Author: Andreea Dimofte, Maksym Sharma, Weibing Yang, Timothy C. Zhu 👨🔬
Affiliation: Department of Radiation Oncology, University of Pennsylvania, University of Pennsylvania 🌍
Purpose:
To assess the effectiveness and dosimetric impact of utilizing a knowledge-based planning model for GI site-specific plans.
Methods:
Six knowledge-based planning models were developed using Varian’s RapidPlan™ for specific gastrointestinal (GI) sites, including Anal, Rectum, Pancreas, Liver, Abdomen, and Pelvis. Each model was trained using data from 70-100 prior patients, incorporating both IMRT and VMAT techniques, with a focus on adhering to institutional constraints for all critical structures. Validation of the models involved assessing PTV coverage (D95% and D2%) and ensuring compliance with dose-volume constraints for all relevant critical structures. Ten patients were excluded from the training dataset and used for model performance validation. Planning efficiency was analyzed by selecting five patients from each site and comparing the number of iterations and active optimization time required for knowledge-based planning versus traditional manual planning.
Results:
For all sites, the knowledge-based models produced clinically acceptable plans that met all planning constraints. PTV D95% improved across all sites, with enhanced homogeneity in target coverage. Dose to critical structures was comparable to manually optimized clinical plans, with significant reductions observed for specific structures. For pancreas and abdomen sites, the spinal cord maximum dose was reduced by up to 7 Gy, while for pelvis plans, bladder doses showed reductions of 2-4 Gy across D50%, D35%, and D5%. Overall, planning efficiency substantially improved without compromising plan quality. RapidPlan required a single optimization iteration (1115±289 seconds) compared to 2-5 iterations for manual planning (2105±997 seconds), resulting in an approximate 50% reduction in active planning time.
Conclusion:
The implementation of RapidPlan knowledge-based planning successfully generated clinically acceptable plans while maintaining the quality observed in prior clinical plans across multiple GI-sites. By delivering comparable plan quality alongside substantially enhanced planning efficiency, the knowledge-based models offer substantial benefits to clinical workflows and help alleviate the strain on clinical resources.