Many computational approaches for identifying the epileptogenic zone (EZ) from intracranial EEG (iEEG) recordings focus on narrow frequency bands (e.g., beta, gamma, delta), limiting their effectiveness in the face of patient heterogeneity and dynamic seizure onset patterns. To overcome these limitations, the authors developed FREEZ (Frequency Range Explorer of the EZ), an interpretable supervised machine learning model that integrates spectral power from a broad frequency range (0.5–150 Hz) using a stacked random forest ensemble. FREEZ simultaneously classifies EZ electrodes and predicts post-operative seizure outcomes, incorporating spatiotemporal features from six canonical frequency bands.

To facilitate adoption and reproducibility, FREEZ is available as an open-source R package with multiple user-friendly deployment options. A web application enables interactive exploration of the algorithm’s results without installation. For researchers or clinicians wishing to run FREEZ on their own iEEG datasets (preprocessed through RAVE), a Docker container is provided that encapsulates all dependencies and computing environments needed to execute the full analysis pipeline locally.

Access FREEZ via:

●      Preprint: https://www.biorxiv.org/content/10.1101/2024.05.31.596825v2

●      Zenodo Archive (Software): https://zenodo.org/records/11440404

●      Docker (Run FREEZ locally with demo data or your own data): https://hub.docker.com/repository/docker/karaslab/rave-freez/general

●      Web Application (Interactive Demo): https://rave-freez.azurewebsites.net/launcher/

●      GitHub (Source Code): https://github.com/seanmnoleary/rave-pipelines/tree/main/modules/multitaper_explorer

FREEZ offers an accessible and reproducible platform for improving EZ localization and surgical planning in drug-resistant epilepsy.

Reference

1. O’Leary, S. et al. Integrating Data Across Oscillatory Power Bands Predicts the Epileptogenic Zone: the Frequency Range Explorer Epileptogenic Zone (FREEZ) Identification Algorithm. 2024.05.31.596825 Preprint at https://doi.org/10.1101/2024.05.31.596825 (2024).