A feasibility study of local adaptation of HFHS Lung SBRT Rapid Plan commercial model.
Objective: To explore the feasibility and optimization effect of modifying the Henry Ford Hospital (HFHS) RapidPlan model for stereotactic body radiation therapy planning based on local requirements. Methods: The following changes were made based on Henry Ford Health System(HFHS) Rapid Plan Lung SBRT model, taking the latest clinical guideline evidence and local clinical practice into account: Internal gross target volume(IGTV) and organ at risk(OAR) structure, lung, were added and set corresponding parameters.The upper value of planning target volume (PTV) was adjusted from 109% to 125%. The original training library was replaced with 73 local historical simultaneous integrated boosting plans, and statistical verification and outlier cleaning of the initial trained model were performed using Model Analytics software. Totally 10 cases not included in the model library were selected for independent verification, and automatic optimization result of the models before and after modifying were compared under the same beam condition. The following dosimetric parameters were compared after target dose normalization: conformal index (CI) of target volume, the mean doses, maximum doses and dose-volume parameters of OARs. Results: The " tail" of the PTV's DVH and the " shoulder" and " tail" of the IGTV's DVH of model M(local) validation plan (M(local)_P) performs higher than the original model HFHS (HFHS_P). The PTV_CI (1.07±0.13) of Mlocal_P were significantly smaller than HFHS_P (1.25±0.24) (Z=-2.497, P<0.05). Except for Heart_D15 cm3 and Heart_Dmax, most of the Mlocal_P dosimetric parameters of OARs were lower than HFHS_P, and the standard deviation was smaller. However, the difference of between two plans was no more than 3.06%. 10 HFHS_P plans don't satisfy dose parameters requirement, two of which PTV_CI values are 1.52 and 1.74, far beyond the clinically acceptable range. Conclusions: Commercial model HFHS could be localized by replacing training library and adjusting parameters. Moreover, plans optimized by the modified model are local clinical acceptable in the aspects of target volume conformity and hotspots, and have a better performance in terms of OAR sparing and plan consistency.