Astrid Morzelewski

Dynamic BOLD functional connectivity in humans and its electrophysiological correlates

Neural oscillations subserve many human perceptual and cognitive operations. Accordingly, brain functional connectivity is not static in time, but fluctuates dynamically following the synchronization and desynchronization of neural populations. This dynamic functional connectivity has recently been demonstrated in spontaneous fluctuations of the Blood Oxygen Level-Dependent (BOLD) signal, measured with functional Magnetic Resonance Imaging (fMRI). We analyzed temporal fluctuations in BOLD connectivity and their electrophysiological correlates, by means of long (≈50 min) joint electroencephalographic (EEG) and fMRI recordings obtained from two populations: 15 awake subjects and 13 subjects undergoing vigilance transitions. We identified positive and negative correlations between EEG spectral power (extracted from electrodes covering different scalp regions) and fMRI BOLD connectivity in a network of 90 cortical and subcortical regions (with millimeter spatial resolution). In particular, increased alpha (8–12 Hz) and beta (15–30 Hz) power were related to decreased functional connectivity, whereas gamma (30–60 Hz) power correlated positively with BOLD connectivity between specific brain regions. These patterns were altered for subjects undergoing vigilance changes, with slower oscillations being correlated with functional connectivity increases. Dynamic BOLD functional connectivity was reflected in the fluctuations of graph theoretical indices of network structure, with changes in frontal and central alpha power correlating with average path length. Our results strongly suggest that fluctuations of BOLD functional connectivity have a neurophysiological origin. Positive correlations with gamma can be interpreted as facilitating increased BOLD connectivity needed to integrate brain regions for cognitive performance. Negative correlations with alpha suggest a temporary functional weakening of local and long-range connectivity, associated with an idling state.

Automatic sleep staging using fMRI functional connectivity data

Recent EEG-fMRI studies have shown that different stages of sleep are associated with changes in both brain activity and functional connectivity. These results raise the concern that lack of vigilance measures in resting state experiments may introduce confounds and contamination due to subjects falling asleep inside the scanner. In this study we present a method to perform automatic sleep staging using only fMRI functional connectivity data, thus providing vigilance information while circumventing the technical demands of simultaneous recording of EEG, the gold standard for sleep scoring. The features to classify are the linear correlation values between 20 cortical regions identified using independent component analysis and two regions in the bilateral thalamus. The method is based on the construction of binary support vector machine classifiers discriminating between all pairs of sleep stages and the subsequent combination of them into multiclass classifiers. Different multiclass schemes and kernels are explored. After parameter optimization through 5-fold cross validation we achieve accuracies over 0.8 in the binary problem with functional connectivities obtained for epochs as short as 60s. The multiclass classifier generalizes well to two independent datasets (accuracies over 0.8 in both sets) and can be efficiently applied to any dataset using a sliding window procedure. Modeling vigilance states in resting state analysis will avoid confounded inferences and facilitate the study of vigilance states themselves. We thus consider the method introduced in this study a novel and practical contribution for monitoring vigilance levels inside an MRI scanner without the need of extra recordings other than fMRI BOLD signals.

Breakdown of long-range temporal dependence in default mode and attention networks during deep sleep

The integration of segregated brain functional modules is a prerequisite for conscious awareness during wakeful rest. Here, we test the hypothesis that temporal integration, measured as long-term memory in the history of neural activity, is another important quality underlying conscious awareness. For this aim, we study the temporal memory of blood oxygen level-dependent signals across the human nonrapid eye movement sleep cycle. Results reveal that this property gradually decreases from wakefulness to deep nonrapid eye movement sleep and that such decreases affect areas identified with default mode and attention networks. Although blood oxygen level-dependent spontaneous fluctuations exhibit nontrivial spatial organization, even during deep sleep, they also display a decreased temporal complexity in specific brain regions. Conversely, this result suggests that long-range temporal dependence might be an attribute of the spontaneous conscious mentation performed during wakeful rest.

To wake or not to wake? The two-sided nature of the human K-complex

Sleep fosters performance but likewise renders creatures insensitive to environmental threat. The brain balances between sleep promotion and protection during light sleep. One associated electrophysiological hallmark is the K-complex (KC), the sleep promoting versus arousal inducing role of which is under debate. We examined 37 subjects using EEG-combined fMRI and found KC-associated positive BOLD signal changes in subcortical (brainstem, thalamus), sensory and motor, midline and regions which form part of the default mode network, and negative changes in the anterior insula. Connectivity analysis revealed the primary auditory cortex as the first region to be influenced during the KC and that midline regions activated successively from front to back in association with the sleep protecting part of the KC. Our findings support thalamic involvement in KC mediation and an association of KCs with subcortical arousal mechanisms: activations in sensory areas suggest the existence of low level information processing during KC limited by anterior insula disengagement suggesting a two-sided nature of the KC: it embodies an arousal with subsequent sleep-guarding counteraction that might on the one hand serve periodical monitoring of the environment with basic information processing and on the other hand protect the continuity of sleep and thus its restoring effect.

