William Hedley Thompson

The mean–variance relationship reveals two possible strategies for dynamic brain connectivity analysis in fMRI

When studying brain connectivity using fMRI, signal intensity time-series are typically correlated with each other in time to compute estimates of the degree of interaction between different brain regions and/or networks. In the static connectivity case, the problem of defining which connections that should be considered significant in the analysis can be addressed in a rather straightforward manner by a statistical thresholding that is based on the magnitude of the correlation coefficients. More recently, interest has come to focus on the dynamical aspects of brain connectivity and the problem of deciding which brain connections that are to be considered relevant in the context of dynamical changes in connectivity provides further options. Since we, in the dynamical case, are interested in changes in connectivity over time, the variance of the correlation time-series becomes a relevant parameter. In this study, we discuss the relationship between the mean and variance of brain connectivity time-series and show that by studying the relation between them, two conceptually different strategies to analyze dynamic functional brain connectivity become available. Using resting-state fMRI data from a cohort of 46 subjects, we show that the mean of fMRI connectivity time-series scales negatively with its variance. This finding leads to the suggestion that magnitude- versus variance-based thresholding strategies will induce different results in studies of dynamic functional brain connectivity. Our assertion is exemplified by showing that the magnitude-based strategy is more sensitive to within-resting-state network (RSN) connectivity compared to between-RSN connectivity whereas the opposite holds true for a variance-based analysis strategy. The implications of our findings for dynamical functional brain connectivity studies are discussed.

Top-down beta enhances bottom-up gamma.

Several recent studies have demonstrated that the bottom-up signaling of a visual stimulus is subserved by interareal gamma-band synchronization, whereas top-down influences are mediated by alpha-beta band synchronization. These processes may implement top-down control of stimulus processing if top-down and bottom-up mediating rhythms are coupled via cross-frequency interaction. To test this possibility, we investigated Granger-causal influences among awake macaque primary visual area V1, higher visual area V4, and parietal control area 7a during attentional task performance. Top-down 7a-to-V1 beta-band influences enhanced visually driven V1-to-V4 gamma-band influences. This enhancement was spatially specific and largest when beta-band activity preceded gamma-band activity by ∼0.1 s, suggesting a causal effect of top-down processes on bottom-up processes. We propose that this cross-frequency interaction mechanistically subserves the attentional control of stimulus selection. SIGNIFICANCE STATEMENT Contemporary research indicates that the alpha-beta frequency band underlies top-down control, whereas the gamma-band mediates bottom-up stimulus processing. This arrangement inspires an attractive hypothesis, which posits that top-down beta-band influences directly modulate bottom-up gamma band influences via cross-frequency interaction. We evaluate this hypothesis determining that beta-band top-down influences from parietal area 7a to visual area V1 are correlated with bottom-up gamma frequency influences from V1 to area V4, in a spatially specific manner, and that this correlation is maximal when top-down activity precedes bottom-up activity. These results show that for top-down processes such as spatial attention, elevated top-down beta-band influences directly enhance feedforward stimulus-induced gamma-band processing, leading to enhancement of the selected stimulus.

A jackknife approach to quantifying single-trial correlation between covariance-based metrics undefined on a single-trial basis.

The quantification of covariance between neuronal activities (functional connectivity) requires the observation of correlated changes and therefore multiple observations. The strength of such neuronal correlations may itself undergo moment-by-moment fluctuations, which might e.g. lead to fluctuations in single-trial metrics such as reaction time (RT), or may co-fluctuate with the correlation between activity in other brain areas. Yet, quantifying the relation between moment-by-moment co-fluctuations in neuronal correlations is precluded by the fact that neuronal correlations are not defined per single observation. The proposed solution quantifies this relation by first calculating neuronal correlations for all leave-one-out subsamples (i.e. the jackknife replications of all observations) and then correlating these values. Because the correlation is calculated between jackknife replications, we address this approach as jackknife correlation (JC). First, we demonstrate the equivalence of JC to conventional correlation for simulated paired data that are defined per observation and therefore allow the calculation of conventional correlation. While the JC recovers the conventional correlation precisely, alternative approaches, like sorting-and-binning, result in detrimental effects of the analysis parameters. We then explore the case of relating two spectral correlation metrics, like coherence, that require multiple observation epochs, where the only viable alternative analysis approaches are based on some form of epoch subdivision, which results in reduced spectral resolution and poor spectral estimators. We show that JC outperforms these approaches, particularly for short epoch lengths, without sacrificing any spectral resolution. Finally, we note that the JC can be applied to relate fluctuations in any smooth metric that is not defined on single observations.

