As you know subcortical structures are bellow cortex, so they are located far from EEG electrodes. Therefor they have little contribution in scalp field potentials. Is it possible to extract their signals from EEG?
I am not sure how you can manage that without going for invasive EEG monitoring procedures. This is usually done in epilepsy surgery preparation work-up.
You should look into some of the EEG Source Analysis methods. You could look at this review paper (http://www.ncbi.nlm.nih.gov/pubmed/18990257). Then, depending on the manufacturer of your system, they may have an integrated solution.
It's difficult to localize scalp-recorded EEG signals to subcortical regions, namely due to anatomical considerations. For example, unlike the layered cortex, a region like the striatum would not have the necessary organization of pyramidal cells to give rise to localizable scalp-recorded EEG.
Michael Cohen has a nice commentary about this issue in HBM (2011):
Nasser , We do not need to do the subcortical invasive monitoring as the signatures of all emotions are specifically found in respiratory patterns. If we include electrodermal monitoring and other aspects of autonomic monitoring we can get almost 100 % correlation of experimentally induced emotions. The following and other articles we have published on emotions and cardiorespiratory rhythms can provide support that EEG is highly influenced by peripheral oscillations .This relationship has not been researched more specifically to tie emotions with relative presence of respiratory , cardiac , alpha, beta and gamma rhythms .
Unless there is some prior knowledge of the location and behavior of the deep brain activity (such as brainstem sources in some auditory experiments), It is not possible to accurately localize a deep brain source. The inverse problem (finding brain activity from scalp activity) is ill-posed without additional assumptions. Many inverse-solution algorithms exist that can produce unique solutions in deep brain regions, but the algorithms must make additional assumptions (e.g. solutions must be single dipoles), that are often too restrictive for EEG (e.g. activity could be distributed over large regions of the cortex).
We talk about this briefly in section 2 of the attached chapter.