Key to linking brain activity to consciousness – or to any brain function – is to be able to decipher causal interactions among different brain regions from neuroimaging data. One useful tool for doing this is Granger causality, a measure of ‘directed functional connectivity’ which is based on precedence and predictability. Put simply, one signal A is said to ‘Granger cause’ a different signal B if (and only if) A contains information that helps predict the future of B, over and above information already in the past of B. At the Sackler Centre we have been pioneering the theory and application of Granger causality to data from neuroscience. A particular controversy has surrounded its application to functional MRI (fMRI) data. This is because the BOLD responses measured by fMRI are sluggish and delayed reflections of the underlying neural activity. This has led many people to assume that Granger causality won’t work on this kind of data. However, we have recently shown - using new theory and advanced computational modelling – that many of these assumptions are wrong. There are still many challenges in applying Granger causality to fMRI, but our work is helping bridge the gap between theory and practice. We are also working on how best to apply Granger causality to other data types, such as surface and intracranial EEG signals.
People:
Reference:
Seth, A.K., Chorley, P., and Barnett, L. (2013). Granger causality analysis of fMRI BOLD signals is invariant to hemodynamic confound but not downsampling. Neuroimage. 65:540-555