Click here to


Are you sure ?

Yes, do it No, cancel

Institutional Experience Implementing a Knowledge-Based Planning Model for Stereotactic Lung Planning

P Irmen*, R Scheuermann , J Marcel , A Tiwari , S Anamalayil , University of Pennsylvania, Philadelphia, PA


(Tuesday, 7/16/2019) 3:45 PM - 4:15 PM

Room: Exhibit Hall | Forum 7

Purpose: To evaluate impact on clinical workflow and plan quality of a knowledge-based model for lung stereotactic body radiation therapy (SBRT).

Methods: Two models were created using Varian’s RapidPlan™ to represent targets qualifying for the RTOG 0915 or RTOG 0813 protocols. Models were trained using 92 and 91 prior plans, respectively. Each model utilizes a ring technique to plan, with an inner ring (0.3cm/0.8cm inner/outer expansion from PTV) and an outer ring (2cm/5cm inner/outer expansion from PTV). 10 patients, not included in model training, were chosen for each model to evaluate model-generated plan quality and planning efficiency. Test plans were submitted to the model and optimized based on model-generated parameters and allowed a single iteration of optimization without intervention. Plan quality was compared with clinical plans based on constraints given in each RTOG protocol.

Results: After a single iteration of the model generated plan, the average change in dose to all OARs from the clinical plan was less than 1Gy with the exception of a reduction in the bronchus D0.03 by 4.49Gy and overall D2cm by 1.2Gy in the RTOG 0915 model. Conformity and gradient measures remained unchanged between clinical and model-generated plans. Dosimetrically equivalent plans were able to be generated after a single iteration without planner intervention reducing the time dedicated to each plan to 1-2 iterations as opposed to 3-4 iterations for manual planning.

Conclusion: Using a knowledge-based model for lung SBRT plan has been observed to achieve plans with equivalent quality to manually generated plans, but in a fraction of the time. Furthermore, with the prioritized use of two standard dosimetry structures, the model has shown to reduce effort in the planning workflow and further increases the potential plan throughout.


Not Applicable / None Entered.


Not Applicable / None Entered.

Contact Email