton Radiation Therapy
Abstract
Purpose
To evaluate the agreement between Varian RapidPlanPT DVH estimates and clinically delivered plans using an independent validation set for head and neck cancer (HNC) treated with proton radiation therapy.
Methods
A cohort of 87 HNC treatment plans used to treat bilateral oropharynx cancer at Mayo Clinic Rochester from 2015-2024 was identified for model development. All plans delivered a prescription dose of 70 Gy using a four-field beam geometry. Eleven additional independent plans were used to determine RapidPlanPT DVH estimation performance. Organ-at-risk DVH metrics identified for analysis included brainstem Dmax and V30%, esophagus Dmean and V55%, larynx Dmean and V50%, mandible Dmax, and spinal cord Dmax. RapidPlanPT model DVH estimation performance was evaluated by determining if the corresponding DVH metrics from clinically delivered plans fell within, above, or below RapidPlanPT estimates.
Results
The majority of clinically delivered plan DVH metrics fell within the RapidPlanPT model estimates. For the validation cases, of the 88 evaluated clinical plan metrics, 59 (67%) fell within RapidPlanPT estimates, 12 (14%) had values greater than upper band estimates, and 17 (19%) had values less than lower band estimates. The esophagus V55% and larynx Dmean had the best agreement between RapidPlanPT estimates and clinical plans with 9 (82%) of the plans being within estimates. The spinal cord Dmax had the least agreement between estimates and clinical plans with 5 (45%) of the plans being within estimates.
Conclusion
RapidPlanPT models have the potential to produce high-quality plans efficiently. In this work, RapidPlanPT DVH estimates were compared with clinically delivered DVH metrics from an independent validation set. Most clinical metrics fell within estimates, suggesting strong model DVH estimation performance. Future work will determine an objective template for planning optimization using RapidPlanPT.