---
title: Real Perfusion Monitoring System - Full Simulation
emoji: 🏥
colorFrom: blue
colorTo: purple
sdk: gradio
sdk_version: 4.44.0
app_file: app.py
pinned: false
license: mit
---

# 🏥 Real Perfusion Monitoring System - Full Simulation

**Complete DQN-powered perfusion monitoring with trained models and real simulation physics.**

## 🚀 Features

- **Real DQN Agents**: Uses actual trained Deep Q-Network models
- **Full Simulation Physics**: Complete perfusion system modeling
- **Real-Time Monitoring**: Live parameter tracking with safety zones
- **Professional Interface**: Medical-grade dashboard design
- **Dual Scenarios**: Eye perfusion and VCA perfusion support

## 🧠 What Makes This Special

### Real AI Models
- Trained on thousands of simulation episodes
- Learned optimal perfusion control strategies
- Validated performance on 24-hour scenarios

### Complete Physics
- Multi-organ system interactions
- Realistic parameter evolution
- Critical event detection
- Professional medical parameter ranges

### Smart Safety System
- 🟢 **Safe Zone**: Optimal parameter ranges
- 🟡 **Warning Zone**: Requires attention
- 🔴 **Critical Zone**: Immediate intervention needed

## 📊 Parameters Monitored

1. **Temperature** (35-40°C): Core/perfusate temperature control
2. **VR (Vascular Resistance)**: Blood vessel resistance management  
3. **pH** (7.0-7.8): Acid-base balance maintenance
4. **pvO2** (50-700): Partial pressure of oxygen
5. **Glucose** (3-20 mM): Blood sugar regulation
6. **Insulin** (5-80 mU): Hormone control

## 🎯 How to Use

1. **Select Scenario**: Choose Eye or VCA perfusion
2. **Start Evaluation**: Watch real AI make decisions
3. **Monitor Progress**: Follow 24-hour simulation
4. **Observe AI Strategy**: See intelligent parameter control

## 🏆 Applications

- **Medical Device Control**: Automated perfusion systems
- **AI Research**: Reinforcement learning in healthcare
- **Clinical Training**: Understanding AI-assisted procedures
- **Safety Validation**: Testing AI decision-making

---

*Powered by PyTorch, Gradio, and advanced deep reinforcement learning.*