A non-voxel based feature extraction to detect cognitive states in fMRI

Over the past few decades, Functional Magnetic Resonance Imaging (fMRI) has evolved into an important utilitarian tool for analyzing brain activity and detecting the cognitive states of a subject. The design and development of effective dimensionality reduction schema for the discovery of discriminatory features for cognitive state classification has become an area of vital interest toward enhanced decision support applications for patients with various brain disorders. In this paper, we present a unique non-voxel based approach using wavelet descriptor differentiation and principal components to extract unique features that reduce slice variability in fMRI data for the enhanced classification of cognitive states. The set of cognitive states that we attempt to classify are 'a person reading a sentence' and 'a person reading a picture.' The discovered feature vector is small and achieves significant classification accuracy using different classifiers and under different Regions of Interest (ROI) constraints. Experimental results using this study demonstrate the effectiveness of the approach when compared to previous voxel-based approaches.