MENU

Click here to

×

Are you sure ?

Yes, do it No, cancel
  1. Home
  2. Meeting Program
  3. All Sessions
  4. All Therapy Scientific Session
  5. An Efficient Workflow for Generation and Refinement of a Knowledge Based Planning Model for Pancreas SBRT Using Automation

An Efficient Workflow for Generation and Refinement of a Knowledge Based Planning Model for Pancreas SBRT Using Automation

T Patton1*, Q Diot1 , B Jones1 , L Olsen2 , Y Vinogradskiy1 , K Goodman1 , T Schefter1 , D Thomas1 , (1) University of Colorado School of Medicine, Aurora, CO, (2) Memorial Hospital, Colorado Springs, CO

Presentations

(Tuesday, 7/16/2019) 1:45 PM - 3:45 PM

Room: Stars at Night Ballroom 1

Purpose: To demonstrate an efficient method to train and validate a pancreas stereotactic body radiotherapy (SBRT) Knowledge Based Planning (KBP) model. Eclipse Scripting API was used to automate plan generation for validation and compare the efficacy of generated versus manual structure priorities.

Methods: Eighty pancreatic VMAT SBRT patients (660cGy x 5fx) were retrospectively analyzed. Sixty plans were used to create an initial KBP model (KBPInitial). Priorities were automatically generated by the DVH Estimation algorithm for organs at risk. An automated planning script was developed to re-plan 20 validation patients not included in the training set. Plan quality was assessed against institutional DVH constraints. To obtain clinically acceptable KBP plans, the priorities in the KBP model were iteratively refined with the aim of meeting DVH constraints for all patients in a final version of the KBP model (KBPFinal).

Results: The automated script re-planned and calculated DVH metrics for the twenty test patients without any human intervention in a total planning time of 5 hours on a single workstation. KBPInitial resulted in plans that met institutional dosimetric constraints in 15/20 patients. Five iterations of the model (requiring a total of 120 automated re-plans) were required in order to increase the number of acceptable plans to 19/20 for KBPFinal. KBPFinal outperformed KBPInitial by reducing the maximum DVH metrics across patients: stomach V15Gy (27.3cc vs. 35.4cc), duodenum V20Gy (13.6cc vs. 15.5cc), small bowel V15Gy (29.3cc vs. 43.2cc), large bowel V15Gy (15.3cc vs. 36.4cc).

Conclusion: A KBP model was developed that produces clinically acceptable plans for pancreas SBRT. We have developed an efficient method to train, refine, and validate a new KBP model using automated planning to reduce the time required to create a clinically usable model. Manual structure priorities arrived at using this iterative approach created more clinically acceptable plans than model-generated priorities.

Keywords

Treatment Planning

Taxonomy

TH- External beam- photons: extracranial stereotactic/SBRT

Contact Email

Additional Links