Better techniques for untangling correlations in big fMRI data
Functional Magnetic Resonance Imaging (fMRI) has revolutionized our ability to measure and assess brain activity. However, when researchers try to compare fMRI data collected from subjects in naturalistic conditions (such as while watching a movie or listening to music), current methods produce false positive rates so high that the validity of these methods is called into question.
In part 1 of this two-part study, IRP researchers led by Robert W. Cox, Ph.D., identified ‘subject-wise bootstrapping’ (SWB) as the best nonparametric method for making inter-subject correlation inferences within a single group, while ‘subject-wise permutation’ (SWP) was found to be both more reliable and required far less computation than the cumbersome method traditionally used to define statistics of the inter-subject correlations in two groups. In part 2, researchers identified a parametric approach—linear mixed-eﬀects modeling (LME)—that is more eﬃcient, more adaptable, easier to use, and more robust than traditional nonparametric methods.
This comprehensive upgrade of modeling and analysis techniques for fMRI data provides both theoretical and practical help to scientists in many different fields who manage big datasets as part of their research. In order to further advance the field, Dr. Cox's team also offers researchers an open source option in the AFNI (Analysis of Functional NeuroImages) suite, where they can access programs that process, analyze and display fMRI data.
Chen G, Shin YW, Taylor PA, Glen DR, Reynolds RC, Israel RB, Cox RW. (2016). Untangling the relatedness among correlations, part I: Nonparametric approaches to inter-subject correlation analysis at the group level. Neuroimage. 142:248-259.
Chen G, Taylor PA, Shin YW, Reynolds RC, Cox RW. (2017). Untangling the relatedness among correlations, Part II: Inter-subject correlation group analysis through linear mixed-effects modeling. Neuroimage. 147:825-840.