How to measure functional connectivity using resting-state fMRI? A comprehensive empirical exploration of different connectivity metrics
BackgroundFunctional connectivity in the context of functional magnetic resonance imaging is typically quantified by Pearso[n]s or partial correlation between regional time series of the blood oxygenation level dependent signal. However, a recent interdisciplinary methodological work proposes more than 230 different metrics to measure similarity between different types of time series. ObjectiveHence, we systematically evaluated how the results of typical research approaches in functional neuroimaging vary depending on the functional connectivity metric of choice. We further explored which metrics most accurately detect neural decline induced by age and malignant brain tumors, aiming to initiate a debate on how best assessing brain connectivity in functional neuroimaging research. MethodsWe addressed both research questions using four independent neuroimaging datasets, comprising multimodal data from a total of 1187 individuals. We analyzed resting-state functional sequences to calculate functional connectivity using 20 representative metrics from four distinct mathematical domains. We further used T1- and T2-weighted images to compute regional brain volumes, diffusion-weighted imaging data to build structural connectomes, and pseudo-continuous arterial spin labeling to measure regional brain perfusion. ResultsFirst, our results demonstrate that the results of typical functional neuroimaging approaches differ fundamentally depending on the functional connectivity metric of choice. Second, we show that correlational and distance metrics are most appropriate to cover neural decline induced by age. In this context, partial correlation performs worse than other correlational metrics. Third, our findings suggest that the FC metric of choice depends on the utilized scanning parameters, the regions of interest, and the individual investigated. Lastly, beyond the major objective of this study, we provide evidence in favor of brain perfusion measured via pseudo-continuous arterial spin labeling as a robust neural entity mirroring age-related neural and cognitive decline. ConclusionOur empirical evaluation supports a recent theoretical functional connectivity framework. Future functional imaging studies need to comprehensively define the study-specific theoretical property of interest, the methodological property to assess the theoretical property, and the confounding property that may bias the conclusions.