Multi-Modal Heart Disease Risk Assessment Platform

AI-powered cardiovascular risk prediction using:

• Multi-modal clinical data (Framingham-based parameters)
• Cardiac imaging analysis (UNet deep learning for medical images)

React frontend with FastAPI backend supporting both assessment methods.

Project Structure

heart-disease-platform/
├── predict_api.py          # FastAPI backend with /api/predict and /api/predict-image endpoints
├── train_model.py          # PyTorch model training script (clinical data)
├── train_image_model.py    # UNet training script (cardiac imaging)
├── run_all.py             # Combined frontend + backend launcher
├── main.py                # Alternative main entry point
├── frontend/              # React + Vite frontend
│   ├── src/
│   │   ├── main.jsx
│   │   ├── HeartDiseaseAssessment.jsx  # Includes image upload tab
│   │   └── index.css
│   ├── package.json
│   ├── vite.config.js     # Proxies /api to backend:8000
│   └── index.html
├── scalers.pkl            # Feature scalers (generated after clinical model training)
├── heart_disease_model_final.pth  # Clinical data model (generated after training)
├── best_cardiac_model.pth        # Cardiac imaging model (generated after image training)
└── framingham.csv         # Training dataset for clinical model

Features

1. Clinical Data Analysis

• Multi-modal neural network for structured health data
• Analyzes demographics, cardiovascular, metabolic, and lab parameters
• Provides modality-specific risk contributions
• Personalized health recommendations

2. Cardiac Imaging Analysis

• UNet-based deep learning for cardiac image segmentation
• Upload cardiac scans (X-ray, CT, MRI, ultrasound)
• Automated structural analysis and risk assessment
• Segmentation-based disease probability calculation

Setup

Prerequisites

• Python 3.8+
• Node.js 16+ and npm
• Miniconda or Anaconda (recommended)

Backend Setup

1. Install Python dependencies:

# Create conda environment (recommended)
conda create -n heart-disease python=3.11
conda activate heart-disease

# Install PyTorch
conda install pytorch torchvision torchaudio cpuonly -c pytorch

# Install other dependencies
pip install fastapi uvicorn pydantic numpy pandas scikit-learn matplotlib seaborn opencv-python pillow python-multipart

2. Train the models (optional):

Clinical Data Model:

python train_model.py

This will generate:

• scalers.pkl - Feature scalers
• heart_disease_model_final.pth - Trained model weights
• best_model.pth - Best checkpoint during training
• training_history.png - Training curves

Cardiac Imaging Model:

python train_image_model.py

This will generate:

• best_cardiac_model.pth - Trained UNet model for cardiac image analysis
• final_cardiac_model.pth - Final model checkpoint

Frontend Setup

1. Install Node dependencies:

cd frontend
npm install
cd ..

Running the Application

Option 1: Combined Server (Recommended)

Run both frontend dev server and backend with one command:

python run_all.py

This will:

• Start the FastAPI backend on http://127.0.0.1:8000
• Start the Vite dev server on http://127.0.0.1:3000
• Automatically open http://127.0.0.1:3000 in your browser
• Proxy /api/* requests to the backend

Option 2: Production Build

Build the frontend and serve everything from FastAPI:

# Build frontend
cd frontend
npm run build
cd ..

# Run combined server (will detect and serve built files)
python run_all.py

The app will be available at http://127.0.0.1:8000

Option 3: Separate Servers

Run backend and frontend separately for development:

Terminal 1 - Backend:

python -m uvicorn predict_api:app --host 127.0.0.1 --port 8000 --reload

Terminal 2 - Frontend:

cd frontend
npm run dev

Open http://127.0.0.1:3000 in your browser.

