---
language: en
tags:
- ecg
- cardiology
- medical
- pediatric
- time-series
- multi-label-classification
- tensorflow
- cnn
datasets:
- Neural-Network-Project/ECG-database
metrics:
- f1
- auc
- precision
- recall
library_name: tensorflow
---

# ECG Disease Classifier - 19 Cardiac Conditions

Multi-label classification model for detecting 19 cardiac conditions from pediatric ECG signals.

## Model Description

Enhanced 1D CNN with Squeeze-Excitation blocks and temporal attention for variable-length ECG classification.

**Architecture:** 64→128→256→512 filters with residual connections
**Training:** Focal loss for class imbalance
**Input:** Variable-length 12-lead ECG (5-120 seconds at 500 Hz)

## Disease Classes

1. Fulminant/Viral Myocarditis
2. Acute Myocarditis
3. Myocarditis Unspecified
4. Dilated Cardiomyopathy
5. Hypertrophic Cardiomyopathy
6. Cardiomyopathy Unspecified
7. Noncompaction Ventricular Myocardium
8. Kawasaki Disease
9. Ventricular Septal Defect
10. Atrial Septal Defect
11. Atrioventricular Septal Defect
12. Tetralogy of Fallot
13. Pulmonary Valve Stenosis
14. Patent Ductus Arteriosus
15. Pulmonary Artery Stenosis
16. Pulmonary Valve Regurgitation
17. Mitral Valve Insufficiency
18. Congenital Heart Malformation
19. Healthy

## Intended Use

⚠️ **Research and educational purposes only** - NOT for clinical diagnosis

## Training Details

- Batch Size: 128
- Epochs: 17
- Loss: Focal Loss (α=0.25, γ=2.0)
- Optimizer: Adam (lr=0.0002)

## Citation

```bibtex
@misc{ecg-classifier-2025,
  author = {Neural-Network-Project},
  title = {ECG Disease Classifier},
  year = {2025},
  publisher = {Hugging Face},
  url = {https://huggingface.co/Neural-Network-Project/ECG-Disease-Classifier}
}
```