Spatiotemporal dynamics of human high gamma discriminate naturalistic behavioral states

Abdulwahab Alasfour, Paolo Gabriel, Xi Jiang, Isaac Shamie, Lucia Melloni, Thomas Thesen, Patricia Dugan, Daniel Friedman, Werner K. Doyle, Orrin Devinsky, David Gonda, Shifteh Sattar, Sonya Wang, Eric Halgren, Vikash Gilja

Abstract: Author summary Systems neuroscience research generally relies on the experimental trial-based paradigm to reveal how the brain works. While this methodology has proved feasible and fruitful for decades, there is a need to move toward studies that leverage unstructured and naturalistic behaviors to reveal the statistical and dynamic structure of neural activity when it is not constrained to a controlled experimental setting. Here we employ a data-driven approach that shows how various signal features of high gamma band activity recorded from electrocorticography (ECoG) and stereo-electroencephalography (sEEG) can differentiate naturalistic behavioral states. These signal features include both static and dynamic aspects of the spatiotemporal neural activity. Dynamic spatiotemporal patterns extracted from high gamma band activity span multiple time scales, have a global-brain spatial representation, and better fit the data in comparison to non-dynamic approaches. These patterns individually and collectively contain valuable information differentiating between naturalistic behavioral states. This work shows that neural activity in a naturalistic setting has multiple axes of variability that must be taken into consideration in the study of the neural basis of unstructured behaviors.