Purpose: An in-house lung SBRT knowledge-based planning (KBP) model was previously validated and deployed on a limited basis to support non-coplanar VMAT treatments for centrally-located NSCLC tumors. Increasingly at our institution, lung SBRT is delivered via a simultaneous integrated boost (SIB) approach to escalate dose to hypoxic tumor cores in tumors >5cm. This study demonstrates a practical limitation of a robust KBP model via SIB-SBRT of large tumors.
Methods: Clinically deployed KBP model trained with 86 patients was adapted by reconfiguring DVH-estimate objectives. Five complex validation cases (average tumor volume, 189.7cc) were selected and reoptimized with risk-adapted KBP model. Patients received 40-50Gy to PTV in 5 fractions using 3-6 non-coplanar 6X-FFF beams. An inner GTV (1.0 cm interior to ITV) was escalated to 60Gy for a larger hotspot. Plans were reoptimized with KBP and renormalized for identical PTV coverage including escalated GTV dose. Plans were compared relative to RTOG-0813 criteria including target metrics, OAR dosing and optimization time.
Results: KBPs were statistically similar and RTOG-0813 compliant, but clinically inferior to clinical plans due to lessened DVH-estimation capability with larger PTVs. KBPs were produced on average in 1.8±0.8 hours and 2.4 optimizations with minimal planner burden and modest manual intervention. For similar conformity and intermediate dose-spillage, KBPs demonstrated mean GTV dose escalation of 1.7Gy (p=0.08). OAR dosing differences were statistically insignificant with minimal plan complexity costs. KBPs on average spared maximum bronchial tree dose 0.04Gy (p=0.87) but increased maximum heart dose by 1.0Gy (p=0.42).
Conclusion: KBP is a powerful tool that can assist with planning in lung SBRT. Complex cases with tumors >5 cm may not be ideal for using KBP model despite potential RTOG compliance. For non-urgent scenarios, use of the KBP model should be limited in complex cases. Further development of sub-model may support complex SIB-SBRT with larger tumors.
Lung, Treatment Planning, Treatment Techniques
TH- External Beam- Photons: Treatment planning using machine learning/Knowledge Based Planning/automation