ABSTRACT
Patterns of increase and decrease in connectivity strength among affected brain regions at specific excitation bands over a (slow) long-range time scale have been associated with preictal periods related to epileptic seizures in existing research. However, the existing solutions analyzed the connectivity patterns among selective regions of interest owing to the otherwise computational inhibition. It is the authors’ argument that several unknown combinations of brain states and frequency bands might be revealed as neuromarkers of a preictal disturbance if a better algorithmic schematic is available. This chapter contributes a critical update and comprehensive seaming of workflows and measures enabling systematic research in the estimation of epileptogenic patterns of slow modulations of sparse causal connectivity. Contemporary methods for the analysis of cortical and subcortical directed and undirected integration have been investigated about their strength and limitations in handling nonlinearity, noise, non-stationarity, high dimensionality, sparsity, directionality, and scalability. A tractable workflow for effective connectivity has been proposed that overcomes the limitations of the traditional methods, namely, the Granger causality model and dynamic causal modeling. Deep learning enables a fast estimation of compressed Koopman eigendecomposition-based encoding of the multichannel observations. Furthermore, the principles of dynamic mode decomposition have been employed for phase-dependence-based causal connectivity estimation.
