Search

BackgroundA common problem in resting-state neuroimaging studies is that subjects become drowsy or fall asleep. Although this could drastically affect neurophysiological measurements, such as magnetoencephalography (MEG), its specific impact remains understudied. We aimed to systematically investigate how often drowsiness is present during resting-state MEG recordings, and how the state changes alter quantitative estimates of oscillatory activity, functional connectivity, and network topology.MethodsAbout 8-min MEG recordings of 19 healthy subjects, split into ~13-s epochs, were scored for the presence of eyes-open (EO), alert eyes-closed (A-EC), or drowsy eyes-closed (D-EC) states. After projection to source-space, results of spectral, functional connectivity, and network analyses in 6 canonical frequency bands were compared between these states on a global and regional levels. Functional connectivity was analyzed using the phase lag index (PLI) and corrected amplitude envelope correlation (AECc), and network topology was analyzed using the minimum spanning tree (MST).ResultsDrowsiness was present in >55% of all epochs that did not fulfill the AASM criteria for sleep. There were clear differences in spectral results between the states (A-EC vs. D-EC) and conditions (EO vs. A-EC). The influence of state and condition was far less pronounced for connectivity analyses, with only minimal differences between D-EC and EO in the AECc in the delta band. There were no effects of drowsiness on any of the MST measures.ConclusionsDrowsiness during eyes-closed resting-state MEG recordings is present in the majority of epochs, despite the instructions to stay awake. This has considerable influence on spectral properties, but much less so on functional connectivity and network topology. These findings are important for interpreting the results of EEG/MEG studies using spectral analyses in neurological disease, where recordings should be evaluated for the presence of drowsiness. For connectivity analyses or studies on network topology, this seems of far less importance.

The oldest-old subjects represent the fastest growing segment of society and are at high risk for dementia with a prevalence of up to 40%. Lifestyle factors, such as lifelong participation in cognitive and leisure activities, may contribute to individual cognitive reserve and reduce the risk for cognitive impairments. However, the neural bases underlying cognitive functioning and cognitive reserve in this age range are still poorly understood. Here, we investigate spectral and functional connectivity features obtained from resting-state MEG recordings in a cohort of 35 cognitively normal (92.2 ± 1.8 years old, 19 women) and 11 cognitively impaired (90.9 ± 1.9 years old, 1 woman) oldest-old participants, in relation to cognitive traits and cognitive reserve. The latter was approximated with a self-reported scale on lifelong engagement in cognitively demanding activities. Cognitively impaired oldest-old participants had slower cortical rhythms in frontal, parietal and default mode network regions compared to the cognitively normal subjects. These alterations mainly concerned the theta and beta band and partially explained inter-subject variability of episodic memory scores. Moreover, a distinct spectral pattern characterized by higher relative power in the alpha band was specifically associated with higher cognitive reserve while taking into account the effect of age and education level. Finally, stronger functional connectivity in the alpha and beta band were weakly associated with better cognitive performances in the whole group of subjects, although functional connectivity effects were less prominent than the spectral ones. Our results shed new light on the neural underpinnings of cognitive functioning in the oldest-old population and indicate that cognitive performance and cognitive reserve may have distinct spectral electrophysiological substrates.

Despite advances in the field of dynamic connectivity, fixed sliding window approaches for the detection of fluctuations in functional connectivity are still widely used. The use of conventional connectivity metrics in conjunction with a fixed sliding window comes with the arbitrariness of the chosen window lengths. In this paper we use multivariate autoregressive and neural mass models with a priori defined ground truths to systematically analyze the sensitivity of conventional metrics in combination with different window lengths to detect genuine fluctuations in connectivity for various underlying state durations. Metrics of interest are the coherence, imaginary coherence, phase lag index, phase locking value and the amplitude envelope correlation. We performed analysis for two nodes and at the network level. We demonstrate that these metrics show indeed higher variability for genuine temporal fluctuations in connectivity compared to a static connectivity state superimposed by noise. Overall, the error of the connectivity estimates themselves decreases for longer state durations (order of seconds), while correlations of the connectivity fluctuations with the ground truth was higher for longer state durations. In general, metrics, in combination with a sliding window, perform poorly for very short state durations. Increasing the SNR of the system only leads to a moderate improvement. In addition, at the network level, only longer window widths were sufficient to detect plausible resting state networks that matched the underlying ground truth, especially for the phase locking value, amplitude envelope correlation and coherence. The length of these longer window widths did not necessarily correspond to the underlying state durations. For short window widths resting state network connectivity patterns could not be retrieved. We conclude that fixed sliding window approaches for connectivity can detect modulations of connectivity, but mostly if the underlying dynamics operate on moderate to slow timescales. In practice, this can be a drawback, as state durations can vary significantly in empirical data.

There is general agreement that the neuropathological processes leading to Alzheimer’s disease (AD) begin decades before the clinical onset. In order to detect early topological changes, we applied functional connectivity and network analysis to magnetoencephalographic (MEG) data obtained from 16 patients with amnestic Mild Cognitive Impairment (aMCI), a prodromal stage of AD, and 16 matched healthy control (HCs). Significant differences between the two groups were found in the theta band, which is associated with memory processes, in both temporal poles (TPs). In aMCI, the degree and betweenness centrality (BC) were lower in the left superior TP, whereas in the right middle TP the BC was higher. A statistically significant negative linear correlation was found between the BC of the left superior TP and a delayed recall score, a sensitive marker of the “hippocampal memory” deficit in early AD. Our results suggest that the TPs, which are involved early in AD pathology and belong to the memory circuitry, have an altered role in the functional network in aMCI.

