B1406 - Status: Accepted
Type of publication: Only Abstract
Type of presentation: Invited Talk for an Organized Session
Session: Statistics in neuroscience
Invited by: Jeff Goldsmith
Title: Spectral causality: Exploring lead-lag dependence structure between oscillatory activities in multivariate signals
Authors: Hernando Ombao - King Abdullah University of Science and Technology (KAUST) (Saudi Arabia) [presenting]
Abdulrahman Althobaiti - King Abdbullah University of Science and Technology (Saudi Arabia)
Keywords: high-dimensional data,spectral analysis,time series
Abstract: The motivation comes from the problem of characterizing multi-scale changes in brain signals following an event (e.g., stroke, epileptic seizure). Preliminary analyses of brain signals following stroke show that there are both short-term (immediate) responses to stroke and long term as affected neuronal populations undergo a reorganization in response to an injury. Spectral analysis will be used to study dependence between neuronal populations. Of prime interest is the notion of ``spectral causality'', which is broadly characterized as the extent to which an oscillatory activity in a population of neurons can predict various oscillatory activities in another region at a future time point. Our approach is to extract different oscillatory components via linear filtering and then examine cross-dependence between the filtered signals. The proposed spectral causality approach overcomes the limitations of classical measures such as coherence and partial coherence since these do not indicate directionality. In addition, the proposed approach is superior to partial directed coherence because it is able to precisely capture the time lag in the between oscillatory activity at different regions. Interesting results from exploratory analyses, showing the immediate changes and long-term brain response, will be reported.
Comments: