After stroke, functional recovery may be promoted during the first six months through rehabilitation. Several events are thought to lead to regaining lost functions, among which the remapping of affected limbs to other regions of the cortex is a frequent occurrence. High spatial resolution brain imaging techniques, like magnetic resonance imaging (MRI), may be used to observe the affected areas and assess the severity of the lesions, but are unable to provide much insight on dynamic changes in brain function. An alternative non-invasive technique possessing an excellent temporal resolution to observe transient events is electroencephalography (EEG), a method limited by a low spatial resolution. This study aims to combine 62-channel EEG recordings with anatomical information derived from structural MRI and diffusion-weighted imaging (DWI) to improve the low resolution of EEG. During EEG acquisition, stroke patients (N = 3) and age-matched healthy controls (N = 2) received electrical impulses at both index fingers, stimulating both brain hemispheres sequentially through somatosensory feedback. After a static estimation of the current time series of candidate sources distributed over the brain cortex, a multivariate autoregressive (MAR) model was used to estimate the causal interactions between those found to be active due to the applied stimulus. The result was a visualization of the information transfer between active sources. A reasonably accurate estimation was achieved, even in presence of low signal to noise ratio (SNR) of EEG. Physiologically plausible source locations and connecting pathways were found, but an explanation of the observed phenomena and the interpretation of the differences between patients and controls is beyond the scope of this thesis, as a more extensive study is needed for this purpose. Consistent results found within subjects provide evidence of the potential value of this method in longitudinal studies, even when a comparison between subjects were not possible due to confounding factors (i.e. SNR differences). The obtained results support the candidacy of this method in the study of stroke, as it was found to be useful to track the information flow in the brain and might constitute a first step towards the development of a precise prognostic model of stroke.