From static to temporal network theory: Applications to functional brain connectivity

Network neuroscience has become an established paradigm to tackle questions related to the functional and structural connectome of the brain. Recently, interest has been growing in examining the temporal dynamics of the brain’s network activity. Although different approaches to capturing fluctuations in brain connectivity have been proposed, there have been few attempts to quantify these fluctuations using temporal network theory. This theory is an extension of network theory that has been successfully applied to the modeling of dynamic processes in economics, social sciences, and engineering article but it has not been adopted to a great extent within network neuroscience. The objective of this article is twofold: (i) to present a detailed description of the central tenets of temporal network theory and describe its measures, and; (ii) to apply these measures to a resting-state fMRI dataset to illustrate their utility. Furthermore, we discuss the interpretation of temporal network theory in the context of the dynamic functional brain connectome. All the temporal network measures and plotting functions described in this article are freely available as the Python package Teneto. Author Summary Temporal network theory is a subfield of network theory that has had limited application to date within network neuroscience. The aims of this work are to introduce temporal network theory, define the metrics relevant to the context of network neuroscience, and illustrate their potential by analyzing a resting-state fMRI dataset. We found both between-subjects and between-task differences that illustrate the potential for these tools to be applied in a wider context. Our tools for analyzing temporal networks have been released in a Python package called Teneto.

On Stabilizing the Variance of Dynamic Functional Brain Connectivity Time Series

Abstract Assessment of dynamic functional brain connectivity based on functional magnetic resonance imaging (fMRI) data is an increasingly popular strategy to investigate temporal dynamics of the brains large-scale network architecture. Current practice when deriving connectivity estimates over time is to use the Fisher transformation, which aims to stabilize the variance of correlation values that fluctuate around varying true correlation values. It is, however, unclear how well the stabilization of signal variance performed by the Fisher transformation works for each connectivity time series, when the true correlation is assumed to be fluctuating. This is of importance because many subsequent analyses either assume or perform better when the time series have stable variance or adheres to an approximate Gaussian distribution. In this article, using simulations and analysis of resting-state fMRI data, we analyze the effect of applying different variance stabilization strategies on connectivity time series. We focus our investigation on the Fisher transformation, the Box–Cox (BC) transformation and an approach that combines both transformations. Our results show that, if the intention of stabilizing the variance is to use metrics on the time series, where stable variance or a Gaussian distribution is desired (e.g., clustering), the Fisher transformation is not optimal and may even skew connectivity time series away from being Gaussian. Furthermore, we show that the suboptimal performance of the Fisher transformation can be substantially improved by including an additional BC transformation after the dynamic functional connectivity time series has been Fisher transformed.

Research: The readability of scientific texts is decreasing over time

Clarity and accuracy of reporting are fundamental to the scientific process. Readability formulas can estimate how difficult a text is to read. Here, in a corpus consisting of 709,577 abstracts published between 1881 and 2015 from 123 scientific journals, we show that the readability of science is steadily decreasing. Our analyses show that this trend is indicative of a growing use of general scientific jargon. These results are concerning for scientists and for the wider public, as they impact both the reproducibility and accessibility of research findings.

