Cognitive States Detection in fMRI using incremental P

The functional Magnetic Resonance Imaging (fMRI) has emerged as a powerful noninvasive method for collecting large amount of data about activities in human brain. Analysis for fMRI is essential to the success in detecting cognitive states. Due to very high dimensionality of feature vectors, feature extraction should be considered as a critical step to preprocess fMRI data before the stage of cognitive state detection. Up to now, different feature extraction methods have been applied to this type of data and they require domain experts to specify the Regions of Interests (Rol). However, none of them can give a dominant approach for precisely detecting cognitive states. In this paper, incremental principal component analysis (iPCA) proves to be an efficient method of feature extraction for fMRI data without using domain experts. Our experimental results show that this approach gives a higher performance compared to other feature extraction methods which require domain experts to select Regions of Interests.