Physiological artefact control in real-time fMRI neurofeedback – A systematic review and meta-analysis

Abstract Physiological artefacts are a general challenge in functional magnetic resonance imaging (fMRI) and are of particular relevance in real-time fMRI neurofeedback (rt-fMRI NF), where they can contaminate the training procedure and lead to undesired effects. However, physiological noise correction has not received corresponding attention in rt-fMRI NF and most attempts have mainly focused on offline data correction. Given the relatively rare use of physiological denoising methods in rt-fMRI NF and little knowledge about how these methods relate to reported outcomes, in this paper we review recent practices in physiology correction, and conduct a meta-analysis to summarize how reported outcomes vary across different correction approaches. A total of 227 studies were included. We found that physiology correction methods are still not very common offline (29.1%) and online (31.3%) and that even basic physiology recordings are still quite rarely reported in rt-fMRI studies (18.1%). This suggests that the implementation of physiological noise correction and use of concurrently recorded physiological signals is still in early stages. When summarizing reported outcomes of real-time target signal regulation, we found that studies employing online physiological correction did not systematically report smaller meta-analytic effect sizes than studies without such correction. Among correction methods, ‘imaging signal-based’ approaches were being the most commonly applied online. From a combined empirical and theoretical perspective, we recommend the use of online physiology corrections in rt-fMRI NF studies.