Room: Karl Dean Ballroom A1
Purpose: 4Ï€ radiotherapy generalizes IMRT to automate beam geometry definition but requires complicated hyperparameter tuning to attain superior plan quality, which can be tedious and inconsistent. In this study, a fully automated 4Ï€ treatment planning was developed using evolving-knowledge-base (EKB) planning guided by dose prediction.
Methods: 20 4Ï€ cases each of lung and head and neck (HN) were included. A statistical voxel dose learning model was initially trained on low quality plans created using default hyperparameter templates. To improve the automated plan quality without being limited by the training data quality, a new 4Ï€ optimization problem was formulated to include a one-sided penalty on the OAR dose deviation from the predicted dose. This directional OAR penalty allows superior OAR sparing. Fast Iterative Shrinkage-Thresholding Algorithm (FISTA) was used to solve the large scale optimization problem. With the new improved plans, new predictions were created to guide the next epoch for a total of 10 epochs. Plan quality was evaluated using a plan quality metric (PQM) points system based on clinical dose constraints and compared with manually-created 4Ï€, clinical VMAT, and automated plans using high quality, manually-created plans as a training set (autHQ).
Results: For the lung cases, the final EKB plans had the highest average PQM and were significantly superior to manual VMAT (+30.5%) and 4Ï€ (+2.60%). The improvements plateaued after the 3rd epoch. The final HN EKB plans and manually-created 4Ï€ plans had comparable PQMs, but had lower PQM compared to autHQ (-3.00% and -4.44%, respectively). The PQM consistently increased up to the 6th epoch. All automated and manually-created 4Ï€ plans had significantly higher PQM than VMAT.
Conclusion: EKB planning is a practical automated planning technique that creates high quality plans more efficiently and consistently than hyperparameter tuning and does not depend on an existing high quality knowledge base.
Funding Support, Disclosures, and Conflict of Interest: NIH U19AI067769 DE-SC0017687 NIH R21CA228160 DE-SC0017057 NIH R44CA183390 NIH R43CA183390 NIH R01CA188300