/RL-Application-in-TPS

Utilizing Reinforcement Learning to make an auto-planning treatment planning system

Primary LanguagePythonMIT LicenseMIT

RL-Application-in-TPS

Utilizing Reinforcement Learning to make an auto-planning treatment planning system

  • This project is intended to make a new toolbox to optimize a rule-based model auto-planning platform on Monaco TPS.
  • To build up a new plan evaluation system for physician as a reference.
  • First to use genetic algorithm to replace the current template modifier. alt text

Based on the current bio-optimization, a function(f) would be developed for updating template.

Currently, f is a rule-based function with the prior knowledge or experience of planners.

https://doi.org/10.1002/acm2.12848

In future, f should be optimized to be more intelligent with evolution algorithm or reinforcement learning.