W. Zhong, S. Kuznetsova, L. Tressel, J. J. Kim, H. Chen, H. S. M. Park, D. J. Carlson, and Z. Chen; Department of Therapeutic Radiology, Yale University School of Medicine, New Haven, CT
Purpose/Objective(s): Knowledge-based planning (KBP) is a valuable tool for generating plans of consistent quality across planners and institutions. It also provides a training tool for planners new to a treatment planning system. This study assessed the feasibility of implementing KBP for a de-novo linear accelerator (LINAC) with integrated positron emission tomography for biology-guided radiotherapy. Materials/
Methods: The focus of this study was lung SBRT plans due to their greater quantity in the small pool of completed image-guided radiotherapy (IGRT) treatments on our PET/CT-LINAC. To reach sufficient statistical strength in the model training, 50 historical clinical plans of right lung SBRT treatments (50 or 55 Gy, 5 fractions) on our C-arm machines were used to train an in-house developed KBP model using the available tool in a commercial treatment planning system (TPS). External validations (open loop) were performed with 9 clinical C-arm plans to fine tune the model. The C-arm backup plans of three right lung SBRT patients on the PET/CT-LINAC were used to assess the performance of the model. The individually derived KBP templates of the three patients were manually converted into the treatment planning system of the PET/CT-LINAC machine as a starting point for plan optimization and dose calculation. The plans generated in this manner (KBP model-informed plans) were compared with their corresponding clinical plans generated in the same TPS of the PET/CT-LINAC in terms of the dose-volume parameters of planning target volumes (PTV) and organs at risk (OAR). Results: The KBP model-informed plans achieved our clinical goals of acceptable ranges within one round of optimization. Compared to their corresponding clinical plans, the KBP plans showed advantages in dose sparing over general OARs, though some OARs resulted in larger maximum doses. The difference of V20Gy in lungs between the KBP-informed and clinical plans was within ±1%. The conformity indices of PTVs in the KBP plans were 0.22±0.20 larger on average. The dose of D2cm was generally higher, which is expected since neither ‘Ring’ structures nor a ‘Body-(PTV+2cm)’ structure was included in the model. Conclusion: This study demonstrated the capability and utility of a KBP tool in handling the geometric and dosimetric variations in plans that were generated and delivered in a de-novo PET/CT-LINAC system. The KBP-derived planning template for lung SBRT cases was able to generate an initial iteration of clinical plans in the TPS of the PET/CT-LINAC that were within clinical guidelines, particularly for OARs. The sub-optimal performance of target conformity and coverage can be improved in the future by including ‘Ring’ structures in the model. The preliminary results suggest a promising potential application of KBP in the PET/CT-LINAC system, as more clinical plans of the system become available. Furthermore, this study is also being extended to other treatment sites with more complicated geometries.
