AANS Online Scientific Session: Stereotactic & Functional-Deep, Self-Supervised Learning For Patient-Specific Anomaly Detection In Stereoelectroencephalography

Deep, Self-Supervised Learning For Patient-Specific Anomaly Detection In Stereoelectroencephalography

Video Transcription

Video Summary

In this video, Michael Martini presents the work of the team at Mount Sinai on deep self-supervised learning for patient-specific anomaly detection in stereoencephalography. The study focuses on epilepsy patients with medically refractory epilepsy who receive intracranial stereo EEG leads and require neurosurgical intervention for treatment. The team adapted a Modified Long Short-Term Memory Network (LSTM) from NASA's Jet Propulsion Laboratory and used dynamic thresholding to enhance anomaly detection in the EEG data. The methods were applied to a dataset of 14 patients, and the results showed improved performance compared to traditional static thresholding approaches. The study demonstrates the potential of deep learning techniques in detecting seizures and improving patient-specific anomaly detection in stereo EEG data. The credits go to Drs. Ohrman, Costa, Betterson, Mako, Panov, and Ghattan at Mount Sinai who contributed to the project.

Asset Subtitle

Michael Martini

Keywords

deep self-supervised learning

patient-specific anomaly detection

stereoencephalography

Modified Long Short-Term Memory Network

dynamic thresholding