An ICA to identify a large number of candidate correlation patterns is carried out based on a time series of image data. The large number of candidate correlation patterns includes a large number of neurophysical events, as well as false patterns owing to noise. The neurophysical events as well as the false patterns are then separated, for example on the basis of a metric, which indicates an intensity of the candidate correlation patterns in a section of the brain, or by a computer-implemented classifier. Techniques of this kind can be used in conjunction with functional magnetic resonance imaging.