DeepPulse is an advanced educational and research platform for visualizing and analyzing physiological signals. Combining medical-grade signal processing with AI-powered interpretation using Google Gemini, DeepPulse provides interactive learning and analysis tools for medical students, researchers, and healthcare professionals.

DeepPulse provides professional-grade visualization and AI-assisted analysis of physiological signals including:

• Cardiac signals: ECG (electrocardiogram), heart rate variability
• Respiratory signals: Respiration waveforms, breathing patterns
• Neural signals: EEG (electroencephalogram) and other biosignals
• Other physiological data: Blood pressure, pulse oximetry, and more

1. Medical-grade visualization with TRUE calibration (25mm/s, 10mm/mV for ECG)
2. AI-driven analysis powered by Google Gemini 3.0 Flash
3. Interactive learning modes including hints and diagnostic quizzes
4. Multi-signal support for comprehensive physiological analysis

DeepPulse integrates with PhysioNet, the world's largest repository of open-source physiological data.

Supported Databases:

• PTB Diagnostic ECG Database (ptbdb): 12-lead ECG recordings with clinical diagnoses
• MIT-BIH Arrhythmia Database (mitdb): Benchmark arrhythmia database
• European ST-T Database (edb): ST and T-wave changes
• MIMIC-III Waveform Database: ICU patient monitoring data
• And many more via PhysioNet search

Data Access:

• Uses wfdb (Waveform Database) Python library
• Local caching for fast repeated access
• Automatic signal categorization and metadata extraction

Modern web application built with:

• React 19 with Vite for fast development
• SVG-based rendering for medical-grade signal visualization
• Medical calibration standards: Pixel-perfect 1mm/5mm grids
• Responsive design with horizontal scrolling for long recordings
• Real-time updates and interactive controls

• FastAPI: Modern Python web framework
• RESTful API: Clean separation of concerns
• Google Gemini 3.0 Flash: Multimodal AI for signal interpretation
• wfdb Integration: Direct PhysioNet database access
• Smart caching: Persistent local storage

• backend/app/main.py: FastAPI application and API endpoints
• backend/app/services/ai_service.py: Gemini AI integration and prompt engineering
• backend/app/services/data_service.py: PhysioNet data loading and management
• frontend/src/components/SignalViewer.jsx: Medical-grade SVG signal renderer
• frontend/src/components/Dashboard.jsx: Main application interface

• Python 3.10+
• Node.js 18+ and npm
• Google Gemini API Key (get from Google AI Studio)

1. Clone the repository:

   git clone https://github.com/Graceliying82/DeepPulse.git
   cd DeepPulse

2. Backend Setup:

   cd backend
   pip install -r requirements.txt

3. Configure API Key: Create backend/.env file:

   GOOGLE_API_KEY=your_gemini_api_key_here

4. Frontend Setup:

   cd frontend
   npm install

Start Backend (Terminal 1):

cd backend
uvicorn app.main:app --reload --port 8000

Start Frontend (Terminal 2):

cd frontend
npm run dev

Access the application:

• Frontend: http://localhost:5173
• Backend API: http://localhost:8000
• API Documentation: http://localhost:8000/docs

• TRUE calibration: 25mm/s time scale, 10mm/mV amplitude (ECG standard)
• Professional grid: 1mm minor lines, 5mm major lines matching medical paper
• DPI-normalized: Accurate pixel-to-millimeter conversion (96 DPI standard)
• Calibration pulse: Standard 1mV reference marker
• Multi-lead display: Synchronized scrolling across all channels
• Print-ready: SVG vector format for documentation

• Full Analysis Mode: Comprehensive clinical interpretation
• Hints Mode: Educational guidance using Socratic method
• Quiz Mode: Interactive diagnostic challenges with explanations
• Multi-signal support: Cardiac, respiratory, neural, and other physiological signals
• Context-aware: Considers patient metadata and signal characteristics

• Database discovery: AI-powered PhysioNet database recommendations
• Smart downloads: Automatic categorization and validation
• Local storage: Persistent caching with data/ directory
• Record browser: Easy navigation through downloaded signals
• Batch processing: Download multiple records efficiently

• Switch analysis modes on the fly (Full/Hints/Quiz)
• Progressive hints that don't reveal the diagnosis
• Multiple-choice quizzes with detailed explanations
• Real patient data for authentic learning experiences
• Instant feedback from AI tutor

• SVG patterns for consistent 1mm/5mm medical grid
• Independent minor/major grid layers for precise styling
• Automatic baseline centering and amplitude scaling
• Downsampling for performance (500 points/second target)
• Horizontal scrolling for recordings of any length
• Time markers every 1 second with clear labels

• Time axis: 25 mm/s (1 small square = 0.04s = 40ms)
• Amplitude axis: 10 mm/mV (1 small square = 0.1 mV)
• Grid: Major lines every 5mm (bold), minor lines every 1mm (thin)
• Calibration: 1mV pulse, 0.2s duration (standard reference)

• Client-side SVG rendering (no server round-trips)
• Efficient downsampling prevents DOM bloat
• React memos and optimized re-renders
• Smart caching reduces repeated downloads
• Lazy loading for large datasets

DeepPulse/
├── backend/                 # FastAPI backend
│   ├── app/
│   │   ├── main.py         # API endpoints
│   │   └── services/       # Business logic
│   │       ├── ai_service.py    # Gemini AI integration
│   │       └── data_service.py  # PhysioNet data access
│   ├── tests/              # Backend tests
│   └── requirements.txt    # Python dependencies
│
├── frontend/               # React frontend
│   ├── src/
│   │   ├── components/     # React components
│   │   │   ├── Dashboard.jsx      # Main UI
│   │   │   └── SignalViewer.jsx   # Medical signal renderer
│   │   ├── App.jsx        # Root component
│   │   └── main.jsx       # Entry point
│   └── package.json       # Node dependencies
│
├── data/                   # Local signal storage
│   ├── cardiac/           # ECG and heart data
│   ├── respiratory/       # Breathing signals
│   └── neural/            # EEG and brain data
│
└── README.md              # This file

Backend:

cd backend
pytest

Frontend:

cd frontend
npm run lint

FastAPI provides automatic interactive API docs at:

• Swagger UI: http://localhost:8000/docs
• ReDoc: http://localhost:8000/redoc

Contributions welcome! Areas of interest:

• Additional signal types and visualizations
• Enhanced AI analysis prompts
• Interactive measurement tools
• Export formats (PDF, DICOM, etc.)
• Real-time monitoring capabilities

This project is licensed under the MIT License - see the LICENSE file for details.

DeepPulse is an educational and research tool. AI-generated analyses are based on anonymized PhysioNet data and should be independently verified before any clinical application.

• PhysioNet for providing open-access physiological data
• Google Gemini for multimodal AI capabilities
• wfdb library maintainers for PhysioNet integration
• Medical signal processing community for standards and best practices

Built with ❤️ for medical education and research