Parallel independent component analysis using an optimized neurovascular coupling for concurrent EEG-fMRI sources

The complexity of the human brain and the limitation of any one imaging approach motivates the need for multimodal measurements to better understand cerebral processing. A very natural goal is to integrate electrophysiological and hemodynamic activity. Among them, concurrent EEG-fMRI studies have shown great promise for understanding intrinsic brain properties yet analyzing such data presents a significant methodological challenge. Here, we propose a multivariate parallel ICA decomposition incorporating dynamic neurovascular coupling for concurrent EEG-fMRI recordings. The goal of our algorithm is to fuse multimodal EEG-fMRI information and detect/interpret the relationship between electrophysiological and hemodynamic sources via a temporal neurovascular connection enhancement. We analyze the performance of the algorithm on a valid simulation based on real EEG and fMRI components (sources) from our previous works and a neurovascular coupling built from an extended `balloon model'. The results from our simulations yield an accurate source tracking and linkage for concurrent EEG-fMRI, and provide a novel and efficient way to combine EEG and hemodynamic responses.

I. Introduction

ONE of the essential problems in brain science is how electrical and hemodynamic signals that are acquired by e.g. electroencephalography (EEG) and functional magnetic resonance (fMRI) relate to each other during different states of cerebral activity. Scalp EEG samples the synchronous post-synaptic potentials in the cerebral cortex which lead to neuronal input processing, whereas blood-oxygenation-level dependent (BOLD) fMRI measures a delayed hemodynamic response to neuronal activity. Previous research has shown a linear relationship between local field potentials and multi-unit activity and the BOLD signal [1]. Methodologically, the motivation for combining electroencephalography (EEG) and functional magnetic resonance imaging (fMRI) is obvious. The advantage of EEG is millisecond temporal resolution and the ability to measure neuronal activity directly. In contrast, fMRI has excellent uniform spatial resolution but measures an indirect metabolic correlate of neuronal function - the BOLD signal, over a considerably longer time period of seconds. EEG and fMRI are complementary and thus an approach which combines fMRI and ERP can potentially draw on the strengths of each and provide additional information not afforded by either technique alone. The neurophysiologic and methodological motivations, along with the technical improvements that allow concurrent data acquisition, make combining EEG and fMRI a popular yet challenging ongoing effort that employs a variety of approaches.