A novel closest white-matter-contact-based referencing scheme for stereotactical EEG recordings

SEEG, due to its spatial resolution, offers a unique opportunity for studying neuronal activity in the human brain (Lachaux, 2003). SEEG recordings are typically analyzed with a bipolar referencing scheme to exclude common volume-conducted signals from neighboring electrode contact pairs. Nevertheless, the Local Field Potentials (LFPs) picked up by SEEG might not be that local (Kajikawa, 2011), hence the bipolar referencing may end up discarding also some of true larger-scale neuronal activity. In addition, it is critical to assess whether each contact is located in the white or in the gray matter. The classical approach relays on visual investigation on post-implant scans instead of automatic tools. However, aiming at analyzing large-scale neuronal dynamics, such as functional connectivity, at the group-level would clearly require an automated approach to accomplish this task. Here, we propose to use a closest white-matter-contact as a reference scheme and we propose an index that can be reliably used in automated identification of how likely a given contact is to pick up true neuronal signals. The Gray Matter Proximity Index (GMPI) has simple formulation which requires the knowledge of the position of the Contact and of its nearest points on both Gray and White matters (A). Each contact point is localized in Talairach space while gray and white vertices have been represented on the corresponding meshes resulting from a cortical segmentation algorithm (Freesurfer). With these information, GMPI can be formalized as follows: (C-W)·(G-W)/|G-W|. GMPI values between [0,1] are more likely to indicate contacts that record neuronal activity within cortical structures, while negative values indicate white matter ones. To test the GMPI reliability, we have acquired SEEG data from six subjects. Differences in spectral power densities (PSD) were highly correlated with GMPI, showing that white matter contacts record reduced power compared to cortical contacts (B). Moreover, GMPI is positively correlated (Spearman rank ~0.8) to PSDs differences in frequency bands among contacts assessing that GMPI variations reflect true changes in signal characteristics (C). We also evaluated the relationship between slow-wave amplitudes after iEEG stimulus onset and GMPI (D-E). Also here, the association contact-GMPI reflects true signal differences. GMPI is thus a promising starting point for moving from bipolar to white-matter based referencing schemes in SEEG data analyses.