Abstracts

Rationale: Patients with medically intractable epilepsy undergo extensive clinical testing to establish candidacy for potential epilepsy surgery. Finding novel mathematical biomarkers underpinning the dynamical organization of brain networks involved in seizure onset and propagation could improve localization of the epileptogenic zone (EZ), leading to better outcomes following epilepsy surgery. Multifractal formalism introduces an invaluable framework for the investigation of nonlinear, scale-invariant features across multiple time scales in non-stationary time-series data^1,2. Interestingly, there have not been many reports about the fractal properties of the brain, although in vitro electrophysiological experiments have previously captured multifractal dynamics in the background noise-like activity recorded from the rodent and human hippocampus³. We sought to explore multifractal features defining temporal correlations in stereoelectroencephalography (sEEG) seizure recordings.


Methods: We collected multi-channel sEEG data from 30 patients with refractory temporal lobe epilepsy treated at Cleveland Clinic, who subsequently underwent standard anteromesial temporal lobectomy (ATL) and achieved at least 1-year of seizure freedom. Multifractal Detrended Fluctuation Analysis (MFDFA) was used to test for fractal scaling in the sEEG signals⁴. Seizure data and electrode contact sites were annotated by clinicians. Stereo-EEG contacts were systematically organized into anatomical groups. Eleven MFDFA-derived features were subjected to uni- and multivariate statistical analysis, controlling for epileptiform states (pre-ictal, ictal and post-ictal) and anatomical structures, using two-way ANOVA with Bonferroni correction.


Results: We identified strong multifractal signatures in human sEEG data, observing spatiotemporal differences across epileptiform states and between different anatomical networks. After analyzing a preliminary cohort of 5 patients, we identified a uniform increase in the degree of multifractality across all contact sites (both resected and spared) in the post-ictal state, suggesting increased network influx. MFDFA-derived metrics (e.g. Holder exponent and multifractal width) identified reduced ictal long-term correlations and multifractal complexity within resected tissue contacts (p< 0.05).


Conclusions: Stereo-EEG seizure recordings contain a multifractal architecture, with spatiotemporal variations observed across the evolution of epileptiform activity. Based on our preliminary results with 5 patients, we now expand the analysis to evaluate MFDFA feature-based variations in a larger cohort of 30 patients undergoing sEEG followed by anteromesial temporal lobectomy. Interpreting and exploring multifractal spectral-based features using machine learning offers an exciting approach to sEEG interpretation and epileptiform network analysis, offering a new tool for surgical decision-making in epilepsy patients.

Funding: None.