An important field of research in functional neuroimaging is the discovery of integrated, distributed brain systems and networks, whose different regions need to work in unison for normal functioning.
The EEG is a non-invasive technique that can provide information for massive connectivity analyses. Cortical signals of time varying electric neuronal activity can be estimated from the EEG. Although such techniques have very high time resolution, two cortical signals even at distant locations will appear to be highly similar due to the low spatial resolution nature of the EEG.
In this study a method for eliminating the effect of common sources due to low spatial resolution is presented. It is based on an efficient estimation of the whole-cortex partial coherence matrix. Using as a starting point any linear EEG tomography that satisfies the EEG forward equation, it is shown that the generalized partial coherences for the cortical grey matter current density time series are invariant to the selected tomography. It is empirically shown with simulation experiments that the generalized partial coherences have higher spatial resolution than the classical coherences. The results demonstrate that with as little as 19 electrodes, lag-connected brain regions can often be missed and misplaced even with lagged coherence measures, while the new method detects and localizes correctly the connected regions using the lagged partial coherences.