HeartWatch AI
AI-Powered 12-Lead ECG Analysis
🚀 Quick Start: Select a sample ECG below and click "Analyze" to see the AI analysis instantly!
Select a Sample ECG
Choose from our collection of real ECG recordings to see the AI analysis in action.
👆 Select a sample and click Analyze to see results.
Upload Your Own ECG Recording
Have your own ECG data? Upload it here for analysis.
Expected Format:
- File type: NumPy array (.npy)
- Shape: (2500, 12) or (12, 2500)
- Leads: I, II, III, aVR, aVL, aVF, V1-V6
- Duration: 10 seconds at 250 Hz
Tip: Use numpy.save('ecg.npy', signal) to create compatible files.
👆 Upload a .npy file and click Analyze to see results.
About HeartWatch AI
HeartWatch AI is a deep learning-based ECG analysis system powered by state-of-the-art models.
🧠 AI Models
| Model | Description |
|---|---|
| 77-Class Diagnosis | Detects 77 different ECG patterns and cardiac conditions |
| LVEF < 40% | Predicts reduced left ventricular ejection fraction |
| LVEF < 50% | Predicts moderately reduced ejection fraction |
| 5-Year AFib Risk | Estimates risk of developing Atrial Fibrillation |
📊 Technical Details
- Architecture: EfficientNetV2 (TorchScript optimized)
- Input: 12-lead ECG, 10 seconds, 250 Hz
- Inference: CPU-optimized for accessibility
- Training Data: Large clinical ECG datasets
⚠️ Important Disclaimer
This is a research demonstration tool.
The predictions provided should NOT be used for clinical decision-making. Always consult qualified healthcare professionals for medical advice and diagnosis.
📚 References
- Models based on the DeepECG project
- Sample ECGs from MIT-BIH Arrhythmia Database (PhysioNet)
Built with Gradio and PyTorch
Based on HeartWise-AI/DeepECG_Docker
Disclaimer: This is a research demo. Not for clinical use.