Program Information
Multi-Institutional Validation of a Knowledge-Based Planning Model for Patients Enrolled On RTOG 0617: Implications for Plan Quality in Cooperative Group Trials
S Holler1*, J Kavanaugh2 , M Mutic2 , E Elder3 , E Ramirez4 , S Mutic2 , J Bradley5 , C Robinson5 , L Olsen6 , (1) Virginia Commonwealth University, Richmond, VA, (2) Washington University in St. Louis, St. Louis, MO, (3) Emory University, Atlanta, Georgia, (4) University of Texas Southwestern Medical Center, Dallas, TX, (5) Washington University in St. Louis School of Medicine, St. Louis, MO, (6) University of Colorado Health, Colorado Springs, CO
Presentations
SU-K-FS1-13 (Sunday, July 30, 2017) 4:00 PM - 6:00 PM Room: Four Seasons 1
Purpose: To validate the process of single institution training of a knowledge-based planning (KBP) model for automated planning and quality control of patients enrolled on a multi-institutional clinical trial.
Methods: A KBP model (RapidPlan, Varian Medical Systems) for Stage III lung cancer was trained using 62clinical IMRT plans from institution A. Three standard beam templates were created for use based on laterality of the tumor (left, right, central). A set of standard optimization objectives was established to systematically achieve the clinical goals outlined by RTOG 0617 when the KBP model was employed. The KBP model was validated on cohort of 23 patients previously enrolled on RTOG 0617 from 3 institutions, including institution A. The plan quality of the generated KBP plans was compared to the submitted clinical trial plans for all clinical goals outlined in RTOG 0617. These metrics included PTV V100, PTV V95, PTV max, spinal cord max, lungs-CTV V20, lungs-CTV mean, heart mean, and esophagus mean dose. A T-test analysis was performed to determine for which metrics the KBP validation plans were statistically better than the clinical plans.
Results: KBP validation plans were on average superior (p<0.05) to the submitted clinical trial plans for PTV V100, PTV max dose, cord max dose, and esophagus mean dose. Validation plans were not statistically different from clinical plans for all other metrics analyzed.
Conclusion: The results indicate that the KBP model achieved plans with improved plan quality to manually developed treatment plans for patients enrolled on RTOG 0617. This demonstrates that a KBP model can be trained with a single institutional training set and used for quality control and automated planning of patients enrolled on multi-institutional clinical trials using similar clinical goals.
Funding Support, Disclosures, and Conflict of Interest: L Olsen: Patent licensed to Varian Medical Systems (US Patent 20120310615A1) S Mutic: Patent licensed to Varian Medical Systems (US Patent 20120310615A1)
Contact Email: