Data and model considerations for estimating time-varying functional connectivity in fMRI.

Functional connectivity (FC) in the brain has been shown to exhibit subtle but reliable modulations within a session. One way of estimating time-varying FC is by using state-based models that describe fMRI time series as temporal sequences of states, each with an associated, characteristic pattern of FC. However, the estimation of these models from data sometimes fails to capture changes in a meaningful way, such that the model estimation assigns entire sessions (or the largest part of them) to a single state, therefore failing to capture within-session state modulations effectively; we refer to this phenomenon as the model becoming static, or model stasis. Here, we aim to quantify how the nature of the data and the choice of model parameters affect the model's ability to detect temporal changes in FC using both simulated fMRI time courses and resting state fMRI data. We show that large between-subject FC differences can overwhelm subtler within-session modulations, causing the model to become static. Further, the choice of parcellation can also affect the model's ability to detect temporal changes. We finally show that the model often becomes static when the number of free parameters per state that need to be estimated is high and the number of observations available for this estimation is low in comparison. Based on these findings, we derive a set of practical recommendations for time-varying FC studies, in terms of preprocessing, parcellation and complexity of the model.

Behavioural relevance of spontaneous, transient brain network interactions in fMRI.

How spontaneously fluctuating functional magnetic resonance imaging (fMRI) signals in different brain regions relate to behaviour has been an open question for decades. Correlations in these signals, known as functional connectivity, can be averaged over several minutes of data to provide a stable representation of the functional network architecture for an individual. However, associations between these stable features and behavioural traits have been shown to be dominated by individual differences in anatomy. Here, using kernel learning tools, we propose methods to assess and compare the relation between time-varying functional connectivity, time-averaged functional connectivity, structural brain data, and non-imaging subject behavioural traits. We applied these methods to Human Connectome Project resting-state fMRI data to show that time-varying fMRI functional connectivity, detected at time-scales of a few seconds, has associations with some behavioural traits that are not dominated by anatomy. Despite time-averaged functional connectivity accounting for the largest proportion of variability in the fMRI signal between individuals, we found that some aspects of intelligence could only be explained by time-varying functional connectivity. The finding that time-varying fMRI functional connectivity has a unique relationship to population behavioural variability suggests that it might reflect transient neuronal communication fluctuating around a stable neural architecture.

Transient spectral events in resting state MEG predict individual task responses.

Even in response to simple tasks such as hand movement, human brain activity shows remarkable inter-subject variability. Recently, it has been shown that individual spatial variability in fMRI task responses can be predicted from measurements collected at rest; suggesting that the spatial variability is a stable feature, inherent to the individual's brain. However, it is not clear if this is also true for individual variability in the spatio-spectral content of oscillatory brain activity. Here, we show using MEG (N â€‹= â€‹89) that we can predict the spatial and spectral content of an individual's task response using features estimated from the individual's resting MEG data. This works by learning when transient spectral 'bursts' or events in the resting state tend to reoccur in the task responses. We applied our method to motor, working memory and language comprehension tasks. All task conditions were predicted significantly above chance. Finally, we found a systematic relationship between genetic similarity (e.g. unrelated subjects vs. twins) and predictability. Our approach can predict individual differences in brain activity and suggests a link between transient spectral events in task and rest that can be captured at the level of individuals.

Discovery of key whole-brain transitions and dynamics during human wakefulness and non-REM sleep.

The modern understanding of sleep is based on the classification of sleep into stages defined by their electroencephalography (EEG) signatures, but the underlying brain dynamics remain unclear. Here we aimed to move significantly beyond the current state-of-the-art description of sleep, and in particular to characterise the spatiotemporal complexity of whole-brain networks and state transitions during sleep. In order to obtain the most unbiased estimate of how whole-brain network states evolve through the human sleep cycle, we used a Markovian data-driven analysis of continuous neuroimaging data from 57 healthy participants falling asleep during simultaneous functional magnetic resonance imaging (fMRI) and EEG. This Hidden Markov Model (HMM) facilitated discovery of the dynamic choreography between different whole-brain networks across the wake-non-REM sleep cycle. Notably, our results reveal key trajectories to switch within and between EEG-based sleep stages, while highlighting the heterogeneities of stage N1 sleep and wakefulness before and after sleep.

A dynamic system of brain networks revealed by fast transient EEG fluctuations and their fMRI correlates.

Resting state brain activity has become a significant area of investigation in human neuroimaging. An important approach for understanding the dynamics of neuronal activity in the resting state is to use complementary imaging modalities. Electrophysiological recordings can access fast temporal dynamics, while functional magnetic resonance imaging (fMRI) studies reveal detailed spatial patterns. However, the relationship between these two measures is not fully established. In this study, we used simultaneously recorded electroencephalography (EEG) and fMRI, along with Hidden Markov Modelling, to investigate how network dynamics at fast sub-second time-scales, accessible with EEG, link to the slower time-scales and higher spatial detail of fMRI. We found that the fMRI correlates of fast transient EEG dynamic networks show highly reproducible spatial patterns, and that their spatial organization exhibits strong similarity with traditional fMRI resting state networks maps. This further demonstrates the potential of electrophysiology as a tool for understanding the fast network dynamics that underlie fMRI resting state networks.

Bayesian Optimisation of Large-Scale Biophysical Networks.

The relationship between structure and function in the human brain is well established, but not yet well characterised. Large-scale biophysical models allow us to investigate this relationship, by leveraging structural information (e.g. derived from diffusion tractography) in order to couple dynamical models of local neuronal activity into networks of interacting regions distributed across the cortex. In practice however, these models are difficult to parametrise, and their simulation is often delicate and computationally expensive. This undermines the experimental aspect of scientific modelling, and stands in the way of comparing different parametrisations, network architectures, or models in general, with confidence. Here, we advocate the use of Bayesian optimisation for assessing the capabilities of biophysical network models, given a set of desired properties (e.g. band-specific functional connectivity); and in turn the use of this assessment as a principled basis for incremental modelling and model comparison. We adapt an optimisation method designed to cope with costly, high-dimensional, non-convex problems, and demonstrate its use and effectiveness. Using five parameters controlling key aspects of our model, we find that this method is able to converge to regions of high functional similarity with real MEG data, with very few samples given the number of parameters, without getting stuck in local extrema, and while building and exploiting a map of uncertainty defined smoothly across the parameter space. We compare the results obtained using different methods of structural connectivity estimation from diffusion tractography, and find that one method leads to better simulations.

How reliable are MEG resting-state connectivity metrics?

MEG offers dynamic and spectral resolution for resting-state connectivity which is unavailable in fMRI. However, there are a wide range of available network estimation methods for MEG, and little in the way of existing guidance on which ones to employ. In this technical note, we investigate the extent to which many popular measures of stationary connectivity are suitable for use in resting-state MEG, localising magnetic sources with a scalar beamformer. We use as empirical criteria that network measures for individual subjects should be repeatable, and that group-level connectivity estimation shows good reproducibility. Using publically-available data from the Human Connectome Project, we test the reliability of 12 network estimation techniques against these criteria. We find that the impact of magnetic field spread or spatial leakage artefact is profound, creates a major confound for many connectivity measures, and can artificially inflate measures of consistency. Among those robust to this effect, we find poor test-retest reliability in phase- or coherence-based metrics such as the phase lag index or the imaginary part of coherency. The most consistent methods for stationary connectivity estimation over all of our tests are simple amplitude envelope correlation and partial correlation measures.

Intrusive memories to traumatic footage: the neural basis of their encoding and involuntary recall.

BACKGROUND: A hallmark symptom after psychological trauma is the presence of intrusive memories. It is unclear why only some moments of trauma become intrusive, and how these memories involuntarily return to mind. Understanding the neural mechanisms involved in the encoding and involuntary recall of intrusive memories may elucidate these questions. METHOD: Participants (n = 35) underwent functional magnetic resonance imaging (fMRI) while being exposed to traumatic film footage. After film viewing, participants indicated within the scanner, while undergoing fMRI, if they experienced an intrusive memory of the film. Further intrusive memories in daily life were recorded for 7 days. After 7 days, participants completed a recognition memory test. Intrusive memory encoding was captured by comparing activity at the time of viewing 'Intrusive scenes' (scenes recalled involuntarily), 'Control scenes' (scenes never recalled involuntarily) and 'Potential scenes' (scenes recalled involuntarily by others but not that individual). Signal change associated with intrusive memory involuntary recall was modelled using finite impulse response basis functions. RESULTS: We found a widespread pattern of increased activation for Intrusive v. both Potential and Control scenes at encoding. The left inferior frontal gyrus and middle temporal gyrus showed increased activity in Intrusive scenes compared with Potential scenes, but not in Intrusive scenes compared with Control scenes. This pattern of activation persisted when taking recognition memory performance into account. Intrusive memory involuntary recall was characterized by activity in frontal regions, notably the left inferior frontal gyrus. CONCLUSIONS: The left inferior frontal gyrus may be implicated in both the encoding and involuntary recall of intrusive memories.

Probabilistic non-linear registration with spatially adaptive regularisation.

This paper introduces a novel method for inferring spatially varying regularisation in non-linear registration. This is achieved through full Bayesian inference on a probabilistic registration model, where the prior on the transformation parameters is parameterised as a weighted mixture of spatially localised components. Such an approach has the advantage of allowing the registration to be more flexibly driven by the data than a traditional globally defined regularisation penalty, such as bending energy. The proposed method adaptively determines the influence of the prior in a local region. The strength of the prior may be reduced in areas where the data better support deformations, or can enforce a stronger constraint in less informative areas. Consequently, the use of such a spatially adaptive prior may reduce unwanted impacts of regularisation on the inferred transformation. This is especially important for applications where the deformation field itself is of interest, such as tensor based morphometry. The proposed approach is demonstrated using synthetic images, and with application to tensor based morphometry analysis of subjects with Alzheimer's disease and healthy controls. The results indicate that using the proposed spatially adaptive prior leads to sparser deformations, which provide better localisation of regional volume change. Additionally, the proposed regularisation model leads to more data driven and localised maps of registration uncertainty. This paper also demonstrates for the first time the use of Bayesian model comparison for selecting different types of regularisation.

A symmetric multivariate leakage correction for MEG connectomes.

Ambiguities in the source reconstruction of magnetoencephalographic (MEG) measurements can cause spurious correlations between estimated source time-courses. In this paper, we propose a symmetric orthogonalisation method to correct for these artificial correlations between a set of multiple regions of interest (ROIs). This process enables the straightforward application of network modelling methods, including partial correlation or multivariate autoregressive modelling, to infer connectomes, or functional networks, from the corrected ROIs. Here, we apply the correction to simulated MEG recordings of simple networks and to a resting-state dataset collected from eight subjects, before computing the partial correlations between power envelopes of the corrected ROItime-courses. We show accurate reconstruction of our simulated networks, and in the analysis of real MEGresting-state connectivity, we find dense bilateral connections within the motor and visual networks, together with longer-range direct fronto-parietal connections.

Modulation of alpha power at encoding and retrieval tracks the precision of visual short-term memory.

Our ability to hold information in mind is strictly limited. We sought to understand the relationship between oscillatory brain activity and the allocation of resources within visual short-term memory (VSTM). Participants attempted to remember target arrows embedded among distracters and used a continuous method of responding to report their memory for a cued target item. Trial-to-trial variability in the absolute circular accuracy with which participants could report the target was predicted by event-related alpha synchronization during initial processing of the memoranda and by alpha desynchronization during the retrieval of those items from VSTM. Using a model-based approach, we were also able to explore further which parameters of VSTM-guided behavior were most influenced by alpha band changes. Alpha synchronization during item processing enhanced the precision with which an item could be retained without affecting the likelihood of an item being represented per se (as indexed by the guessing rate). Importantly, our data outline a neural mechanism that mirrors the precision with which items are retained; the greater the alpha power enhancement during encoding, the greater the precision with which that item can be retained.

Task induced modulation of neural oscillations in electrophysiological brain networks.

In recent years, one of the most important findings in systems neuroscience has been the identification of large scale distributed brain networks. These networks support healthy brain function and are perturbed in a number of neurological disorders (e.g. schizophrenia). Their study is therefore an important and evolving focus for neuroscience research. The majority of network studies are conducted using functional magnetic resonance imaging (fMRI) which relies on changes in blood oxygenation induced by neural activity. However recently, a small number of studies have begun to elucidate the electrical origin of fMRI networks by searching for correlations between neural oscillatory signals from spatially separate brain areas in magnetoencephalography (MEG) data. Here we advance this research area. We introduce two methodological extensions to previous independent component analysis (ICA) approaches to MEG network characterisation: 1) we show how to derive pan-spectral networks that combine independent components computed within individual frequency bands. 2) We show how to measure the temporal evolution of each network with millisecond temporal resolution. We apply our approach to ~10h of MEG data recorded in 28 experimental sessions during 3 separate cognitive tasks showing that a number of networks could be identified and were robust across time, task, subject and recording session. Further, we show that neural oscillations in those networks are modulated by memory load, and task relevance. This study furthers recent findings on electrodynamic brain networks and paves the way for future clinical studies in patients in which abnormal connectivity is thought to underlie core symptoms.

Inferring task-related networks using independent component analysis in magnetoencephalography.

A novel framework for analysing task-positive data in magnetoencephalography (MEG) is presented that can identify task-related networks. Techniques that combine beamforming, the Hilbert transform and temporal independent component analysis (ICA) have recently been applied to resting-state MEG data and have been shown to extract resting-state networks similar to those found in fMRI. Here we extend this approach in two ways. First, we systematically investigate optimisation of time-frequency windows for connectivity measurement. This is achieved by estimating the distribution of functional connectivity scores between nodes of known resting-state networks and contrasting it with a distribution of artefactual scores that are entirely due to spatial leakage caused by the inverse problem. We find that functional connectivity, both in the resting-state and during a cognitive task, is best estimated via correlations in the oscillatory envelope in the 8-20 Hz frequency range, temporally down-sampled with windows of 1-4s. Second, we combine ICA with the general linear model (GLM) to incorporate knowledge of task structure into our connectivity analysis. The combination of ICA with the GLM helps overcome problems of these techniques when used independently: namely, the interpretation and separation of interesting independent components from those that represent noise in ICA and the correction for multiple comparisons when applying the GLM. We demonstrate the approach on a 2-back working memory task and show that this novel analysis framework is able to elucidate the functional networks involved in the task beyond that which is achieved using the GLM alone. We find evidence of localised task-related activity in the area of the hippocampus, which is difficult to detect reliably using standard methods. Task-positive ICA, coupled with the GLM, has the potential to be a powerful tool in the analysis of MEG data.

Measuring functional connectivity in MEG: a multivariate approach insensitive to linear source leakage.

A number of recent studies have begun to show the promise of magnetoencephalography (MEG) as a means to non-invasively measure functional connectivity within distributed networks in the human brain. However, a number of problems with the methodology still remain--the biggest of these being how to deal with the non-independence of voxels in source space, often termed signal leakage. In this paper we demonstrate a method by which non-zero lag cortico-cortical interactions between the power envelopes of neural oscillatory processes can be reliably identified within a multivariate statistical framework. The method is spatially unbiased, moderately conservative in false positive rate and removes linear signal leakage between seed and target voxels. We demonstrate this methodology in simulation and in real MEG data. The multivariate method offers a powerful means to capture the high dimensionality and rich information content of MEG signals in a single imaging statistic. Given a significant interaction between two areas, we go on to show how classical statistical tests can be used to quantify the importance of the data features driving the interaction.

The danger of systematic bias in group-level FMRI-lag-based causality estimation.

Schippers, Renken and Keysers (NeuroImage, 2011) present a simulation of multi-subject lag-based causality estimation. We fully agree that single-subject evaluations (e.g., Smith et al., 2011) need to be revisited in the context of multi-subject studies, and Schippers' paper is a good example, including detailed multi-level simulation and cross-subject statistical modelling. The authors conclude that "the average chance to find a significant Granger causality effect when no actual influence is present in the data stays well below the p-level imposed on the second level statistics" and that "when the analyses reveal a significant directed influence, this direction was accurate in the vast majority of the cases". Unfortunately, we believe that the general meaning that may be taken from these statements is not supported by the paper's results, as there may in reality be a systematic (group-average) difference in haemodynamic delay between two brain areas. While many statements in the paper (e.g., the final two sentences) do refer to this problem, we fear that the overriding message that many readers may take from the paper could cause misunderstanding.

Partial volume correction of multiple inversion time arterial spin labeling MRI data.

The accuracy of cerebral blood flow (CBF) estimates from arterial spin labeling (ASL) is affected by the presence of both gray matter (GM) and white matter within any voxel. Recently a partial volume (PV) correction method for ASL has been demonstrated (Asllani et al. Magn Reson Med 2008; 60:1362-1371), where PV estimates were used with a local linear regression to separate the GM and white matter ASL signal. Here a new PV correction method for multi-inversion time ASL is proposed that exploits PV estimates within a spatially regularized kinetic curve model analysis. The proposed method exploits both PV estimates and the different kinetics of the ASL signal arising from GM and white matter. The new correction method is shown, on both simulated and real data, to provide correction of GM CBF comparable to a linear regression approach, whilst preserving greater spatial detail in the CBF image. On real data corrected GM CBF values were found to be largely independent of GM PV, implying that the correction had been successful. Increases of mean GM CBF after correction of 69-80% were observed.

Multiple-subjects connectivity-based parcellation using hierarchical Dirichlet process mixture models.

We propose a hierarchical infinite mixture model approach to address two issues in connectivity-based parcellations: (i) choosing the number of clusters, and (ii) combining data from different subjects. In a Bayesian setting, we model voxel-wise anatomical connectivity profiles as an infinite mixture of multivariate Gaussian distributions, with a Dirichlet process prior on the cluster parameters. This type of prior allows us to conveniently model the number of clusters and estimate its posterior distribution directly from the data. An important benefit of using Bayesian modelling is the extension to multiple subjects clustering via a hierarchical mixture of Dirichlet processes. Data from different subjects are used to infer on class parameters and the number of classes at individual and group level. Such a method accounts for inter-subject variability, while still benefiting from combining different subjects data to yield more robust estimates of the individual clusterings.

A Bayesian framework for global tractography.

We readdress the diffusion tractography problem in a global and probabilistic manner. Instead of tracking through local orientations, we parameterise the connexions between brain regions at a global level, and then infer on global and local parameters simultaneously in a Bayesian framework. This approach offers a number of important benefits. The global nature of the tractography reduces sensitivity to local noise and modelling errors. By constraining tractography to ensure a connexion is found, and then inferring on the exact location of the connexion, we increase the robustness of connectivity-based parcellations, allowing parcellations of connexions that were previously invisible to tractography. The Bayesian framework allows a direct comparison of the evidence for connecting and non-connecting models, to test whether the connexion is supported by the data. Crucially, by explicit parameterisation of the connexion between brain regions, we infer on a parameter that is shared with models of functional connectivity. This model is a first step toward the joint inference on functional and anatomical connectivity.

Probabilistic diffusion tractography with multiple fibre orientations: What can we gain?

We present a direct extension of probabilistic diffusion tractography to the case of multiple fibre orientations. Using automatic relevance determination, we are able to perform online selection of the number of fibre orientations supported by the data at each voxel, simplifying the problem of tracking in a multi-orientation field. We then apply the identical probabilistic algorithm to tractography in the multi- and single-fibre cases in a number of example systems which have previously been tracked successfully or unsuccessfully with single-fibre tractography. We show that multi-fibre tractography offers significant advantages in sensitivity when tracking non-dominant fibre populations, but does not dramatically change tractography results for the dominant pathways.

Automated single-trial measurement of amplitude and latency of laser-evoked potentials (LEPs) using multiple linear regression.

OBJECTIVE: Laser stimulation of Adelta-fibre nociceptors in the skin evokes nociceptive-specific brain responses (laser-evoked potentials, LEPs). The largest vertex complex (N2-P2) is widely used to assess nociceptive pathways in physiological and clinical studies. The aim of this study was to develop an automated method to measure amplitudes and latencies of the N2 and P2 peaks on a single-trial basis. METHODS: LEPs were recorded after Nd:YAP laser stimulation of the left hand dorsum in 7 normal volunteers. For each subject, a basis set of 4 regressors (the N2 and P2 waveforms and their respective temporal derivatives) was derived from the time-averaged data and regressed against every single-trial LEP response. This provided a separate quantitative estimate of amplitude and latency for the N2 and P2 components of each trial. RESULTS: All estimates of LEP parameters correlated significantly with the corresponding measurements performed by a human expert (N2 amplitude: R2=0.70; P2 amplitude: R2=0.70; N2 latency: R2=0.81; P2 latency: R2=0.59. All P<0.0001). Furthermore, regression analysis was able to extract an LEP response from a subset of the trials that had been classified by the human expert as without response. CONCLUSIONS: This method provides a simple, fast and unbiased measurement of different components of single-trial LEP responses. SIGNIFICANCE: This method is particularly desirable in several experimental conditions (e.g. drug studies, correlations with experimental variables, simultaneous EEG/fMRI and low signal-to-noise ratio data) and in clinical practice. The described multiple linear regression approach can be easily implemented for measuring evoked potentials in other sensory modalities.