HealthProcessAI - Process Mining Framework | SMAILE, Karolinska Institutet

Why Process Mining in Healthcare?

analytics

Data-Driven Insights

Transform electronic health records into actionable insights about patient pathways and clinical processes.

psychology

Clinical Pattern Discovery

Identify patterns in sepsis progression, treatment variations, and patient outcomes using advanced algorithms.

smart_toy

AI-Powered Analysis

Leverage multiple LLMs including Claude, GPT-4, and Gemini to generate clinical reports and recommendations.

code

Python & R Support

Complete implementations in both Python (PM4PY) and R (bupaR) with modular, reusable components.

menu_book

Educational Framework

Comprehensive tutorials, exercises, and real-world examples designed for healthcare professionals.

verified

Production Ready

Includes unit tests, integration tests, and best practices for deploying in clinical environments.

assessment

LLM Report Evaluation

Compare and score reports from different AI models using healthcare-specific evaluation rubrics. View Dashboard →

The Process Mining Pipeline

Data Loading

Load and validate event logs from CSV files or EHR systems

Data Preparation

Clean, transform, and enrich data for process mining

Process Discovery

Apply algorithms to discover actual clinical pathways

AI Analysis

Generate insights using Large Language Models

Report Generation

Create actionable reports for clinical teams

Quick Start Examples

# Install and use HealthProcessAI
pip install healthprocessai

# Load and analyze healthcare data
import healthprocessai as hpai

# Step 1: Load clinical event data
loader = hpai.EventLogLoader("patient_events.csv")
data = loader.load_and_prepare()

# Step 2: Discover healthcare processes
miner = hpai.ProcessMiner()
event_log = miner.create_event_log(data)
dfg, starts, ends = miner.discover_dfg()

# Step 3: Advanced analytics
analyzer = hpai.AdvancedProcessAnalyzer(event_log)
clusters = analyzer.cluster_patient_pathways()

# Step 4: Generate AI-powered insights
ai = hpai.LLMAnalyzer(api_key)
report = ai.generate_clinical_report(results)

print(f"✅ Analysis complete! Found {clusters['n_clusters']} patient patterns")

# Process mining in R with bupaR
library(bupaR)
library(httr)

# Load event log
mydata <- read.csv("sepsis_events.csv")
mydata$timestamp <- as.POSIXct(mydata$timestamp)

# Create event log object
event_log <- mydata %>%
  eventlog(case_id="case",
           activity_id="activity",
           timestamp="timestamp")

# Discover clinical process
process_map(event_log, 
            type_nodes = frequency("relative_case"))

# Generate AI insights with OpenRouter
query_openrouter(model_name, messages)

# Complete pipeline for sepsis analysis
def analyze_sepsis_progression(data_path, api_key):
    """
    Complete pipeline for sepsis progression analysis.
    """
    # Load and prepare clinical data
    loader = EventLogLoader(data_path)
    data = loader.load_data()
    prepared = loader.prepare_data()
    
    # Filter sepsis cases for analysis
    sepsis_cases = loader.filter_by_outcome(sepsis_only=True)
    
    # Discover clinical process
    miner = ProcessMiner()
    event_log = miner.create_event_log(sepsis_cases)
    dfg, starts, ends = miner.discover_dfg()
    variants = miner.discover_variants(top_k=10)
    metrics = miner.calculate_process_metrics()
    
    # AI-powered clinical analysis
    analyzer = LLMAnalyzer(api_key)
    prompt = analyzer.create_clinical_prompt(metrics, "sepsis")
    results = analyzer.analyze_with_multiple_models(prompt)
    
    # Generate clinical report
    report = analyzer.generate_clinical_report(
        metrics, 
        results['claude']['content']
    )
    
    return report

# Execute analysis
report = analyze_sepsis_progression("data.csv", API_KEY)
print("✅ Clinical report generated successfully!")

Interactive Learning Modules

school Fundamentals

Master the basics of process mining in healthcare

check_circleWhat is process mining?
check_circleEvent logs and data structures
check_circleProcess discovery algorithms
check_circleHealthcare applications
check_circleHands-on exercises

biotech Sepsis Analysis

Complete walkthrough of sepsis progression analysis

check_circleUnderstanding sepsis pathways
check_circleData preparation techniques
check_circlePattern discovery methods
check_circleRisk stratification
check_circleClinical recommendations

psychology AI Integration

Enhance analysis with Large Language Models

check_circleOpenRouter API setup
check_circlePrompt engineering
check_circleMulti-model comparison
check_circleReport generation
check_circleBest practices

Clinical Use Cases

Infection Progression

Analyze how infections develop and progress to sepsis, focusing on temperature patterns and early warning signs for timely intervention.

Organ Damage Analysis

Compare process patterns between patients who develop organ damage vs. those who don't, identifying critical divergence points.

III

Glomerular Filtration

Study kidney function progression in sepsis patients, tracking GFR changes and predicting acute kidney injury risk.

Kidney Disease

Examine the relationship between sepsis episodes and chronic kidney disease development over extended time periods.