anesthesim 0.1.0

Gymnasium environments for reinforcement learning in automated anesthesia drug dosing

AnestheSim

Gymnasium environments for reinforcement learning in automated anesthesia drug dosing

---

Three Gymnasium-compatible RL environments for automated anesthesia drug dosing: propofol infusion control via the Marsh three-compartment pharmacokinetic model with Hill pharmacodynamic BIS prediction, remifentanil effect-site concentration targeting via the Minto model, and combined propofol-remifentanil anesthesia management using the Greco synergistic interaction surface. Includes configurable difficulty tiers with patient variability and surgical stimulation events, heuristic TCI (Target-Controlled Infusion) clinical baselines, PPO RL agents, and a benchmark suite across three difficulty levels. 109 tests, MIT licensed.

Installation

pip install anesthesim              # Core (numpy, scipy, gymnasium)
pip install anesthesim[train]       # + SB3, PyTorch for RL training
pip install anesthesim[all]         # Everything

Development install:

git clone https://github.com/HassDhia/anesthesim.git
cd anesthesim
pip install -e ".[all]"

Quick Start

import gymnasium as gym
import anesthesim

env = gym.make("anesthesim/PropofolControl-v0")
obs, info = env.reset(seed=42)
for _ in range(100):
    action = env.action_space.sample()
    obs, reward, terminated, truncated, info = env.step(action)
    if terminated or truncated:
        obs, info = env.reset()
env.close()

Environments

Environment	Task	Observation	Action	Key Challenge
PropofolControl-v0	BIS-guided propofol infusion	Box(6): Cp, Ce, BIS, error, rate, time	Box(1): rate [0,200] mcg/kg/min	PK/PD delay, surgical stimulation
RemifentanilDosing-v0	Effect-site concentration targeting	Box(6): Cp, Ce, target, error, rate, time	Box(1): rate [0,0.5] mcg/kg/min	Rapid onset/offset, changing targets
CombinedAnesthesia-v0	Multi-drug BIS control	Box(8): 2x Cp, 2x Ce, BIS, error, 2x rate	Box(2): propofol + remifentanil rates	Drug synergy exploitation

Architecture

AnestheSim implements established pharmacokinetic/pharmacodynamic models:

• Marsh PK Model (Marsh et al., 1991): Three-compartment model for propofol distribution
• Minto PK Model (Minto et al., 1997): Three-compartment model for remifentanil with effect-site
• Hill PD Model (Schnider et al., 1999): Sigmoid Emax model mapping propofol Ce to BIS
• Greco Interaction Model (Bouillon et al., 2004): Response surface for propofol-remifentanil synergy

Paper

The accompanying paper is available at:

• PDF (GitHub)

Citation

If you use anesthesim in your research, please cite:

@software{dhia2026anesthesim,
  author = {Dhia, Hass},
  title = {AnestheSim: Gymnasium Environments for Reinforcement Learning in Automated Anesthesia Drug Dosing},
  year = {2026},
  publisher = {Smart Technology Investments Research Institute},
  url = {https://github.com/HassDhia/anesthesim}
}

License

MIT License. See LICENSE for details.

Contact

Hass Dhia - Smart Technology Investments Research Institute

• Email: [e-mail]
• Web: smarttechinvest.com/research