Abstracts

Rationale: What do intracranial neural recordings in white matter tissue tell us about the underlying function of the brain and epilepsy disease biology? Stereoelectroencephalography (SEEG) implantations capture neural activity in both gray matter (GM) and white matter (WM) tissues and are being increasingly adopted at many institutions. Yet we currently lack an understanding about what WM recordings reveal about seizure dynamics. Any differences in amplitude, frequency, or timing of activity are pertinent to localizing the seizure onset zone, a vital step in achieving eventual seizure freedom in medically refractory epilepsy patients. Here we aim to understand the differences in GM and WM signals, including if WM contacts capture information redundant or complementary to those emanating from nearby GM tissue, and if WM contacts are functionally distinct from GM contacts in relation to seizure activity.

Methods: We employed univariate, bivariate, and multivariate methods to analyze the differences between WM and GM recordings (Fig. 1). Univariate methods included power spectral density and signal-to-noise ratio (SNR); bivariate methods analyzed the statistical relationship between signals from pairs of electrodes in GM and WM (i.e. functional connectivity, including Pearson correlation and coherence), and multivariate methods analyzed the differences in properties from networks composed of the signals each tissue type separately. We replicated our findings using common average referencing and bipolar montaging.

Results: We found differences in power and SNR at seizure onset (Fig. 2). Power was consistently lower in WM tissue (p< 0.05, Wilcoxon signed-rank test, Fig. 2a) for interictal, preictal, ictal and postictal states, and was elevated for both tissue types during the ictal and postictal states compared to preictal states (p< 0.05). Power decreased with distance from GM (Fig. 2b). Compared to the interictal baseline, SNR increased 4 to 8 fold during seizure activity (Fig. 2c). Ictal SNR was lower for WM contacts >6mm from GM tissue, and remained elevated postictally compared to contacts in GM tissue (p< 0.05) supporting the hypothesis that WM contacts measure different activity during seizures than GM contacts. Bivariate analysis demonstrated that functional connectivity of WM tissue was higher than GM tissue at seizure onset. Multivariate analysis showed topologically distinct brain networks represented by each GM and WM contacts, further demonstrating that WM recordings are functionally distinct from GM recordings. Finally, we relate the functional connectivity to the underlying structural connectivity of the brain (Fig. 1e) and show that the neural activity between GM regions is related to underlying structural connectivity, but not in WM.