The Blind Source Separation (BSS) problem applies in many fields, in the context of biomedical signal processing it is usually applied to multichannel recordings in order to a) de-noise and b) discover meaningful signals in the mixture. The BSS framework is usually instantiated through Independent Component Analysis (ICA), which uses measures of independence as a means of unmixing signal recordings [1]. In the case of brain signal recordings, such as the electroencephalogram (EEG), mixed signal recordings from the scalp over many channels are used to gain an insight into brain function (or dysfunction). Whereas ICA in its standard form extracts independent components based on spatial information inherent in the multichannel recordings, our novel Spatio-Temporal ICA (ST-ICA) framework [2] uses both spatial and temporal information derived from multi-channel time-series, to extract underlying sources. To this end STICA can be used to extract information that is both spectrally as well as spatially overlapping – two confounding issues when reverting to “simple” ICA alone. This provides a very powerful means of functional brain imaging, both denoising and extracting multiple underlying sources of interest embedded in the brain signals – in one action. However, a necessary part of the ST-ICA process is the so-called component selection of components of relevance – this is where domain knowledge is paramount: the separation of spatially and spectrally overlapping sources based on their contribution to unique sources of interest. This process becomes problematic due to the large number of candidate components that the ST-ICA process generates. We have devised an innovative, systematic approach to component selection such that we can automate the extraction of multiple functional brain sources in the multichannel EEG in an efficient and reproduceable way.