API Endpoints

Backend (Port 8000)

• GET / - Root endpoint info

• GET /api/health - Health check

  {
    "status": "healthy",
    "model_loaded": true,
    "image_model_loaded": true,
    "timestamp": "2025-12-23T..."
  }

• POST /api/predict - Heart disease risk prediction from clinical data

  Request body:

  {
    "age": 55,
    "sex": "male",
    "bmi": 27,
    "systolic_bp": 130,
    "diastolic_bp": 80,
    "heart_rate": 75,
    "prevalent_hypertension": 0,
    "total_cholesterol": 200,
    "hdl": 50,
    "ldl": 120,
    "triglycerides": 150,
    "fasting_glucose": 95,
    "diabetes": 0,
    "sodium": 140,
    "potassium": 4.2,
    "calcium": 9.5,
    "creatinine": 1.0,
    "egfr": 90,
    "smoking": 0,
    "physical_activity": "moderate",
    "family_history": 0
  }

  Response:

  {
    "probability": 0.234,
    "risk_category": "Low",
    "modalities": {
      "cardiovascular": 0.093,
      "metabolic": 0.082,
      "labs": 0.035,
      "demographics": 0.023
    },
    "feature_importance": [
      {"name": "systolic_bp", "value": 10},
      {"name": "ldl", "value": 20},
      ...
    ],
    "recommendations": [...]
  }

• POST /api/predict-image - Heart disease risk prediction from cardiac imaging

  Request:

  • Multipart form data
  • Field name: file
  • Accepted formats: PNG, JPG, JPEG, etc.

  Example using curl:

  curl -X POST http://127.0.0.1:8000/api/predict-image \
    -F "file=@cardiac_scan.jpg"

  Response:

  {
    "probability": 0.65,
    "risk_probability": 0.65,
    "risk_category": "Medium",
    "predicted_class": 1,
    "num_classes": 2,
    "analysis_type": "cardiac_imaging",
    "recommendations": [
      "Cardiac imaging shows structural analysis completed",
      "MODERATE RISK indicated in cardiac imaging",
      "Schedule follow-up cardiac imaging and consultation",
      "Monitor cardiac symptoms closely"
    ],
    "timestamp": "2025-01-15T...",
    "message": "Image analyzed using 2-class segmentation model"
  }

Testing

Test Backend API

# Health check
curl http://127.0.0.1:8000/api/health

# Prediction (PowerShell)
Invoke-RestMethod -Method POST -Uri http://127.0.0.1:8000/api/predict `
  -ContentType "application/json" `
  -Body '{"age":55,"sex":1,"bmi":27,"systolic_bp":130,"diastolic_bp":80,"heart_rate":75,"prevalent_hypertension":0,"total_cholesterol":200,"hdl":50,"ldl":120,"triglycerides":150,"fasting_glucose":95,"diabetes":0,"sodium":140,"potassium":4.2,"calcium":9.5,"creatinine":1.0,"egfr":90,"smoking":0,"physical_activity":1,"family_history":0}'

Troubleshooting

Frontend not opening automatically

• Ensure you're running python run_all.py from the heart-disease-platform directory
• Check if port 3000 is already in use: netstat -ano | findstr :3000
• Try running backend and frontend separately (Option 3 above)

Model/Scalers not found

• If you haven't trained the model, the API will use a demo fallback (heuristic-based predictions)
• Run python train_model.py to generate the model files
• Files will be saved in the heart-disease-platform directory

CORS / Connection errors

• Ensure backend is running on port 8000
• Check vite.config.js has the proxy configured to http://127.0.0.1:8000
• Verify CORS middleware is enabled in predict_api.py

Port conflicts

To use different ports:

Backend:

python -m uvicorn predict_api:app --host 127.0.0.1 --port 8001

Frontend (update vite.config.js first):

cd frontend
PORT=3001 npm run dev

Development

Frontend Development

The React app uses:

• Vite - Build tool and dev server
• React 18 - UI framework
• Tailwind CSS (CDN) - Styling (currently loaded via CDN in index.html)

To add proper Tailwind support:

cd frontend
npm install -D tailwindcss postcss autoprefixer
npx tailwindcss init -p

Backend Development

The FastAPI backend uses:

• FastAPI - Web framework
• PyTorch - Deep learning framework
• Scikit-learn - Data preprocessing and metrics
• Pydantic - Request/response validation

Production Deployment

1. Build the frontend:

cd frontend
npm run build

2. Serve with Gunicorn + Uvicorn workers:

pip install gunicorn
gunicorn predict_api:app -w 4 -k uvicorn.workers.UvicornWorker -b 0.0.0.0:8000

3. Or use the built-in runner:

python run_all.py