Global signal regression

Just a very expensive breathing training? Risk of respiratory artefacts in functional connectivity-based real-time fMRI neurofeedback

Real-time functional magnetic resonance imaging neurofeedback (rtfMRI NFB) is a promising method for targeted regulation of pathological brain processes in mental disorders. But most NFB approaches so far have used relatively restricted regional activation as a target, which might not address the complexity of the underlying network changes. Aiming towards advancing novel treatment tools for disorders like schizophrenia, we developed a large-scale network functional connectivity-based rtfMRI NFB approach targeting dorsolateral prefrontal cortex and anterior cingulate cortex connectivity with the striatum. In a double-blind randomized yoke-controlled single-session feasibility study with N ​= ​38 healthy controls, we identified strong associations between our connectivity estimates and physiological parameters reflecting the rate and regularity of breathing. These undesired artefacts are especially detrimental in rtfMRI NFB, where the same data serves as an online feedback signal and offline analysis target. To evaluate ways to control for the identified respiratory artefacts, we compared model-based physiological nuisance regression and global signal regression (GSR) and found that GSR was the most effective method in our data. Our results strongly emphasize the need to control for physiological artefacts in connectivity-based rtfMRI NFB approaches and suggest that GSR might be a useful method for online data correction for respiratory artefacts.