Connectivity Assessment and Training: A Partial Directed Coherence Approach

Abstract

Background. The multivariate autoregressive (MVAR) method to generate a linear model of multichannel signal processes has been employed in many fields but not applied to the assessment of quantitative electroencephalographic (QEEG) connectivity neurofeedback. A measure known as Partial Directed Coherence (PDC) derived in the MVAR framework can offer insensitivity to volume conduction and ability to provide information relating to the direction of information flow between electrode locations, as a function of frequency during QEEG assessment and neurofeedback. Method. This article outlines a variety of reasons why PDC and other related metrics could play a more fundamental role in elucidating the causal relationships underlying EEG connectivity than can be provided though a multivariate analysis of coherence alone. Results. Real-time PDC neurofeedback implementation issues are discussed, technical challenges are outlined, and research questions are proposed. Conclusion. MVAR-based methods are an additional means of relating global to local EEG activity as well as helping to bridge QEEG assessment and neurofeedback protocol generation and treatment.