Personalization of hybrid brain models from neuroimaging and electrophysiology data

Personalization is rapidly becoming standard practice in medical diagnosis and treatment. This study is part of an ambitious program towards computational personalization of neuromodulatory interventions in neuropsychiatry. We propose to model the individual human brain as a network of neural masses embedded in a realistic physical matrix capable of representing measurable electrical brain activity. We call this a hybrid brain model (HBM) to highlight that it encodes both biophysical and physiological characteristics of an individual brain. Although the framework is general, we provide here a pipeline for the integration of anatomical, structural and functional connectivity data obtained from magnetic resonance imaging (MRI), diffuse tensor imaging (DTI connectome) and electroencephalography (EEG). We personalize model parameters through a comparison of simulated cortical functional connectivity with functional connectivity profiles derived from cortically-mapped, subject-specific EEG. We show that individual information can be represented in model space through the proper adjustment of two parameters (global coupling strength and conduction velocity), and that the underlying structural information has a strong impact on the functional outcome of the model. These findings provide a proof of concept and open the door for further advances, including the model-driven design of non-invasive brain-stimulation protocols.