Computational tool for multimodal brain imaging

Abstract

After stroke, motor recovery usually occurs with functional changes in the sensorimotor network rather than healing damaged brain areas. Clinical assessments and structural brain imaging techniques (CT/MRI) provide an overview of stroke severity and anatomical damages but cannot reveal the dynamic changes of brain function. Electroencephalography (EEG) as a non-invasive electrophysiological technique provides excellent temporal resolution to capture the dynamics of cortical activity. The state-of-the-art challenge of EEG is its limited spatial resolution. This study aims to introduce a computational tool based on the multimodal brain imaging to improve the spatial resolution of EEG for studying stroke. The 62-channel EEG, structural MRI and diffusion weighted MRI (DWI) were recorded from 10 chronic stroke patients and 8 age-matched healthy controls. During EEG acquisition they received electrical somatosensory stimulation of index finger. We used a multivariate autoregressive model to formulate the directional interactions between EEG sources. The parameters in the model are constrained by the subject-specific anatomical connections inferred from the DWI to improve the spatial resolution of EEG. Moreover, the source dynamics are combined to track the signal propagation in the cortex using sub-space representation. We found the expected source activity at the contralateral sensorimotor areas for the healthy controls. The differences of active sources and their connectivity are shown between stroke patients and control subjects, although the interpretation of these results regarding to neuroplasticity is yet to be performed. This study paves the way towards the precise modelling of functional changes in the brain after stroke using EEG.