healthprocessai

⚡ HealthProcessAI Quick Start Guide

Get Started in 10 Minutes

This guide will have you analyzing healthcare processes with AI insights in under 10 minutes.

Contributors:

🚀 Option 1: Google Colab (Fastest - 2 minutes)

No installation required!

  1. Open our Colab notebook: Open In Colab

  2. Run all cells (Runtime → Run all)

  3. Done! You’ll see:

    • Sample sepsis progression analysis
    • Process visualization
    • AI-powered clinical insights

🐍 Option 2: Local Python Setup (5 minutes)

Prerequisites

Quick Install

# 1. Clone the repository (30 seconds)
git clone https://github.com/ki-smile/HealthProcessAI.git
cd HealthProcessAI

# 2. Install dependencies (2 minutes)
pip install pm4py pandas numpy matplotlib requests

# 3. Run quick example (1 minute)
python examples/quickstart_example.py

What You’ll Get

📊 What This Quick Start Demonstrates

🏥 Healthcare Process Mining in Action

Sample Output:

✓ Loaded 1,000 events from 150 sepsis cases
✓ Discovered 12 distinct patient pathways
✓ Generated process visualization
✓ Most common pathway (34% of patients):
  Admission → Triage → Blood Test → High Temperature → Antibiotics → Recovery

📊 Key Insights:
- Average case duration: 18.5 hours
- Early antibiotic administration reduces complications by 23%
- Temperature monitoring is critical within first 6 hours

Visual Output:

🎯 Quick Example: Sepsis Analysis

Here’s the complete code that runs in our quick start:

# Quick Start Example - Sepsis Process Analysis
from core.step1_data_loader import EventLogLoader
from core.step2_process_mining import ProcessMiner

# 1. Load sample data (15 seconds)
print("🔄 Loading sepsis progression data...")
loader = EventLogLoader("data/sepsisAgregated_Infection.csv")
data = loader.load_data()
prepared_data = loader.prepare_data()

# 2. Get quick statistics
stats = loader.get_statistics()
print(f"📊 Dataset: {stats['num_events']:,} events from {stats['num_cases']} patients")
print(f"📈 Sepsis rate: {stats.get('sepsis_rate', 0):.1%}")

# 3. Discover patient pathways (30 seconds)
print("🔍 Discovering patient pathways...")
miner = ProcessMiner()
event_log = miner.create_event_log(prepared_data)
dfg, starts, ends = miner.discover_dfg()

# 4. Find common pathways
variants = miner.discover_variants(top_k=3)
print("\n🏥 Top 3 Patient Pathways:")
for i, row in variants.iterrows():
    print(f"  {i+1}. {row['percentage']:.1f}% of patients ({row['cases']} cases)")
    pathway = row['variant'].split(' → ')[:4]  # Show first 4 steps
    print(f"     {' → '.join(pathway)}...")

# 5. Generate quick insights
print(f"\n💡 Key Findings:")
print(f"- Found {len(dfg)} unique activity transitions")
print(f"- Most common start: {list(starts.keys())[0]}")
print(f"- Most common end: {list(ends.keys())[0]}")
print(f"- Process coverage: Top 3 pathways cover {variants['percentage'].head(3).sum():.1f}% of cases")

print("\n✅ Quick analysis complete! Check generated files for visualizations.")

🎓 Next Steps (Choose Your Path)

For Healthcare Professionals

👩‍⚕️ → Clinician Tutorial

For Data Scientists/Developers

🐍 → Python Tutorial

For Comprehensive Learning

📚 → Complete Tutorial

🔧 Troubleshooting

Common Issues

1. Import Errors

# Solution: Install missing packages
pip install pm4py pandas numpy

2. Visualization Not Showing

# Solution: Install Graphviz
# Windows: Download from graphviz.org
# Mac: brew install graphviz  
# Linux: apt-get install graphviz

3. Data File Not Found

# Solution: Check file path
import os
print("Current directory:", os.getcwd())
print("Data files:", os.listdir("data/"))

Need Help?

📊 Sample Results You Should See

Process Statistics

Dataset Overview:
- Total events: 15,214
- Unique patients: 1,050
- Unique activities: 16
- Sepsis rate: 31.4%
- Date range: 2020-01-01 to 2023-12-31

Top Patient Pathways

1. 34.2% of patients (359 cases)
   Admission → Triage → Blood Culture → High Temperature → Antibiotics

2. 18.7% of patients (196 cases) 
   Admission → Triage → Lab Test → Normal Temperature → Discharge

3. 12.4% of patients (130 cases)
   Admission → High Temperature → Infection → ICU Transfer → Recovery

Key Process Insights

💡 Clinical Insights Generated:
- Average progression time from admission to sepsis: 14.2 hours
- Early warning indicators: Temperature elevation + Lab abnormalities
- Critical intervention window: First 6 hours post-admission
- Successful treatment pattern: Rapid antibiotic administration

🎯 What You’ve Accomplished

In just 10 minutes, you’ve:

✅ Loaded real healthcare data (sepsis progression events) ✅ Discovered process patterns (patient pathways through care) ✅ Generated visualizations (process flow diagrams) ✅ Extracted insights (clinical patterns and timings) ✅ Learned the framework (ready for advanced analysis)

🚀 Ready for More?

Immediate Next Steps

  1. Try your own data: Replace sample data with your CSV files
  2. Explore parameters: Adjust thresholds and filters
  3. Add AI insights: Set up OpenRouter API for LLM analysis
  4. Generate reports: Export professional analysis reports

Advanced Features to Explore

🏥 Clinical Applications

This quick start demonstrates techniques applicable to:

🎉 Congratulations! You’re now ready to apply process mining to healthcare data.

Choose your next tutorial based on your role and interests:

This quick start guide is part of HealthProcessAI - Developed at SMAILE, Karolinska Institutet

This site is open source. Improve this page.