Click here to

Session: Therapy General ePoster Viewing [Return to Session]

Hippocampal Sparing Whole Brain: RapidPlan Model V2.0 Validation Process

R Foster1*, R Clark2, A Magliari3, (1) Northwestern Medicine Chicago Proton Center, Warrenville, IL, (2) Varian Medical Systems Inc,(3) Varian Medical Systems Inc,


PO-GePV-T-253 (Sunday, 7/10/2022)   [Eastern Time (GMT-4)]

ePoster Forums

Purpose: Validation of HSWBv2 model was required to ensure model produces high-quality and consistent HSWB treatment plans.

Methods: Five datasets, separate from the model training set, were utilized for validation. Model based plans were validated with varying treatment delivery systems (Halcyon, TrueBeam, C-series linacs), beam arrangements (coplanar/non-coplanar, hyperarc, 3-arc/4-arc VMAT), optimization settings (convergence modes), and calculation algorithms (AcurosXB/AAA) to ensure that HSWBv2 produced improved dosimetric results from HSWBv1. Model based plans were generated with no user interaction within the optimizer thus realizing treatment planning efficiency gains.V2.0 plan scorecard was created to provide a single, objective score to quantify relative dosimetric plan quality with a basis of metrics derived from a V1.0 scorecard (NRG-CC001). Changes to V2.0 were based on community feedback from planning results of HSWBv1.0. Refining of plan quality metrics was employed to provide a comprehensive evaluation of HSWBv2.0.

Results: Delivery system validation resulted in higher scores across all platforms for HSWBv2.0 against HSWBv1.0. Halcyon 4-arc coplanar HSWBv1.0/HSWBv2.0 averaged scores of 120.49/133.00 while the TrueBeam 4-arc coplanar averaged scores of 114.11/127.19. Non-coplanar, HyperArc-based planning resulted in scores of 120.36/129.93 where TrueBeam, non-coplanar 4-arc results were 120.05/129.08. 4-arc arrangements did result in an improvement over 3-arc delivery. Convergence mode testing showed that longer optimization time and earlier/multiple returns correlated to higher plan scores. The use of AcurosXB resulted in improved scoring against AAA as well.

Conclusion: HSWBv2 model validation shows improved results over HSWBv1.0. Refined scorecard metrics aided in better quantifying desired improvements according to community feedback from using original HSWBv1.0. Testing HSWBv2 across multiple delivery systems, beam arrangements, optimization processing, and algorithms gives confidence in model quality and future use expectations.


Not Applicable / None Entered.


TH- External Beam- Photons: Treatment planning using machine learning/Knowledge Based Planning/automation

Contact Email