EEG-Derived Neurophenotypes

Abstract

In this chapter, scalp electrophysiological measurements using electroencephalographs or EEGs are examined in light of new developments in complex systems theory. At the most fundamental level, brain function is electrical. The neural network that comprises the brain and peripheral nervous system, along with all the specialized cellular structures for propagating electrical impulses, is designed to support exquisitely fine control over the electrical patterns that determine all thought and behavior. It is not an exaggeration to say that the most fundamental medium of the mind is an electric field. Measurements of brain electrical activity may thus in principle contain information about cognitive phenotypes, if recurring patterns can be found that reliably correlate with them. The brain meets the mathematical definition of a complex dynamical system and EEG measurements are time series or signals produced by local clusters of neurons in this system. This chapter presents a methodology for discovering patterns in the complex systems parameters that can be derived from EEG measurements that is based on machine learning or pattern recognition algorithms. Without attempting to describe these complex features in neurobiological terms, machine learning algorithms can be used to find significant mappings from these measurable features to cognitive phenotypes, thus creating neuropsychiatric biomarkers that may be clinically useful. Experimental results from research to find early biomarkers for autism illustrate the approach.