An AI-assisted radiology platform that integrates HOPPR AI for medical imaging analysis, enabling radiologists to annotate DICOM images side-by-side with AI findings and visualize consensus regions.

This project combines:

• Frontend: React + TypeScript + Vite web application for DICOM viewing and annotation
• AI Backend: Flask API server with HOPPR AI integration for chest radiography analysis
• Features: Side-by-side comparison, consensus highlighting, combined overlay, and multi-model AI analysis

• React 18
• TypeScript
• Vite
• shadcn-ui components
• Tailwind CSS
• React Router
• TanStack Query

• Python 3
• Flask (REST API server)
• Flask-CORS
• HOPPR AI SDK (hopprai)
• DICOM image processing
• Cornerstone.js (DICOM rendering)

• Node.js & npm (install with nvm)
• Python 3.x
• HOPPR AI API key (get one from HOPPR)

1. Clone the repository

   git clone <YOUR_GIT_URL>
   cd MTCHacks-2025

2. Set up the frontend

   # Install Node dependencies
   npm install

3. Set up the Python backend

   # Create and activate virtual environment
   python3 -m venv venv
   source venv/bin/activate  # On macOS/Linux
   # or: venv\Scripts\activate  # On Windows
   
   # Install Python dependencies
   pip install hopprai python-dotenv flask flask-cors

4. Configure environment variables

   Create a .env file in the project root:

   HOPPR_API_KEY=your_api_key_here

Start the backend API server (in one terminal):

source venv/bin/activate
python api_server.py

Start the frontend development server (in another terminal):

npm run dev

Test the HOPPR AI integration:

source venv/bin/activate
python test_hoppr.py

MTCHacks-2025/
├── src/                        # React frontend source
│   ├── components/             # UI components
│   │   ├── Demo.tsx           # Main interactive demo
│   │   ├── AnnotatableDicomViewer.tsx  # Annotation interface
│   │   ├── CombinedDicomViewer.tsx     # Combined overlay view
│   │   └── DicomViewerWithOverlay.tsx  # AI findings overlay
│   ├── pages/                  # Page components
│   ├── hooks/                  # Custom React hooks
│   └── lib/                    # Utility functions
├── public/                     # Static assets
├── api_server.py              # Flask REST API server
├── hoppr_model.py             # HOPPR AI integration module
├── pathology_utils.py         # Pathology region mapping
├── test_hoppr.py              # AI model test script
├── main.py                    # Example HOPPR usage
├── dicom_images/              # Sample DICOM files
├── venv/                      # Python virtual environment
└── .env                       # Environment variables (not in git)

• Side-by-Side Comparison: View clinician annotations next to AI findings simultaneously
• Interactive Annotations: Click to add annotations with custom comments on DICOM images
• Combined Overlay: Merge both clinician and AI views with visual consensus highlighting
• Consensus Detection: Automatically detect and highlight regions where AI and clinician agree (green glow)
• Multi-Model AI Analysis: Analyze images across 13+ chest pathology models via HOPPR AI
• Anatomically Accurate Regions: AI findings mapped to medically appropriate locations on X-rays
• Real-time Analysis: Instant AI inference with visual feedback

The hoppr_model.py module provides a simple interface:

from hoppr_model import HopprModelInterface

# Initialize the model
model = HopprModelInterface()

# Analyze a DICOM image
results = model.analyze_dicom_image("path/to/scan.dcm")

# Access AI insights
if results["success"]:
    insights = results["results"]
    print(insights)

Atelectasis, Cardiomegaly, Pneumothorax, Lung Opacity, Pleural Effusion, Consolidation, Infiltration, Pleural Thickening, Aortic Enlargement, Calcification, Pulmonary Fibrosis, ILD, and Normal

Frontend:

npm run dev      # Start dev server
npm run build    # Build for production
npm run lint     # Run linter

Python/AI:

# Always activate venv first
source venv/bin/activate

# Test AI integration
python test_hoppr.py

# Run example
python main.py

Synapse is a hackathon prototype intended for research and educational purposes only. This is not a medical device and is not approved for primary diagnostic use. All images shown are de-identified. Clinical decisions must be made by licensed healthcare professionals. The developers assume no liability for clinical outcomes.

This project was built for MTCHacks 2025.

MIT License - See LICENSE file for details.