HealEdge

Inspiration

Modern healthcare is rapidly adopting AI for disease diagnosis, but two major barriers remain. First, medical data is highly sensitive, and regulations such as GDPR and HIPAA prevent hospitals from freely sharing patient information. Second, diagnostic outcomes can vary across institutions due to differences in experience and clinical interpretation.

We were inspired to bridge this gap — enabling collaborative medical intelligence without compromising patient privacy or trust.

---

What it does

HealEdge is a privacy-first Edge-AI diagnostic platform that enables hospitals to collaboratively train AI models without sharing raw patient data.

The platform:

• Trains AI models locally on hospital edge devices
  
• Shares only model parameters (weights), not medical images or PHI
  
• Uses federated / swarm learning to improve a global model
  
• Integrates a blockchain-based trust layer to verify model updates
  

This ensures secure collaboration while keeping patient data on-premise.

---

How we built it

HealEdge consists of three main components:

1. Edge AI Layer
   AI models are deployed locally within each hospital. All training and inference occur on institutional data, ensuring privacy.

2. Federated / Swarm Learning Framework
   Instead of sharing datasets, hospitals exchange model weights. These updates are aggregated to improve a shared global model.

3. Blockchain Trust Layer
   Cryptographic hashes of model updates are recorded on a blockchain. This provides:

   • Tamper-proof logging
     
   • Verifiable training contributions
     
   • Transparent and immutable update history
     

To demonstrate the system, we implemented a Diabetic Retinopathy detection model that classifies retinal images into:

• No DR
  
• Mild DR
  
• Moderate DR
  
• Severe DR
  

---

Challenges we ran into

• Securely aggregating model weights without exposing sensitive information
  
• Coordinating decentralized training rounds
  
• Integrating blockchain without storing any medical data
  
• Ensuring performance efficiency on edge devices
  
• Balancing technical depth with usability in the UI
  

These challenges required careful system design to preserve both privacy and performance.

---

Accomplishments that we're proud of

• Building a functional federated learning prototype
  
• Deploying AI inference locally on edge infrastructure
  
• Successfully integrating blockchain for secure verification
  
• Developing a full end-to-end demo for diabetic retinopathy detection
  
• Creating a scalable architecture adaptable to other medical imaging domains
  

We are especially proud of demonstrating collaborative AI without raw data exchange.

---

What we learned

Through this project, we learned that:

• Privacy and trust are foundational in healthcare AI
  
• Decentralized AI systems require strong coordination mechanisms
  
• Blockchain can enhance trust without compromising data privacy
  
• Regulatory awareness must guide technical architecture
  
• Usability is as important as technical sophistication
  

---

What's next for HealEdge

HealEdge is designed to be modality-agnostic and scalable. Next steps include:

• Expanding to X-ray, MRI, CT, and dermatology imaging
  
• Enhancing differential privacy techniques
  
• Optimizing blockchain efficiency for larger networks
  
• Conducting pilot testing with healthcare institutions
  
• Benchmarking performance across multi-hospital environments
  

Our long-term vision is to build a global, privacy-preserving AI ecosystem for healthcare — enabling collaborative intelligence without compromising trust.

Built With

• blockchain
• cnn
• fast-api
• react.js
• vercel