Objective: Epilepsy surgery results in seizure freedom in the majority of drug-resistant patients. To improve surgery outcome we studied whether MEG metrics combined with machine learning can improve localization of the epileptogenic zone, thereby enhancing the chance of seizure freedom.Methods: Presurgical interictal MEG recordings of 94 patients (64 seizure-free >1y post-surgery) were analyzed to extract four metrics in source space: delta power, low-to-high-frequency power ratio, functional connectivity (phase lag index), and minimum spanning tree betweenness centrality. At the group level, we estimated the overlap of the resection area with the five highest values for each metric and determined whether this overlap differed between surgery outcomes. At the individual level, those metrics were used in machine learning classifiers (linear support vector machine (SVM) and random forest) to distinguish between resection and non-resection areas and between surgery outcome groups.Results: The highest values, for all metrics, overlapped with the resection area in more than half of the patients, but the overlap did not differ between surgery outcome groups. The classifiers distinguished the resection areas from non-resection areas with 59.94% accuracy (95% confidence interval: 59.67–60.22%) for SVM and 60.34% (59.98–60.71%) for random forest, but could not differentiate seizure-free from not seizure-free patients [43.77% accuracy (42.08–45.45%) for SVM and 49.03% (47.25–50.82%) for random forest].Significance: All four metrics localized the resection area but did not distinguish between surgery outcome groups, demonstrating that metrics derived from interictal MEG correspond to expert consensus based on several presurgical evaluation modalities, but do not yet localize the epileptogenic zone. Metrics should be improved such that they correspond to the resection area in seizure-free patients but not in patients with persistent seizures. It is important to test such localization strategies at an individual level, for example by using machine learning or individualized models, since surgery is individually tailored.

Subjects with mild cognitive impairment (MCI) have an increased risk of developing Alzheimer’s disease (AD), and their functional brain networks are presumably already altered. To test this hypothesis, we compared magnetoencephalography (MEG) eyes-closed resting-state recordings from 29 MCI subjects and 29 healthy elderly subjects in the present exploratory study. Functional connectivity in different frequency bands was assessed with the phase lag index (PLI) in source space. Normalized weighted clustering coefficient (normalized Cw) and path length (normalized Lw), as well as network measures derived from the minimum spanning tree [MST; i.e., betweenness centrality (BC) and node degree], were calculated. First, we found altered PLI values in the lower and upper alpha bands in MCI patients compared to controls. Thereafter, we explored network differences in these frequency bands. Normalized Cw and Lw did not differ between the groups, whereas BC and node degree of the MST differed, although these differences did not survive correction for multiple testing using the False Discovery Rate (FDR). As an exploratory study, we may conclude that: (1) the increases and decreases observed in PLI values in lower and upper alpha bands in MCI patients may be interpreted as a dual pattern of disconnection and aberrant functioning; (2) network measures are in line with connectivity findings, indicating a lower efficiency of the brain networks in MCI patients; (3) the MST centrality measures are more sensitive to detect subtle differences in the functional brain networks in MCI than traditional graph theoretical metrics.

Synaptic loss is an early pathological finding in Alzheimer’s disease (AD) and correlates with memory impairment. Changes in macroscopic brain activity measured with electro- and magnetoencephalography (EEG and MEG) in AD indicate synaptic changes and may therefore serve as markers of intervention effects in clinical trials. EEG peak frequency and functional networks have shown, in addition to improved memory performance, to be sensitive to detect an intervention effect in mild AD patients of the medical food Souvenaid containing the specific nutrient combination Fortasyn® Connect, which is designed to enhance synapse formation and function. Here, we explore the value of MEG, with higher spatial resolution than EEG, in identifying intervention effects of the nutrient combination by comparing MEG spectral measures, functional connectivity and – networks between an intervention and a control group. Quantitative markers describing spectral properties, functional connectivity and graph theoretical aspects of MEG from the exploratory 24-week, double blind, randomized controlled Souvenir II MEG sub-study (NTR1975, http://www.trialregister.nl) in drug-naïve patients with mild AD were compared between a test group (n=27), receiving Souvenaid, and a control group (n=28), receiving an isocaloric control product. The groups were unbalanced at screening with respect to Mini-Mental State Examination (MMSE). Peak frequencies of MEG were compared to EEG peak frequencies, recorded in the same patients at similar time points, were compared with respect to sensitivity to intervention effects. No consistent statistically significant intervention effects were detected. In addition, we found no difference in sensitivity between MEG and EEG peak frequency. This exploratory study could not unequivocally establish the value of MEG in detecting interventional effects on brain activity, possibly due to small sample size and unbalanced study groups. We found no indication that the difference could be attributed to a lack of sensitivity of MEG compared to EEG. MEG in randomized controlled trials is feasible but its value to disclose intervention effects of Souvenaid in mild AD patients needs to be studied further.

A network of 32 or 64 connected neural masses, each representing a large population of interacting excitatory and inhibitory neurons and generating an EEG / MEG like output signal, was used to demonstrate how an interaction between dynamics and connectivity might explain the emergence of complex network features, in particular modularity. Network evolution was modeled by two processes: (i) synchronization dependent plasticity (SDP) and (ii) growth dependent plasticity (GDP). In the case of SDP, connections between neural masses were strengthened when they were strongly synchronized, and were weakened when they were not. GDP was modeled as a homeostatic process with random, distance dependent outgrowth of new connections between neural masses. GDP alone resulted in stable networks with distance dependent connection strengths, typical small-world features, but no degree correlations and only weak modularity. SDP applied to random networks induced clustering, but no clear modules. Stronger modularity evolved only through an interaction of SDP and GDP, with the number and size of the modules depending on the relative strength of both processes, as well as on the size of the network. Lesioning part of the network, after a stable state was achieved, resulted in a temporary disruption of the network structure. The model gives a possible scenario to explain how modularity can arise in developing brain networks, and makes predictions about the time course of network changes during development and following acute lesions.