SzCORE: Seizure Community Open-Source Research Evaluation Framework for EEG-Based Seizure Detection
Abstract
Ambulatory and long-term EEG monitoring relies on automated seizure detection, yet variability in datasets, evaluation methodologies, and performance metrics makes fair comparison difficult. The SzCORE framework introduces a unified set of recommendations for validating EEG-based seizure detection algorithms, including standardized datasets, file formats, seizure annotations, cross-validation strategies, and performance metrics. It proposes a 10–20 seizure-detection benchmark assembled from public datasets converted to a common format and defines reporting practices to assess clinical significance. The accompanying open-source software library enables rigorous, reproducible evaluation and aims to foster community-driven improvement of seizure detection systems.
Key Highlights
- Provides a standardized evaluation framework and benchmark for EEG-based seizure detection algorithms.
- Harmonizes dataset formats, annotations, cross-validation strategies, and reporting metrics to enable fair comparison.
- Ships with an open-source software stack and 10–20 seizure detection benchmark built from widely used public datasets.
Thorir Mar Ingolfsson
Postdoctoral Researcher
I develop efficient machine learning systems for biomedical wearables that operate under extreme resource constraints. My work bridges foundation models, neural architecture design, and edge deployment to enable real-time biosignal analysis on microwatt-scale devices.