Multi-Agent Reinforcement Learning Meets Leaf Sequencing in Radiotherapy

Authors: Riqiang Gao, Florin-Cristian Ghesu, Simon Arberet, Shahab Basiri, Esa Kuusela, Martin Kraus, Dorin Comaniciu, Ali Kamen

ICML 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We have conducted experiments on four datasets with four metrics and compared our model with a leading optimization sequencer.
Researcher Affiliation Industry 1Digital Technology and Innovation, Siemens Healthineers, Princeton NJ, USA 2Digital Technology and Innovation, Siemens Healthineers, Erlangen, Germany 3Varian Medical Systems, Siemens Healthineers, Helsinki, Finland.
Pseudocode Yes Algorithm 1 RLS training
Open Source Code No The paper does not provide an explicit statement or link for open-source code of the described methodology.
Open Datasets Yes The HN site includes three datasets: 1) the HNd set contains 493 patients after quality assurance used for training and test with a train/validation/test split, 2) two external test sites from TCIA: HNe1 (Bejarano et al., 2018) with 31 patients and HNe2 (Grossberg et al., 2020) with 140 patients after filtering. The Pros site is from a public dataset with access permission requirements, including 555 patients after filtering.
Dataset Splits Yes HNd set contains 493 patients after quality assurance used for training and test with a train/validation/test split, 2) two external test sites from TCIA: HNe1 (Bejarano et al., 2018) with 31 patients and HNe2 (Grossberg et al., 2020) with 140 patients after filtering. The Pros site is from a public dataset with access permission requirements, including 555 patients after filtering.
Hardware Specification Yes GPU type: NVIDIA RTX A4500
Software Dependencies Yes deep learning platform: Py Torch 1.13
Experiment Setup Yes RL iterations: 20000 Update epochs: 2 batch_size: 96... discount factor gamma: 0.99... learning rate: 1e-4... optimizer: Adam W