ECG Classification with Parallel xLSTM

Deep learning application for classifying cardiac conditions from ECG signals using Parallel xLSTM architecture.

Quick Start

Run Locally

streamlit run app.py

Deploy to Cloud

See START_HERE.md for step-by-step guide

📊 Features

• 🫀 Classifies 5 cardiac conditions
• 📁 Supports .npy, .csv, and WFDB formats
• 🎨 Interactive web interface
• 📈 Real-time probability visualization

🏥 Classifications

1. NORM - Normal ECG
2. MI - Myocardial Infarction
3. STTC - ST/T Change
4. CD - Conduction Disturbance
5. HYP - Hypertrophy

🔧 Technical Stack

• Framework: PyTorch, Streamlit
• Model: Parallel xLSTM (sLSTM + mLSTM)
• Dataset: PTB-XL (22k+ ECG recordings)
• Input: 1000 timesteps × 12 leads @ 100Hz

📦 Installation

pip install -r requirements.txt

🎮 Usage

1. Upload ECG signal file
2. Enter patient metadata (age, sex {0: Male, 1: Female})
3. View predictions and probabilities

📁 Project Structure

.
├── app.py                 # Main Streamlit app
├── model_inference.py     # Model loading & inference
├── utils.py               # Preprocessing utilities
├── requirements.txt       # Dependencies
├── xlstm_100hz_parallel_final.ckpt  # Model weights
└── normalization_params.npz         # Normalization params

👤 Author

Mohamed Ahmed AL Mansoury - 2025

🙏 Acknowledgments

• PTB-XL dataset creators
• xlstm library developers
• Streamlit team