Robust detection of brain function activity using blind signal separation

Independent component analysis (ICA) is regarded as a useful technique for processing a wide range of practical signals, such as speech, radar and biomedical recordings. In the biomedical image processing, functional magnetic resonance imaging become a common tool for investigating the brain function and cognitive process. However, much debate on the preferred technique for analyzing these functional activation images is still a problem. In this contribution, blind signal separation via ICA is proposed to detect the brain function activities. Several experiment with digital image data were also carried out based on the presented fastICA algorithm. ICA technique is employed to separate the independent components of the observation and restrain the impact caused by the additive noise. The results using common method and ICA technique were also demonstrated and compared to show that the proposed ICA method significantly reduces the physiological baseline fluctuation and the background interfaces.