Stephen Smith1, Karla Miller1, Steen Moeller2, Junqian Xu2, Edward J. Auerbach2, Mark W. Woolrich3, Christian F. Beckmann4, 5, Mark Jenkinson1, Jesper Andersson1, Matthew F. Glasser6, David Van Essen6, David Feinberg7, 8, Essa Yacoub2, Kamil Ugurbil2
1FMRIB, Oxford University, Oxford, Oxfordshire, United Kingdom; 2Center for Magnetic Resonance Research, University of Minnesota; 3Oxford Centre for Human Brain Activity, Oxford University; 4Donders Institute for Brain, Cognition and Behaviour, Radboud University Nijmegen; 5MIRA Institute for Biomedical Technology and Technical Medicine, University of Twente; 6Anatomy and Neurobiology, Washington University School of Medicine; 7Advanced MRI Technologies; 8University of California, Berkeley
Current correlation-based approaches for resting-state networks measure average functional connectivity between regions over time, but this is not very meaningful if regions are part of multiple networks. One wants a network model that allows overlap, allowing a regions activity level to reflect one networks activity at some points in time and another networks activity at others. However, even approaches that do allow overlap have often maximised spatial independence, which may be suboptimal if networks have significant overlap. Here we identify functionally distinct networks by virtue of temporal independence, taking advantage of additional temporal richness via improvements in FMRI sampling rate.