Dynamic Source Localization and Functional Connectivity Estimation With State-Space Models: Preliminary Feasibility Analysis

Abstract

Dynamic imaging of source and functional connectivity (FC) using electroencephalographic (EEG) signals is essential for understanding the brain and cognition with sufficiently affordable technology to be widely applicable for studying changes associated with healthy ageing and the progression of neuropathology. We present an application for group analysis of recently developed state-space models and algorithms for simultaneously estimating the large-scale EEG inverse and FC problems. This approach reduces estimation bias and facilitates a detailed exploration and investigation of neuronal dynamics compared to current techniques. We present feasibility analyses for simulated and real EEG event-related data. The latter analysis uses a sixteen subjects EEG (Wakeman and Henson’s) database, with signals recorded during a face-processing task. We implement a state-space methodology efficiently using an alternating least squares (ALS) algorithm. This application to neuroimaging analysis may be critical to reliably capture the brain dynamics despite interindividual variability, as demonstrated by the results presented.