Chapter 18 – Biomedical applications

Publisher Summary This chapter focuses on the use of independent component analysis (ICA) in biomedical systems. ICA method is employed for identification and removal of the reference signal contribution from intracranial electro-encephalography (iEEG) recorded with a scalp reference signal. All EEG potential measurements reflect the difference between two potentials, and can then be factorized as a static mixture of two source subspaces corresponding to the reference signal and signals of interest, respectively. The former subspace is one dimensional by definition, while the latter subspace may have a higher dimension. As far as the statistical independence between both subspaces is concerned, it should at least be approximately true because of the high resistivity of the skull between scalp electrode and intracranial electrodes. Then the FastICA method is used in order to automatically extract the scalp reference signal based on its independence from the bipolar iEEG data. When iEEG and scalp EEG are recorded simultaneously using the same scalp reference, the scalp reference signal R2 extracted by ICA from iEEG can also be used in order to clean the scalp EEG.