Several studies dealing with brain computer interfaces (BCI) were conducted using the concept of independent components analysis (ICA). Most of these studies only used a reduced number of ICA techniques, mainly the two algorithms FastICA and InfoMax. The main goal of this paper is to present some key points regarding ICA to help BCI researchers not familiar with ICA techniques to select the best appropriate method to tackle the questions understudy. Therefore, the concept of ICA is briefly introduced as well as a short description of algorithms widely used in ICA community, namely SOBI, COM2, JADE, ICAR, FastICA and InfoMax algorithms. The implementation of the ICA technique in the field of the BCI is also reported. Finally, a comparative study of these algorithms, conducted on physiologically plausible simulated EEG data, shows that an appropriate selection of an ICA algorithm can significantly improve the overall capabilities of BCI systems.