Biophysically Principled Computational Neural Network Modeling of Magneto-/Electro-Encephalography Measured Human Brain Oscillations

Brain rhythms are the most prominent signal measured noninvasively in humans with magneto-/electro-encephalography (MEG/EEG). MEG/EEG measured rhythms have been shown to be functionally relevant and signature changes are used as markers of disease states. Despite the importance of understanding the underlying neural mechanisms creating these rhythms, relatively little is known about their in vivo origin in humans. There are obvious challenges in linking the extracranially measured signals directly to neural activity with invasive studies in humans, and although animal models are well suited for such studies, the connection to human brain function under cognitively relevant tasks is often lacking. Biophysically principled computational neural modeling provides an attractive means to bridge this critical gap. Here, we describe a method for creating a computational neural model capturing the laminar structure of cortical columns and how this model can be used to make predictions on the cellular and circuit level mechanisms of brain oscillations measured with MEG/EEG. Specifically, we describe how the model can be used to simulate current dipole activity, the common macroscopic signal inferred from MEG/EEG data. We detail the development and application of the model to study the spontaneous somatosensory mu-rhythm, containing mu-alpha (7–14 Hz) and mu-beta (15–29 Hz) components. We describe a novel prediction on the neural origin on the mu-rhythm that accurately reproduces many characteristic features of MEG data and accounts for changes in the rhythm with attention, detection, and healthy aging. While the details of the model are specific to the somatosensory system, the model design and application are based on general principles of cortical circuitry and MEG/EEG physics, and are thus amenable to the study of rhythms in other frequency bands and sensory systems.