product-spec.md

Product Specification

Overview

ECG Axis Lab is a visual-first web application for exploring ECG lead geometry and frontal-plane projections. It is designed for educational purposes, learning, and transparent signal exploration—not for clinical diagnosis.

Core Philosophy

1. Transparent: All mathematics are simple, documented, and inspectable
2. Visual: Focus on geometric relationships and spatial understanding
3. Adaptive: Responds gracefully to available lead data
4. Non-diagnostic: Never classifies, diagnoses, or makes clinical recommendations
5. Learner-focused: Includes simulation and teaching modes

Main Capabilities by Input Level

Level 1: Single Lead

• Waveform viewer
• Single-axis projection
• Optional beat/QRS selection
• No multi-lead geometry

Level 2: Two Frontal Leads

• Both waveforms displayed
• Both lead axes shown
• Limited 2D vector estimate (marked as partial)

Level 3: Three Bipolar Leads (I, II, III)

• Einthoven triangle rendering
• Triple-lead projection relationships
• Frontal vector estimate

Level 4: Six Frontal Leads

• Hexaxial reference system
• Full frontal-plane geometry
• Robust vector estimation
• Projection comparison

Level 5: Full 12-Lead ECG

• All waveforms viewable
• Geometry focus remains on frontal leads
• Precordial context

MVP Feature Set

Essential

• Signal waveform visualization (Plotly)
• CSV/JSON upload and parsing
• Normalized lead handling
• Frontal-plane vector estimation (least-squares)
• Einthoven triangle detection
• Adaptive geometry rendering (SVG)
• Built-in demo data
• Simulation mode
• Data summary panel

Nice to Have (Future)

• Drag-and-drop CSV upload
• Advanced preprocessing pipelines
• Export functionality
• Beat annotation
• Time-series animation
• Multi-patient comparison
• Offline mode

Out of Scope

• PDF/image parsing
• Delineation algorithms
• Diagnostic classifiers
• Clinical validation

User Workflows

Workflow 1: Explore a Demo

1. User lands on homepage
2. Clicks "Enter App"
3. Selects a demo from sidebar
4. Waveforms and geometry displayed
5. Modifies simulation parameters
6. Observes changed projections

Workflow 2: Upload Own Data

1. User uploads CSV or JSON file
2. System auto-detects leads
3. Reconstruction computed
4. Results displayed
5. User can manipulate via simulation mode

Workflow 3: Learn via Simulation

1. User sets vector angle (slider)
2. User sets magnitude (slider)
3. Specifies which leads to show
4. Synthetic ECG generated
5. Geometric projection visualized
6. User understands lead-vector relationship

Non-Goals

• Rhythm classification (normal sinus, atrial fib, etc.)
• Axis categories ("normal", "left deviation", etc.)
• Pathology labels ("posterior MI", "LVH", etc.)
• Machine learning predictions
• Clinical decision support
• Real-time monitoring
• Integration with hospital systems

Success Criteria for MVP

1. Robustness: Handles single lead to 12-lead inputs without crashing
2. Clarity: UI is intuitive; controls are obvious
3. Accuracy: Vector estimates match synthetic ground truth within tolerance
4. Speed: User interactions < 500ms latency
5. Usability: New user can load demo and adjust parameters in < 30 seconds
6. Documentation: Code is well-commented; API is self-documenting

Performance Targets

• Page load: < 2 seconds
• Waveform rendering (500-1000 samples): < 100ms
• Vector reconstruction: < 50ms
• Simulation generation: < 100ms
• JSON parsing: < 50ms

Data Privacy & Security

• No data sent to third parties
• All APIs CORS-enabled for localhost development
• No authentication required (educational tool)
• No data persistence (client-side session only)
• No analytics or tracking

Accessibility

• Clean, readable typography
• High-contrast color scheme
• Keyboard navigation support (future)
• Semantic HTML

Browser Support

• Modern Chrome/Edge (latest)
• Firefox (latest)
• Safari (latest)
• Mobile-responsive design (optional for MVP)