Item

Abstract

Neurofeedback is a type of biofeedback that uses the real-time display of neural signals to facilitate self-regulation of brain activity. During neurofeedback training, participants try to up- or down-regulate a certain dynamical parameter calculated from the recorded neuronal activity (e.g., average amplitude of oscillations). It has been shown that these training sessions can lead to clinical improvements in a number of brain disorders: ADHD, stroke, Parkinson’s disease, PTSD, etc. However, previous studies report that up to 50% of the participants are unable to properly modulate their brain activity via neurofeedback, and the reasons why this occurs are still unclear. While neurofeedback learning success is typically evaluated by examining changes in the target neuronal activity, the co-occurring modulation of distributed neuronal networks is usually not investigated. We hypothesize that the observed success in self-regulation can be attributed to subject-specific spatio-temporal patterns of distributed activity and connectivity preceding up- or down-regulation of the target neuronal parameter during neurofeedback. In the present study, we plan to develop a computational approach for the extraction of such patterns and apply it to publicly available datasets of neurofeedback training protocols. Since patterns of self-regulation can greatly depend on the mental strategy that participants use, we will first validate our method using the data from a brain-computer interface (BCI) training with feedback, where subjects had a fixed and well-investigated mental strategy – motor imagery of left and right-hand movements. After validation, the approach will be applied to neurofeedback recordings where subjects have no explicit strategy, in order to analyze individual aspects of their neurofeedback learning.