EEG microstates of wakefulness and NREM sleep

EEG-microstates exploit spatio-temporal EEG features to characterize the spontaneous EEG as a sequence of a finite number of quasi-stable scalp potential field maps. So far, EEG-microstates have been studied mainly in wakeful rest and are thought to correspond to functionally relevant brain-states. Four typical microstate maps have been identified and labeled arbitrarily with the letters A, B, C and D. We addressed the question whether EEG-microstate features are altered in different stages of NREM sleep compared to wakefulness. 32-channel EEG of 32 subjects in relaxed wakefulness and NREM sleep was analyzed using a clustering algorithm, identifying the most dominant amplitude topography maps typical of each vigilance state. Fitting back these maps into the sleep-scored EEG resulted in a temporal sequence of maps for each sleep stage. All 32 subjects reached sleep stage N2, 19 also N3, for at least 1 min and 45 s. As in wakeful rest we found four microstate maps to be optimal in all NREM sleep stages. The wake maps were highly similar to those described in the literature for wakefulness. The sleep stage specific map topographies of N1 and N3 sleep showed a variable but overall relatively high degree of spatial correlation to the wake maps (Mean: N1 92%; N3 87%). The N2 maps were the least similar to wake (mean: 83%). Mean duration, total time covered, global explained variance and transition probabilities per subject, map and sleep stage were very similar in wake and N1. In wake, N1 and N3, microstate map C was most dominant w.r.t. global explained variance and temporal presence (ratio total time), whereas in N2 microstate map B was most prominent. In N3, the mean duration of all microstate maps increased significantly, expressed also as an increase in transition probabilities of all maps to themselves in N3. This duration increase was partly--but not entirely--explained by the occurrence of slow waves in the EEG. The persistence of exactly four main microstate classes in all NREM sleep stages might speak in favor of an in principle maintained large scale spatial brain organization from wakeful rest to NREM sleep. In N1 and N3 sleep, despite spectral EEG differences, the microstate maps and characteristics were surprisingly close to wakefulness. This supports the notion that EEG microstates might reflect a large scale resting state network architecture similar to preserved fMRI resting state connectivity. We speculate that the incisive functional alterations which can be observed during the transition to deep sleep might be driven by changes in the level and timing of activity within this architecture.

Large-scale brain functional modularity is reflected in slow electroencephalographic rhythms across the human non-rapid eye movement sleep cycle

Large-scale brain functional networks (measured with functional magnetic resonance imaging, fMRI) are organized into separated but interacting modules, an architecture supporting the integration of distinct dynamical processes. In this work we study how the aforementioned modular architecture changes with the progressive loss of vigilance occurring in the descent to deep sleep and we examine the relationship between the ensuing slow electroencephalographic rhythms and large-scale network modularity as measured with fMRI. Graph theoretical methods are used to analyze functional connectivity graphs obtained from fifty-five participants at wakefulness, light and deep sleep. Network modularity (a measure of functional segregation) was found to increase during deeper sleep stages but not in light sleep. By endowing functional networks with dynamical properties, we found a direct link between increased electroencephalographic (EEG) delta power (1-4 Hz) and a breakdown of inter-modular connectivity. Both EEG slowing and increased network modularity were found to quickly decrease during awakenings from deep sleep to wakefulness, in a highly coordinated fashion. Studying the modular structure itself by means of a permutation test, we revealed different module memberships when deep sleep was compared to wakefulness. Analysis of node roles in the modular structure revealed an increase in the number of locally well-connected nodes and a decrease in the number of globally well-connected hubs, which hinders interactions between separated functional modules. Our results reveal a well-defined sequence of changes in brain modular organization occurring during the descent to sleep and establish a close parallel between modularity alterations in large-scale functional networks (accessible through whole brain fMRI recordings) and the slowing of scalp oscillations (visible on EEG). The observed re-arrangement of connectivity might play an important role in the processes underlying loss of vigilance and sensory awareness during deep sleep.

Narcoleptic Patients Show Fragmented EEG-Microstructure During Early NREM Sleep

Narcolepsy is a chronic disorder of the sleep-wake cycle with pathological shifts between sleep stages. These abrupt shifts are induced by a sleep-regulating flip-flop mechanism which is destabilized in narcolepsy without obvious alterations in EEG oscillations. Here, we focus on the question whether the pathology of narcolepsy is reflected in EEG microstate patterns. 30 channel awake and NREM sleep EEGs of 12 narcoleptic patients and 32 healthy subjects were analyzed. Fitting back the dominant amplitude topography maps into the EEG led to a temporal sequence of maps. Mean microstate duration, ratio total time (RTT), global explained variance (GEV) and transition probability of each map were compared between both groups. Nine patients reached N1, 5 N2 and only 4 N3. All healthy subjects reached at least N2, 19 also N3. Four dominant maps could be found during wakefulness and all NREM- sleep stages in healthy subjects. During N3, narcolepsy patients showed an additional fifth map. The mean microstate duration was significantly shorter in narcoleptic patients than controls, most prominent in deep sleep. Single maps’ GEV and RTT were also altered in narcolepsy. Being aware of the limitation of our low sample size, narcolepsy patients showed wake-like features during sleep as reflected in shorter microstate durations. These microstructural EEG alterations might reflect the intrusion of brain states characteristic of wakefulness into sleep and an instability of the sleep-regulating flip-flop mechanism resulting not only in pathological switches between REM- and NREM-sleep but also within NREM sleep itself, which may lead to a microstructural fragmentation of the EEG.

Corrigendum to “Automatic sleep staging using fMRI functional connectivity data” [Neuroimage 63/1 (2012) 63–72]

The authors regret that an accidental mislabelling of four subwith and without noise regression should read: [0.86/0.86,0.12/ jects in the two sleep + wake datasets of Fig. 3 resulted in a 0.12,0.01/0.01,0.00/0.00; 0.59/0.60,0.36/0.34,0.04/0.05,0.00/0.00; 0.48/ computation error of the results reported for the testing set #1. 0.49,0.06/0.06,0.42/0.43,0.02/0.02; 0.01/0.01,0.00/0.00,0.00/0.00,0.99/ Accuracy using a 2 minute window for this dataset is 75%/74%, 0.99]. with and without noise regression respectively. Confusion matrices The authors would like to apologise for any inconvenience caused. NeuroImage 81 (2013) 506