Click here to

Session: Multi-Disciplinary General ePoster Viewing [Return to Session]

Optimizing the Combination of Radiotherapy with Immunotherapy Using Reinforcement Learning

M Bleile1, 2*, Yixun Xing 2,3, Dan Nguyen 2,5, Hein Nguyen 6, Casey Moore 4, Debabrata Saha 5, Benjamin Chen 5, Michael Story 5, Daniel F. Heitan 2,7, Robert Timmerman5, Steve Jiang 2,5 (1) Department of Statistical Science, Southern Methodist University, Dallas, TX (2) Medical Artificial Intelligence and Automation (MAIA) lab, University of Texas Southwestern Medical Center, Dallas, TX (3) Department of Advanced Data Analytics, University of North Texas, Denton, TXs(4) Department of Pathology, University of Texas Southwestern Medical Center, Dallas, TX (5) Department of Radiation Oncology, University of Texas Southwestern Medical Center, Dallas, TX, (6) Department of Electrical and Computer Engineering, University of Houston, Houston, TX (7) Department of Population & Data Sciences, University of Texas Southwestern Medical Center, Dallas, TX


PO-GePV-M-53 (Sunday, 7/10/2022)   [Eastern Time (GMT-4)]

ePoster Forums

Purpose: Recent pre-clinical studies have indicated that, when combined with immunotherapy, current fractionation schemes for radiotherapy are far from optimal. Deriving the optimal combination of the two modalities, which may depend on individual characteristics, is a clinically important yet unresolved issue. This work aims to develop and test a reinforcement learning (RL) method to identify synergistic, personalized combinations of radiation with immunotherapy.

Methods: We have developed a mechanistic differential equation model based on a series pre-clinical experiments exploring various combinations of radiotherapy and immunotherapy. We used this model to simulate data to train the RL agent. The agent takes a sequence of tumor volumes as well as historical treatment as inputs, and outputs the number of days to wait before applying the next pulse of radiation. The action generated from the RL agent can then be used to drive the mechanistic model and simulate the treatment outcome. We trained models to apply 2, 3, 4, and 5 pulses of radiation, using fixed-spacing treatment and random choice as references.

Results: The difference between the agent and each reference favored of the agent in each case.

Conclusion: We have developed a RL agent to explore the optimal combination of radiotherapy and immunotherapy. While the system needs to be further improved, especially for clinical use, our results are encouraging.


Not Applicable / None Entered.


Not Applicable / None Entered.

Contact Email