Authors :
Presenting Author: Hanu Skanda Banappa, MS – Cleveland Clinic
Himanshu Kumar, PhD – Cleveland Clinic
Guhan Seshadri, PhD – Cleveland Clinic
Andreas Alexopoulos, MD, MPH – Cleveland Clinic
David Martinez, MD – Cleveland Clinic
Imad Najm, MD – Cleveland Clinic
Juan C. Bulacio, MD – Cleveland Clinic, Cleveland, United States
Demitre Serletis, MD, PhD – Cleveland Clinic Epilepsy Center, USA
Balu Krishnan, PhD – Cleveland Clinic
Rationale: Epileptic seizures result from dynamic interactions across distributed brain regions. Precise localization of the epileptic foci requires capturing both spatial patterns and rapid temporal changes. While prior studies focused on second-scale dynamics, emerging evidence points to the relevance of millisecond-scale functional microstates in detecting subtle transitions between interictal and ictal states [1]. In this study, we used stereo-EEG (SEEG) to examine peri-ictal functional connectivity microstates and introduced a geometry-based framework to detect subtle network shifts. Our goal was to identify the optimal temporal resolution for capturing these microstates in SEEG data.
Methods:
We analyzed SEEG data from 7 temporal lobe epilepsy patients who underwent successful surgery. Preprocessing followed Krishnan et al. [2], with electrode contacts mapped to a cortical atlas and the highest-entropy contact selected per region. Signals were filtered into delta, theta, alpha, beta and gamma bands, and segmented into 50–500 ms windows (50 ms steps).
Functional connectivity was estimated using Pearson correlation and projected onto eigenspaces; eigenvectors with below-mean eigenvalues defined invariant spaces representing stable network states [3]. Each window was treated as a functional microstate, and transitions between microstates were quantified using principal angles—defined as the smallest angles between subspaces—where smaller angles indicate greater similarity.
The most frequent principal angle index across time defined the representative microstate trajectory. To identify the optimal window size, we analyzed 10-second preictal and ictal segments per seizure, modeling the slope of principal angles over time. The window with the largest slope change was selected, assuming a state transition at seizure onset.
Results: Across all seizures recorded from 7 patients, we calculated seizure durations and selected those exceeding the mean minus one standard deviation to ensure sufficient sampling of peri-ictal and ictal dynamics. This yielded 29 seizures for further analysis. Our analysis showed that a 150 ms window most effectively captured functional microstate transitions during epileptic seizures recorded with SEEG (Fig. 1). At seizure onset, changes in principal angle first appeared in the high-frequency gamma band, followed by changes in the lower-frequency theta band. Toward termination, the principal angle for both bands gradually returned to patterns resembling those seen during preictal periods (Fig. 2).
Conclusions: This study is the first to identify functional microstates in SEEG and define their optimal temporal resolution at 150 ms. Sequential activation of gamma- and theta-band microstates reveals a frequency-specific pattern of network reorganization during seizures. These findings improve our understanding of seizure dynamics and offer a foundation for real-time monitoring, seizure staging, and personalized treatment planning.
References[1] Dev R et al.
IEEE Sens Lett. 2023;7(5):1–4.
[2] Krishnan B et al.
Clin Neurophysiol. 2024;161:80–92.
[3] Maheshwari J et al.
IEEE Trans Neural Syst Rehabil Eng. 2020;28:1742–9.
Funding: Transformative research grant-Cleveland Clinic