Automatic Plan Quality Assurance Using Knowledge-Based DVH Prediction for Head and Neck Cancer Patients
W Cao1*, M Gronberg2, A Olanrewaju3, C Cardenas4, B Beadle5, L Court6, (1) University of Texas MD Anderson Cancer Center, Houston, TX, (2) University of Texas MD Anderson Cancer Center, Houston, TX, (3) University of Texas MD Anderson Cancer Center, Houston, TX, (4) University of Texas MD Anderson Cancer Center, Houston, TX, (5) Stanford University, Stanford, CA, (6) University of Texas MD Anderson Cancer Center, Houston, TX
Presentations
PO-GePV-T-193 (Sunday, 7/25/2021) [Eastern Time (GMT-4)]
Purpose: To investigate the use of knowledge-based planning to identify poor quality VMAT plans for head and neck cancer (HNC) patients.
Methods: We used a commercial knowledge-based planning system (RapidPlan, Varian Medical Systems) to create two dose-volume histogram (DVH) prediction models using data from 25 HNC patients. One was trained by clinical plans (manual plans: MP); the other used automatically created plans (AP). 7 normal tissues and 2 automatically-generated standardized planning structures were included. DVHs were then predicted for an evaluation cohort of 25 patients (EC1) and compared with achieved DVHs of MP and AP plans. Next we predicted DVHs for 25 other patients (EC2) on which we intentionally generated plans with suboptimal normal tissue sparing while satisfying dose-volume limits of standard practice. These plans were also reviewed by one radiation oncologist when blinded to DVH predictions. Oncologist reviews on plan quality were analyzed together with the predicted DVHs.
Results: Predicted DVH ranges (upper – lower predictions) were consistently wider for the MP model than the AP model for all structures. The average range of mean dose predictions among all structures was 9.9 Gy (MP model) and 3.4 Gy (AP model) for EC1 patients. RapidPlan models identified 7 MP plans as outliers according to mean dose or D1% on one of normal tissues while none of AP plans was flagged. For EC2 patients, 22 suboptimal plans were identified by prediction. While re-generated AP plans validated that these plans could be improved and pass RapidPlan QA, 19 out 45 structures with predicted poor sparing were also identified by oncologist review for additional planning due to inadequate sparing.
Conclusion: Our study shows that RapidPlan could provide accurate DVH prediction for HN cancer patients. Agreement between prediction and oncologist review also shows promising use of knowledge-based DVH prediction for detecting poor quality plans.
Funding Support, Disclosures, and Conflict of Interest: This work is supported by Varian Medical Systems.
ePosters
Keywords
Not Applicable / None Entered.
Taxonomy
Not Applicable / None Entered.
Contact Email