General anesthesia, sedation and sleep correspond to distinct physiological states on a spectrum of unconsciousness. Using electroencephalogram (EEG) data from healthy volunteers, we provide the first direct quantitative comparison between brain activity under propofol general anesthesia, sedation with dexmedetomidine, and natural sleep. We focus on the slow oscillation (0.1-1Hz), which has been shown to be a biomarker of depth of unconsciousness. First, we compare slow oscillation power and spatial distribution in EEG sensor space in the different states of unconsciousness. We apply minimum norm estimate (MNE) source localization to estimate slow oscillation activity on the cortical surface based on the EEG recordings. Second, with the goal of examining changes in functional connectivity between resting state networks in different states of unconsciousness, we use canonical coherence analysis to estimate source-space functional connectivity within the slow frequency band. We explore the limitations of the MNE technique for source localization and their impact on the accuracy of canonical coherence estimation. Third, with the goal of improving functional connectivity estimates in source space, we employ the Matsuda-Komaki oscillator model in an expectation maximization (EM) algorithm to improve the accuracy of source localization. We demonstrate in simulation studies that the Matsuda-EM method provides a clear improvement over MNE source localization. Finally, we apply the Matsuda-EM method to localize the slow, alpha and spindle oscillations in the propofol, dexmedetomidine and sleep datasets. We calculate the slow-band canonical coherence between resting state networks in source space based on the Matsuda-EM localization results.

Join Zoom Meeting https://mit.zoom.us/j/94932454648?pwd=NDk0ZkZXRE53RXRvRlUxQ3lGQWx6QT09

Password: 834981