Comparison of Independent Component Analysis Algorithms for Removal of Ocular Artifacts from Electroencephalogram

The electroencephalogram (EEG) is useful for clinical diagnosis and in biomedical research. EEG recordings are distorted by electrooculogram (EOG) artifacts causing serious problem for EEG interpretation and analysis. An often preferable method is to apply Independent Component Analysis (ICA) to multichannel EEG recordings and remove a wide variety of artifacts from EEG recordings by eliminating the contributions of artifactual sources onto the scalp sensors. The estimated sources should be as independent as possible, for better removal of artifacts from EEG. In this paper, the actual independence of the components obtained from various ICA algorithms like OGWE, MS-ICA, SHIBBS, KernelICA, JADE and RADICAL are assessed and compared by a recently introduced Mutual Information (MI) Estimator based on k-neighbor statistics without using the probability density functions. The results show that RADICAL algorithm performs best at separating the source signals from the observed (mixed) EEG signals and is recommended for.