Skip to Main Content
This exploratory study proposes the use of artificial neural networks to analyze whole brain fMRI data. Because fMRI data is dimensionally exorbitant, the first step is to reduce the amount of data to a tractable size, which is accomplished using probabilistic independent component analysis (PICA). Then data enters a simple back propagation feed forward neural network. This network outputs correct predictions above chance level in a different sample of subjects. More interestingly, it is found that hidden nodes segregate and concentrate different, but coherent, brain networks, which are the target of interpretations to support cognitive processes during the assessment of brands' logos.