Multivariate prediction of pain perception based on pre-stimulus activity

Abstract

The perception of pain is modulated by different processes such as, for example, expectations and attention regarding the upcoming stimulus. Such processes are initiated prior to the actual stimulus and are reflected in ongoing brain activity. Different processes that are by definition also complex in itself are reflected in pre-stimulus activity and therefore the detection of this activity pattern should benefit from a multivariate approach. To identify specific pre-stimulus EEG activity patterns related to subsequent pain perception in humans, we contrasted painful with non-painful sensations delivered at the individual threshold level during EEG measurements. The results of the multivariate EEG analysis revealed a high level of accuracy (group mean 68%) in predicting the pain categorization solely based on pre-stimulus activity. In particular, fronto-central regions and activity in the higher gamma band (60:120 Hz) were of maximal importance for classification. Additional analyses supported the specific role of the pattern of high gamma band activity prior to the stimulus for predicting the behavioral outcome and demonstrated that the informational value embedded in the pre-stimulus activity is nearly as informative as the post-stimulus processing and reflects a specific preparatory state. Further, a close relation between pre- and post-stimulus processing in the high gamma band was observed. These findings support the important role of a multivariate cognitive state prior to stimulus appearance for the emergence of the subjective perception of pain and the functional role of widespread high gamma band activity.