Brain connectivity analysis for real-time fMRI neurofeedback experiments

Functional MRI has evolved to become one of the most innovative methods to study the human brain. With a real-time fMRI system, it is possible to perform neurofeedback experiments, where stimuli are adapted in real time to the participants' measured brain activity. Cognitive processes depend on the communication between regions of the brain, and so the analysis of connectivity before, during and after a neurofeed-back experiment is of great interest. This work analyses and benchmarks some of the methods currently used on fMRI data: Correlation, Coherence and Granger Causality Mapping. Tests performed on simulated datasets showed that coherence presents robustness to downsampling, unlike correlation and Granger causality. Preliminary analysis of connectivity on an experimental dataset allowed for interesting insights on the working memory network functioning. The work developed has raised a number of questions related to the application of connectivity methods to fMRI data, results validation and optimisation of fMRI experimental setups, as well as a number of challenges to be surpassed in future work.