Interactive chat interface for interpreting pediatric cardiac MRI measurements using evidence-based reference values.
Live Demo: miracle-chat.netlify.app
This web application demonstrates the practical implementation of the MIRACLE API using functional tool calling by LLMs by fetching reference values for pediatric cardiovascular magnetic resonance (CMR) through an intuitive chat interface.
- 🤖 Natural language processing of clinical measurements
- 📊 Real-time z-score and percentile calculations
- 📈 Evidence-based reference values via MIRACLE API
- 🔄 Functional tool calling architecture
- 🌐 Serverless deployment with secure API handling
-
Fork & Clone
git clone https://github.com/drankush/MIRACLE-ChatBot.git cd MIRACLE-ChatBot -
Setup Netlify
- Create a Netlify account
- Connect your GitHub repository
- Configure build settings:
- Publish directory:
public - Functions directory:
netlify/functions
- Publish directory:
-
Environment Variables
- Add in Netlify dashboard:
GROQ_API_KEY=<your-groq-api-key>
- Add in Netlify dashboard:
- GROQ API: Obtain key from groq.com
- MIRACLE API: Public API, no key required
- Atrial Volumes (LAV, RAV)
- Ventricular Volumes (LVEDV, LVESV, RVEDV, RVESV)
- Ventricular Function (LVSV, RVSV, LVEF, RVEF)
- Myocardial Mass (LVM, RVM)
If modifying the LLM:
- Must support OpenAI-compatible function calling
- Recommended: Use models with medical domain knowledge
- Current: GROQ LLaMA 3.3 70B
- Integration with vascular MR parameters
- Support for additional MIRACLE API endpoints
- Enhanced visualization options
- Multi-language support
This tool implements reference values from peer-reviewed publications in pediatric cardiovascular magnetic resonance, making them accessible through a modern interface.
MIT License - see LICENSE
This is a demonstration project. For clinical use, always verify calculations and consult appropriate medical professionals.
Made with ❤️ for the CMR Community