Click here to

Session: Automation in Treatment Planning [Return to Session]

A Comparison of In-House and Shared Knowledge-Based Planning (KBP) Models for Bilateral Head and Neck Treatment Planning

O Trejo*, H Li, K Guida, The University of Kansas Cancer Center, Kansas City, KS


MO-H345-IePD-F7-3 (Monday, 7/11/2022) 3:45 PM - 4:15 PM [Eastern Time (GMT-4)]

Exhibit Hall | Forum 7

Purpose: Radiotherapy planning can vary considerably, based on planner experience, time constraints and physician preference. KBP can improve planning efficiency by standardizing the treatment planning process. The goal of this study is to compare the dosimetric impact of an in-house KBP model (KBP_IH) to a publicly available (KBP_PA) model for head and neck disease.

Methods: A comparison study was performed using 10 previously treated patients. VMAT plans were optimized using both the KBP_PA and KBP_IH models. The KBP_IH model was generated using over 50 bilateral head and neck patients. Evaluation of the models was done by considering our in-house protocol for PTV coverage and OAR sparring based on NRG HN005 constraints. CI100% and dice similarity coefficients (DSC) were measured for the high risk PTV to evaluate dose conformality. The V10, V20 and V30 were quantified to analyze the extent of low dose spill.

Results: CI100% and DSC remained consistent amongst the KBP_IH and KBP_PA models (<3% difference). Similarly, volume measurements of low dose spill vary by less than 3%, proving both models produce clinically acceptable plans in terms of coverage and conformity. On average, the KBP_IH model improves the submandibular gland mean dose and brainstem D0.03cc by over 5Gy and 4Gy, respectively. Esophagus and lips mean dose and spine cord max dose are spared marginally under 1Gy in favor of the local KBP_IH model. The KBP_PA model spared oral cavity by 2.5Gy.

Conclusion: In-house KBP modeling has demonstrated a reduction in OAR dose while maintaining desired PTV coverage for head and neck cases. By generating in-house KBP models, the radiation oncology team can tailor models based on clinical contouring guidelines, dosimetric goals, and delivery techniques to attain desired radiotherapy plans.


Treatment Planning, Dosimetry Protocols, Dose Volume Histograms


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

Contact Email