Characterizing the relationship of non-invasive and invasive neural recordings: An information-theoretic analysis of the scalp and stereo electroencephalography

Stereo electroencephalography (sEEG) is a powerful recording technique to measure intracranial electrical activity. Scalp electroencephalography (EEG) offers a noninvasive alternative to measuring neuronal activity at the cost of much lower spatial resolution. Although research has shown that scalp EEG can help localize and image intracranial activity, there has not been a systematic analysis of the relationship between scalp EEG and sEEG. The present work investigates the spatial, temporal, and spectral connection between synchronous recordings of scalp EEG and sEEG, using a suite of tools including Spearman's rank correlation coefficient, mutual information, and directed information to capture any nonlinear relationships. The results confirm an inverse relationship between spatial depth and frequency content between intracranial recordings and scalp EEG. Additionally, this work investigates the directed information between sEEG and EEG to determine optimal temporal windows for comparison. The findings of this work suggest the feasibility of estimating intracranial activity from noninvasive recordings while illustrating the physical challenges inherent to measuring subcortical activity noninvasively.