Functional magnetic resonance imaging (fMRI) has identified dysfunctional network dynamics underlying a number of psychopathologies, including post-traumatic stress disorder, depression and schizophrenia. There is tremendous potential for the development of network-based clinical biomarkers to better characterize these disorders. However, to realize this potential requires the ability to track brain networks using a more affordable imaging modality, such as Electroencephalography (EEG). Here we present a novel analysis pipeline capable of tracking brain networks from EEG alone, after training on supervisory signals derived from data simultaneously recorded in EEG and fMRI, while people engaged in various cognitive tasks. EEG-based features were then used to classify three cognitively-relevant brain networks with up to 75\% accuracy. These findings could lead to affordable and non-invasive methods to objectively diagnose brain disorders involving dysfunctional network dynamics, and to track and even predict treatment responses.

No abstract