Autonomous Treatment Planning
Modern radiation oncology is poised for a significant breakthrough, focusing on delivering personalized care by adapting treatments in real-time to patients' anatomical changes. While advancements in image-guided systems and delivery techniques have paved the way, the full potential of real-time adaptive strategies remains untapped due to limitations in automation and speed within current Treatment Planning Systems (TPS).
Our groundbreaking project aims to overcome these challenges by integrating cutting-edge Deep Learning (DL) and Reinforcement Learning (RL) technologies into the treatment planning process. We have three core objectives:
⦁ Pioneer a Fully Autonomous Treatment Planning Pipeline: Develop the world's first radiation therapy planning system that operates independently of conventional TPS, leveraging advanced DL architectures.
⦁ Explore Reinforcement Learning in Treatment Planning: Investigate how RL can enhance the planning process by optimizing machine parameters in radiotherapy plans
⦁ Establish a Next-Generation Workflow: Create a real-time adaptive treatment workflow that significantly improves the effectiveness and safety of radiation therapy.
Selected publications:
Ultra-fast, one-click radiotherapy treatment planning outside a treatment planning system (2025)
https://doi.org/10.1016/j.phro.2025.100724Generating deliverable DICOM RT treatment plans for prostate VMAT by predicting MLC motion sequences with an encoder-decoder network (2023)
https://doi.org/10.1002/mp.16545Technical Note: Dose prediction for radiation therapy using feature-based losses and One Cycle Learning (2021) https://doi.org/10.1002/mp.14774