SzCORE - Epilepsy Benchmarks

Seizure Detection Challenge (2025)

In partnership with The International Conference on Artificial Intelligence in Epilepsy and Other Neurological Disorders (2025), EPFL, ETH, and partners.

---

Table of Contents

• Background and Impact
• Objective
• Installation
• Usage
• Contributing
• License
• Citations

---

Background and Impact

Epilepsy is the most common chronic brain disease and affects people of all ages. More than 50 million people worldwide have epilepsy. Scalp EEG-based seizure detection algorithms can optimize and facilitate the diagnostic workup performed on people with epilepsy to improve patients’ care and quality of life [1].

EEG-based seizure detection aims to detect the onset and duration of all seizures in an EEG recording. The task has benefited from advances in machine learning. However, a relative scarcity of public datasets and a lack of standardization hinder progress in the field. This likely explains the lack of adoption of state-of-the-art algorithms in clinical practices. Recently, SzCORE has proposed a method to standardize dataset formats, evaluation methodology, and performance metrics.

---

Objective

This challenge aimed to build a seizure detection model that accurately detected the onset and duration of all epileptic seizures in a recording from long-term EEG data collected in an epilepsy monitoring unit.

---

Installation

Prerequisites

• Python 3.9 or higher
• BIDS-formatted EEG dataset

Create a virtual environment

conda create -n eeg python=3.9
conda activate eeg

Install dependencies

pip install -r requirements.txt

---

Usage

EEG Seizure Detection Pipeline

A deep learning-based pipeline for detecting seizures in EEG data using BIDS-formatted datasets.

Data Preparation

Ensure your EEG data is organized in BIDS format. The structure should look like:

BIDS_Siena/
└── BIDS_Siena/
    ├── sub-00/
    │   └── ses-01/
    │       └── eeg/
    │           ├── sub-00_ses-01_task-szMonitoring_run-00_eeg.edf
    │           ├── sub-00_ses-01_task-szMonitoring_run-00_events.tsv
    │           └── ...
    ├── sub-01/
    └── ...

Running the Pipeline

Option 1: Jupyter Notebook (Recommended for Exploration)

1. Start Jupyter Notebook:

jupyter notebook

2. Open seizure_detection.ipynb and run the cells sequentially to:
   • Load and preprocess EEG data
   • Train the seizure detection model
   • Visualize results
   • Evaluate model performance

Option 2: Python Script

1. Update the path to your BIDS dataset in the script:

# In seizure_detection.py
initial_path = Path(r"path/to/your/BIDS_dataset")

2. Run the script:

python seizure_detection.py

3. The script will:
   • Load and process the EEG data
   • Train a CNN model on the data
   • Output metrics and visualizations
   • Save the best model to best_model.pth

Customizing the Pipeline

• Adjust hyperparameters in main():

# Change learning rate
optimizer = torch.optim.Adam(model.parameters(), lr=0.0005)

# Adjust batch size
train_loader = DataLoader(train_dataset, batch_size=64, shuffle=True)

# Change training epochs
n_epochs = 20

• Modify model architecture in SeizureDetectionModel class
• Change data preprocessing in load_single_recording function

Using the Trained Model for Inference

# Load the best model
checkpoint = torch.load('best_model.pth')
model = SeizureDetectionModel(n_channels=19)
model.load_state_dict(checkpoint['model_state_dict'])
model.eval()

# Process new EEG data
# ... [code to load and preprocess new EEG data] ...

# Make predictions
with torch.no_grad():
    predictions = model(new_data)
    predictions = (torch.sigmoid(predictions.squeeze()) > 0.5).float()

---

Contributing

Outline guidelines for contributing to the project.

---

License

This project is licensed under the Open Data Commons Attribution License v1.0.

---

Citations

• Beniczky, S., et al. (2013). Standardized computer‐based organized reporting of EEG: SCORE. Epilepsia, 54(6), 1112-1124.
• Beniczky, S., et al. (2017). Standardized computer-based organized reporting of EEG: SCORE–second version. Clinical Neurophysiology, 128(11), 2334-2346.
• Dan, J., & Detti, P. (2024). BIDS Siena Scalp EEG Database (v1.0.0) [Data set]. EPFL. https://doi.org/10.5281/zenodo.10640762.
• Goldberger, A., Amaral, L., Glass, L., Hausdorff, J., Ivanov, P. C., Mark, R., ... & Stanley, H. E. (2000). PhysioBank, PhysioToolkit, and PhysioNet: Components of a new research resource for complex physiologic signals. Circulation [Online], 101(23), e215–e220.
• Guttag, J. (2010). CHB-MIT Scalp EEG Database (version 1.0.0). PhysioNet. https://doi.org/10.13026/C2K01R.
• Shoeb, A. (2009). Application of Machine Learning to Epileptic Seizure Onset Detection and Treatment. PhD Thesis, Massachusetts Institute of Technology.
• Tal Pal Attia, et al. (in prep). Hierarchical Event Descriptor library schema for clinical EEG data annotation. https://arxiv.org/abs/2310.15173.

In partnership with The International Conference on Artificial Intelligence in Epilepsy and Other Neurological Disorders (2025), EPFL, ETH, and partners.