Bursty properties revealed in large-scale brain networks with a point-based method for dynamic functional connectivity

The brain is organized into large scale spatial networks that can be detected during periods of rest using fMRI. The brain is also a dynamic organ with activity that changes over time. We developed a method and investigated properties where the connections as a function of time are derived and quantified. The point based method (PBM) presented here derives covariance matrices after clustering individual time points based upon their global spatial pattern. This method achieved increased temporal sensitivity, together with temporal network theory, allowed us to study functional integration between resting-state networks. Our results show that functional integrations between two resting-state networks predominately occurs in bursts of activity. This is followed by varying intermittent periods of less connectivity. The described point-based method of dynamic resting-state functional connectivity allows for a detailed and expanded view on the temporal dynamics of resting-state connectivity that provides novel insights into how neuronal information processing is integrated in the human brain at the level of large-scale networks.

Functional resting-state fMRI connectivity correlates with serum levels of the S100B protein in the acute phase of traumatic brain injury

The S100B protein is an intra-cellular calcium-binding protein that mainly resides in astrocytes in the central nervous system. The serum level of S100B is used as biomarker for the severity of brain damage in traumatic brain injury (TBI) patients. In this study we investigated the relationship between intrinsic resting-state brain connectivity, measured 1–22 days (mean 8 days) after trauma, and serum levels of S100B in a patient cohort with mild-to-severe TBI in need of neuro-intensive care in the acute phase. In line with previous investigations, our results show that the peak level of S100B acquired during the acute phase of TBI was negatively correlated with behavioral measures (Glasgow Outcome Score, GOS) of functional outcome assessed 6 to 12 months post injury. Using a multi-variate pattern analysis-informed seed-based correlation analysis, we show that the strength of resting-state brain connectivity in multiple resting-state networks was negatively correlated with the peak of serum levels of S100B. A negative correspondence between S100B peak levels recorded 12–36 h after trauma and intrinsic connectivity was found for brain regions located in the default mode, fronto-parietal, visual and motor resting-state networks. Our results suggest that resting-state brain connectivity measures acquired during the acute phase of TBI is concordant with results obtained from molecular biomarkers and that it may hold a capacity to predict long-term cognitive outcome in TBI patients.

A simulation and comparison of dynamic functional connectivity methods

There is a current interest in quantifying brain dynamic functional connectivity (DFC) based on neuroimaging data such as fMRI. Many methods have been proposed, and are being applied, revealing new insight into the brain’s dynamics. However, given that the ground truth for DFC in the brain is unknown, many concerns remain regarding the accuracy of proposed estimates. Since there exists many DFC methods it is difficult to assess differences in dynamic brain connectivity between studies. Here, we evaluate five different methods that together represent a wide spectrum of current approaches to estimating DFC (sliding window, tapered sliding window, temporal derivative, spatial distance and jackknife correlation). In particular, we were interested in each methods’ ability to track changes in covariance over time, which is a key property in DFC analysis. We found that all tested methods correlated positively with each other, but there were large differences in the strength of the correlations between methods. To facilitate comparisons with future DFC methods, we propose that the described simulations can act as benchmark tests for evaluation of methods. In this paper, we present dfcbenchmarker, which is a Python package where researchers can easily submit and compare their own DFC methods to evaluate its performance.

A common framework for the problem of deriving estimates of dynamic functional brain connectivity

&NA; The research field of dynamic functional connectivity explores the temporal properties of brain connectivity. To date, many methods have been proposed, which are based on quite different assumptions. In order to understand in which way the results from different techniques can be compared to each other, it is useful to be able to formulate them within a common theoretical framework. In this study, we describe such a framework that is suitable for many of the dynamic functional connectivity methods that have been proposed. Our overall intention was to derive a theoretical framework that was constructed such that a wide variety of dynamic functional connectivity techniques could be expressed and evaluated within the same framework. At the same time, care was given to the fact that key features of each technique could be easily illustrated within the framework and thus highlighting critical assumptions that are made. We aimed to create a common framework which should serve to assist comparisons between different analytical methods for dynamic functional brain connectivity and promote an understanding of their methodological advantages as well as potential drawbacks. HighlightsDifferent approaches to compute dynamic functional brain connectivity have been proposed, each with their own assumptions.We present a theoretical framework that encompasses a large majority of proposed methods.Our common framework facilitates comparisons between different methods by highlighting their underlying assumptions.