Cognitive states classification from fMRI data using support vector machines

It is useful to find the sequence of hidden cognitive states that subjects pass through when performing some complex task. We present a method to classify cognitive states from fMRI data using SVMs. A dataset of the study of Chinese character vs. Pinyin is considered. The data images are firstly processed and transformed to normalized coordinates. The features are extracted based on activities of voxel and index of Brodmann's areas. Then, they are used as input vectors to train the classifiers of SVMs. The results indicate that it is feasible for either single subject cognitive classification or multiple subjects'. The method is helpful to decode cognitive states.