CTCAE Symptom Standardizer

A tool for standardizing clinical symptom descriptions to CTCAE terminology using RAG (Retrieval Augmented Generation) with InterSystems IRIS Vector Store.

Features

• Processes the official CTCAE v5.0 Excel file from NCI
• Maps free-text symptom descriptions to standardized CTCAE terms
• Determines appropriate symptom grades based on severity descriptions
• Uses vector embeddings stored in InterSystems IRIS for efficient similarity search
• Provides API and command-line interfaces

Quick Start

Docker Setup (Recommended)

1. Clone this repository:

git clone https://github.com/yourusername/ctcae-standardizer.git
cd ctcae-standardizer

2. Run the setup script:

python setup_docker.py

3. Access the services:
   • REST API: http://localhost:8000
   • Jupyter Notebook: http://localhost:8888
   • IRIS Management Portal: http://localhost:5274/csp/sys/UtilHome.csp

Command Line Usage

# Match a symptom to CTCAE terminology
python scripts/run_symptom_matcher.py "severe headache with nausea" --details "occurs daily, pain level 8/10"

API Usage

# Using curl
curl -X POST "http://localhost:8000/match" \
  -H "Content-Type: application/json" \
  -d '{"symptom": "severe headache with nausea", "details": "occurs daily, pain level 8/10"}'

How It Works

1. Data Processing: The system extracts structured data from the CTCAE v5.0 Excel file, including terms, grades, and descriptions.

2. Vector Indexing: CTCAE terms and grade descriptions are embedded and stored in InterSystems IRIS Vector Store for similarity search.

3. Symptom Matching: When a user submits a symptom description:

   • The system retrieves the most similar CTCAE terms using vector search
   • It then finds the most appropriate grade descriptions for the term
   • An LLM makes the final match decision based on retrieved context

4. Result: The user receives the standardized CTCAE term, grade, and description that best matches their symptom.

Project Structure

• scripts/: Command-line tools for data processing and symptom matching
• src/: Core application code and API implementation
• notebooks/: Jupyter notebooks for data exploration and examples
• data/: Storage for CTCAE data and processed files

Requirements

• Docker and Docker Compose
• Python 3.8+
• OpenAI API key
• InterSystems IRIS (included in Docker setup)

License

MIT